The growing demand for ride-hailing services has led to an increasing need for accurate taxi demand prediction. Existing systems are limited to specific regions, lacking generalizability to unseen areas. This paper presents a novel taxi demand prediction system, harnessing the strengths of multiview graph neural networks to capture spatial-temporal dependencies and patterns in urban environments. Additionally, the proposed systemCROSS-Netemploys a spatially transferable approach, enabling it to train a model that can be deployed to previously unseen regions. To achieve this, the framework incorporates the power of a Variational Autoencoder to disentangle the input features into region-specific and region-agnostic components. The region-agnostic features facilitate cross-region taxi demand predictions, allowing the model to generalize well across different urban areas. Experimental results demonstrate the effectiveness ofCROSS-Netin accurately forecasting taxi demand, even in previously unobserved regions, thus showcasing its potential for optimizing taxi services and improving transportation efficiency on a broader scale.

Modern mobility platforms have stored vast streams of GPS trajectories, temporal metadata, free-form textual notes, and other unstructured data. Privacy statutes such as the GDPR require that any individual's contribution be unlearned on demand, yet retraining deep models from scratch for every request is untenable. We introduce MobText-SISA, a scalable machine-unlearning framework that extends Sharded, Isolated, Sliced, and Aggregated (SISA) training to heterogeneous spatio-temporal data. MobText-SISA first embeds each trip's numerical and linguistic features into a shared latent space, then employs similarity-aware clustering to distribute samples across shards so that future deletions touch only a single constituent model while preserving inter-shard diversity. Each shard is trained incrementally; at inference time, constituent predictions are aggregated to yield the output. Deletion requests trigger retraining solely of the affected shard from its last valid checkpoint, guaranteeing exact unlearning. Experiments on a ten-month real-world mobility log demonstrate that MobText-SISA (i) sustains baseline predictive accuracy, and (ii) consistently outperforms random sharding in both error and convergence speed. These results establish MobText-SISA as a practical foundation for privacy-compliant analytics on multimodal mobility data at urban scale.

This paper presents a privacy-preserving monitoring system for elderly individuals living alone, utilizing collaborative lightweight Large Language Model (LLM) agents deployed on edge devices. The system integrates non-invasive sensors with a novel three-agent architecture: an activity recognition agent processes sensor data, an hourly summarization agent generates intermediate reports in 3-hour segments, and a daily summarization agent produces comprehensive summaries. Our evaluation on real-world data from two elderly households demonstrates that the system achieves 95.8% accuracy in activity recognition and generates natural language summaries comparable to GPT-4o, while maintaining privacy through local processing on affordable Raspberry Pi hardware. The results indicate that our approach effectively balances monitoring accuracy, summary quality, and practical deployment constraints, making it suitable for widespread adoption in elderly care applications.

This paper presents a novel autonomous navigation framework that integrates Large Language Models (LLMs) with multimodal sensor fusion to enable dynamic obstacle avoidance and human-aware path planning in diverse environments. The proposed system leverages an FPGA-accelerated fusion pipeline, combining LiDAR and vision data for real-time perception. A Hungarian algorithm-based object matching technique ensures robust tracking, while a bird's-eye view (BEV) representation enhances spatial reasoning and occlusion handling. The fused sensory inputs are processed by a fine-tuned LLM, which contextualizes pedestrian behavior and environmental constraints to generate adaptive, human-centric navigation strategies. Unlike traditional rule-based methods, LLMs provide generalization capabilities to novel scenarios, significantly improving interaction with vulnerable pedestrians such as children, elderly individuals, and wheelchair users. Extensive evaluations in both simulated and real-world scenarios confirm the system's ability to reduce collisions and enhance navigation efficiency in high-density environments. By bridging semantic reasoning and robotic control, this work lays the foundation for next-generation intelligent navigation systems that are both safety-aware and scalable across autonomous platforms.

This paper introduces a novel approach for coordinated evacuation route planning during disasters, focusing on guiding groups of individuals-such as families or neigh-bors-who are spatially separated but socially connected. Recognizing the importance of group cohesion during evacuations, we designed and implemented a prototype system, which was validated through a real-world evacuation drill conducted in Kobe City during the 30th memorial year of the Great HanshinEarthquake. We introduce and formulate a new problem, termed the ‘Evacuation Routing Problem’ (ERP), using integer linear programming (ILP). Our system recommends that group members rendezvous at specific points along their evacuation routes. The core idea is to model each evacuation route as a tree on a predefined road network, where the leaves correspond to evacuees (group members) and the root corresponds to the shelter. This problem can be viewed as a variant of the Steiner Tree Problem, augmented with an additional constraint to ensure that members of a group meet at an optimal rendezvous node while en route to the shelter. To validate our method, we conducted a pilot experiment in Kobe City, Japan. We developed a prototype system featuring a web-based interface that generates high-quality evacuation routes derived from ILP solutions. The experiment involved two family groups, and the developed system navigated a real road network that included various points of interest such as parks, shopping malls, offices, and schools. The results showed that one family successfully gathered at the proposed rendezvous point and evacuated together, while the other group was unable to meet due to a suggested detour. Drawing from these experimental outcomes and user feedback, we analyze system limitations and discuss potential improvements from a user-centric perspective.

Accurate indoor positioning is essential for mobile computing, human-computer interaction, and next-generation smart environments, enabling applications in indoor navigation, augmented reality, personalized services, healthcare, and emergency response. Cellular signal fingerprinting has emerged as a widely adopted solution, with deep learning models achieving state-of-the-art performance. However, existing approaches face critical deployment challenges, including labor-intensive fingerprinting, sparse reference points, and missing RSS values caused by environmental interference, hardware variability, and dynamic signal fluctuations. These limitations hinder their scalability, adaptability, and real-world usability in complex indoor environments. To address these challenges, we present GPT2Loc a novel indoor localization framework that integrates LLM with label-aware contrastive learning, improving accuracy while reducing reliance on extensive fingerprinting. LLMs effectively extract meaningful spatial features from incomplete and noisy RSS data, enabling robust localization even in sparsely finger-printed areas. Our label-aware contrastive learning approach further enhances generalization by aligning latent representations with spatial relationships, allowing GPT2Loc to interpolate user locations in unseen areas and mitigate signal inconsistencies.

Effective crowd tracking at railway station platforms is essential for ensuring passenger safety and optimizing pedestrian flow, particularly in high-density urban transit hubs. However, traditional tracking methods, such as object detection and multi-object tracking, face limitations in congested environments due to severe occlusions and overlapping individuals. This paper proposes a novel approach for modeling pedestrian flow on train station platforms by coupling deep learning-based motion analysis with crowd simulation. In the proposed method, we utilize RAFT, a state-of-the-art optical flow model, to extract pixel-level motion vectors, which are clustered to identify human movement patterns. These motion data are mapped onto a calibrated 2D platform model, providing a top-down representation of pedestrian trajectories. To simulate realistic crowd dynamics, Unity's NavMesh is employed alongside an enhanced Simulated Annealing approach to generate high-accuracy origin-destination (OD) data. This is a new digital twin concept where the analysis from the vision in the real world is projected onto the virtual world model to simulate and reproduce the pedestrian flows. The proposed method was evaluated using synthetic crowd simulation data, demonstrating high accuracy in destination estimation. The experimental results indicate that the OD estimation outperforms conventional approaches, with error rates reduced to half of those observed in YOLOv8x-based tracking systems. These findings suggest that the integration of optical flow-based motion analysis with digital twin simulation can significantly enhance crowd monitoring and congestion management in railway stations.

This doctoral research proposes a framework for developing sociallyaware robot navigation systems by integrating the cognitive capabilities of Large Language Models (LLMs) with the demands of real-world Human-Robot Interaction (HRI).Our work follows a four-stage plan that systematically addresses the challenges of applying LLMs to time-sensitive, safety-critical tasks.This paper details the completion of the first two stages, wherein we developed and evaluated a foundational navigation model.Our system features a meticulously designed multimodal fusion pipeline that integrates LiDAR and camera data, processed by a YOLO model and a Hungarian algorithm for semantic association, providing rich, contextual input to the LLM.Through knowledge distillation and fine-tuning on data from a custom simulator, our model demonstrates robust spatial reasoning and superior performance in low-frequency decision-making scenarios compared to traditional reinforcement learning methods.We successfully validated this foundational model and identified its inference latency as a key challenge.These results establish a solid basis for our future work.This includes developing a "brain-cerebellum" hybrid architecture for real-time performance and exploring multi-robot social compliance.This research contributes to HRI by creating more predictable and trustworthy robots, and to the LLM field by investigating the symbol grounding problem through embodied intelligence.

Due to global climate change, developing robust and accurate disaster prediction systems has become essential worldwide, as severe disasters are increasingly occurring even in previously safe regions.Two key technical challenges in disaster prediction are data scarcity and regional heterogeneity.Disasters are rare and extreme, resulting in limited data, which hinders data-driven approaches like machine learning (ML).Moreover, hydro-meteorological disasters depend on region-specific factors such as rainfall patterns, soil, vegetation, and infrastructure.These, combined with dynamic environmental changes from earthquakes, climate, and urbanization, complicate the generalization of models across time and space.We propose a self-adaptive physics-hybrid disaster prediction system that autonomously adapts to diverse and evolving environments to address this.Our system consists of: (1) a physics-hybrid disaster prediction model integrating physics-based hydro-meteorological simulation and multi-modal ML models, and (2) an environmental adaptation utilizing test-time adaptation (TTA) to update parameters without labeled disaster data(i.e., disaster event data).Our physics-hybrid model is anchored in the physics model to balance stability and adaptability, leveraging strengths from both physics and ML models.Furthermore, the strong inductive bias of the physics model mitigates overfitting and catastrophic forgetting during TTA, enabling robust adaptation to unseen regions.

Electric scooters (e-scooters) are reshaping urban mobility, providing eco-friendly and cost-effective transport.However, the surge in their usage in mixed-traffic environments poses significant safety risks, especially for vulnerable road users (VRUs).This paper presents a lightweight vision-based safety assistance system optimized for real-time inference on edge devices.The core modules include semantic segmentation enhanced by semi-supervised learning, accurate bird's-eye view (BEV) transformation, real-time motion prediction, and adaptive path planning.Extensive evaluations demonstrate high segmentation accuracy and low-latency execution suitable for resource-constrained hardware.Future research directions include knowledge distillation for model adaptability, cooperative multi-scooter perception, intersection-based AI infrastructure, and the application of large language models (LLMs) to optimize urban traffic flows.

Accurate monitoring of indoor environments via wireless sensing is essential for applications such as adaptive crowd flow control, energy management in smart buildings, and rapid response in healthcare and emergency scenarios.Traditional vision-based methods suffer from lighting issues, occlusion, and privacy concerns, while RSSI-based approaches are vulnerable to multipath fading and device inconsistencies, and CSI-based systems lack interoperability.We present a device-free and device-independent framework that continuously records round-trip signal delays as a rich time series, supporting both occupancy estimation and spatial localization without specialized hardware.A deep feature extraction strategy suppresses environmental noise and non-line-of-sight distortions, delivering accurate real-time performance on commodity WiFi equipment even in cluttered spaces.This approach establishes a foundation for ubiquitous privacy-preserving indoor environment monitoring with potential extensions to multimodal sensor fusion and integration with next-generation wireless standards.

Electric scooters (e-scooters) are increasingly used in urban areas but face serious safety concerns, especially in mixed traffic environments. This paper presents a real-time vision-based safety assistance system designed for e-scooters and optimized for edge deployment. The system uses a lightweight segmentation model trained via semi-supervised learning to detect sidewalks and road areas from non-vehicle perspectives. Dynamic objects such as pedestrians and cyclists are tracked using Kalman filtering, and potential risks are assessed in real time. A hybrid path planning module uses A* and artificial potential fields to recommend safe forward paths. Experimental results show that the system achieves real-time performance on Jetson Orin NX while improving perception accuracy and navigation safety in sidewalk environments.

Abstract Accurate human counting in indoor environments is essential for optimizing people-centric applications, such as crowd management, disaster response, and monitoring in settings like shopping malls and healthcare facilities. Traditional vision approaches face challenges with poor lighting conditions and raise privacy concerns. WiFi-based solutions enable device-free human counting by detecting disruptions in wireless signals caused by human presence. However, methods using received signal strength indicator are unreliable due to physical obstructions, multipath fading, radio interference, and fluctuating access point power. While WiFi channel state information-based systems are more sensitive to environmental changes, they lack standardization, limiting their practicality. To overcome these limitations, this paper presents Time4Count, an innovative device-free indoor human counting system that leverages round trip time measurements to achieve high accuracy and scalability. Time4Count capitalizes on human-induced fluctuations in signal propagation time to accurately estimate the number of individuals in a space. By employing a multivariate transformer-based feature extraction method, the system effectively mitigates non-line-of-sight errors and signal distortions, ensuring robust performance even in cluttered indoor environments. Additionally, Time4Count integrates spatial discretization and multi-label classification techniques, enabling it to count an unlimited number of individuals in real-time. The system was rigorously evaluated in two realistic, cluttered environments using commodity hardware, involving up to 15 participants. Experimental results reveal that Time4Count achieves an high counting accuracy of 92.7%. To our knowledge, Time4Count is the first RTT-based indoor counting system, providing a precise solution for indoor monitoring. Implementation is available at: https://github.com/mclab-osaka/time4count .

This paper proposes a method for estimating household activities based only on the cumulative power consumption data obtained from the HEMS home distribution board every 30 min. The proposed method predicts the activity of each 30 min timeslot from the eight activity labels; household-level waking-up, household-level going-to-bed, room-level waking-up, room-level going-to-bed, cooking, laundry, dishwashing, and bathing. For the prediction, we first identify the branch circuit that is strongly correlated with each activity label and detect the turn-on/off of home appliances on the circuit to detect those activities. We also incorporate machine learning for estimating the other activities based on the circuit’s time series of power consumption. Furthermore, to cope with the difference among households, we apply transfer learning to the constructed model. In collaboration with a Japanese home builder, we conducted an experiment on five households using their HEMS data. In parallel, we obtained verifiable activity labels as our ground truth by the installation of specialized sensors in the respective homes. Under a ±30 min tolerance (i.e. allowing a prediction in the immediately preceding or following half-hour slot), our model achieved an average F1 score of 0.689 across all activities. We also confirmed that transfer learning improved the F1 score of each activity recognition and achieved an average improvement of 0.260 in household-level waking-up, household-level going-to-bed, room-level waking-up, room-level going-to-bed, and bathing activities.

The increasing need for pedestrian tracking within public and commercial domains necessitates innovative solutions that respect privacy concerns. Traditional surveillance methods employing RGB video footage for the monitoring of pedestrian movements have raised significant privacy issues, attributed to their capability to easily identify individuals. This paper shifts focus towards the utilization of 3D point cloud data as a less intrusive alternative for pedestrian monitoring, which inherently minimizes the risk of individual identification. Notwithstanding, traditional tracking methodologies frequently encounter difficulties in environments characterized by high pedestrian density, where occlusions caused by individuals obscuring one another from the sensors’ perspective are commonplace. This research introduces an advanced approach leveraging 3D point cloud data to address the challenges inherent in pedestrian tracking. A particular obstacle addressed is the frequent obstruction caused by pedestrians, who may be accompanied by objects such as baby strollers and/or luggage, thus complicating their detection and tracking when positioned behind others or in close proximity to the sensors. Such scenarios significantly undermine the efficacy of pedestrian segmentation and tracking, particularly in the context of Kalman-Filter-based methodologies. To mitigate these challenges, the proposed solution incorporates a method to spatially enhance the missing segments within the framework of Kalman-Filter-based multi-object tracking. The effectiveness of this approach has been rigorously assessed through the application of 3D point cloud data derived from diverse real-world settings, including laboratory environments, shopping mall entrances, and a densely populated urban area in Kita-ku, Osaka City, during a festival event.

Accurate taxi-demand prediction is essential for optimizing taxi operations and enhancing urban transportation services. However, using customers' data in these systems raises significant privacy and security concerns. Traditional federated learning addresses some privacy issues by enabling model training without direct data exchange but often struggles with accuracy due to varying data distributions across different regions or service providers. In this paper, we propose CC-Net: a novel approach using collaborative learning enhanced with contrastive learning for taxi-demand prediction. Our method ensures high performance by enabling multiple parties to collaboratively train a demand-prediction model through hierarchical federated learning. In this approach, similar parties are clustered together, and federated learning is applied within each cluster. The similarity is defined without data exchange, ensuring privacy and security. We evaluated our approach using real-world data from five taxi service providers in Japan over fourteen months. The results demonstrate that CC-Net maintains the privacy of customers' data while improving prediction accuracy by at least 2.2% compared to existing techniques.

This paper presents a novel approach to smart home simulation by integrating Large Language Models (LLMs) for generating human daily schedules and activities of smart home simulator agents. The use of LLMs aims to reduce the complexity of user settings and increase the variety and generality of generated activities. This study explores the potential of LLMs to simulate human-like activity-generating by exploiting their experiential knowledge and adaptability. In addition, this paper discusses fine-tuning techniques for LLMs to optimize their performance within a simulator. By utilizing LoRA and task-specific fine-tuning datasets, while maintaining the ability to generate proper activities, we achieve as much as 4.3% performance improvement about the number of queries. Overall, the integration of LLMs into smart home simulators offers promising opportunities for enhancing the intelligence and adaptability of virtual agents in smart home environments.

Indoor localization has become a critical area of research with increasing relevance in applications. While numerous technologies have been explored, WiFi-based fingerprinting solutions using Received Signal Strength Indicators from multiple access points have garnered substantial attention due to the ubiquity of WiFi networks. However, these methods often encounter challenges, including fluctuation of access points and their noise signal measurements. Furthermore, they struggle to adapt effectively to cluttered or dynamically changing environments. In this paper, we introduce GraphLy: a Graph Neural Network-based model explicitly designed to tackle these challenges. GraphLy captures complex spatial relationships between different locations and adapts to environmental complexities and clutter, offering a robust solution for indoor localization. Our experiments demonstrate that GraphLy outperforms state-of-the-art WiFi-based localization techniques in two cluttered and propagation complex testbeds. In particular, we achieved a performance improvement of at least 37% and 65% in the two environments, respectively. These findings underscore the potential of GraphLy to enhance indoor localization accuracy and reliability for various real-world applications.

WiFi-based human sensing has exhibited remarkable potential to analyze user behaviors in a non-intrusive and device-free manner. However, most previous works focus on single-user sensing, which has limited practicability in scenarios involving multiple users. In this paper, we introduce YOSO a novel system employing a multi-label multi-view Transformer-based architecture to address these issues, enabling simultaneous localization and activity recognition. By applying advanced preprocessing techniques and utilizing the Transformer's self-attention mechanism, our system effectively learns high-dimensional representations of human activities and locations from CSI data. This approach overcomes traditional sequential data processing limitations, offering precise activity recognition and localization in multiperson environments. Our experimental results showcase superior performance in both localization and activity recognition tasks, surpassing existing methods. The real-time processing capability of our system paves the way for applications in smart environments, security, and healthcare monitoring, providing an efficient tool for situational awareness and advancing wireless sensing technology.

Poster: This study focuses on automatically identifying and classifying objects within indoor environments. Traditional methods struggle with this task due to the high cost of manually labeling each object and the inherent ambiguity of written descriptions. To overcome these limitations, we propose a novel instance segmentation approach that utilizes a visual-language model. This system is trained on extensive indoor environment data, including detailed point clouds (3D representations) and RGB images, readily collected by modern smartphone sensors. By eliminating the need for pre-defined labels, the system allows users to search for items using natural language. We evaluate the system's effectiveness through experiments analyzing performance metrics like accuracy and efficiency with public datasets, demonstrating its practicality and potential usefulness.

Precise disaster prediction plays a critical role in saving lives. Traditional landslide prediction methods have predominantly relied on deep neural networks such as CNNs and LSTMs. However, these methods face two main challenges. The first is the class imbalance issue, as landslides are rare, disrupting the training process. The second challenge stems from decentralized data management, with variations in volume and characteristics across regions. Typically, local municipalities manage disaster data within their regions, and sharing or migrating this data is not straightforward. This paper presents SlideSafe: a novel landslide prediction system that combines spatio-temporal contrastive learning and collaborative learning11Our implementation is available here (https://github.com/mclab-osaka/slidesafe). It begins by training a contrastive learning model to extract meaningful representations of land characteristics in each region. Subsequently, these trained models are merged among regions with similar characteristics, leveraging federated learning. The federated models are then fine-tuned and customized for the landslide event prediction using the data specific to each region. Experimental results indicate that the proposed system achieves higher precision under 100% recall compared to state-of-the-art federated learning methods, which are often adversely affected by non-iid data and data scarcity.

Hybrid-metaverses, integrating physical and virtual spaces, face a critical challenge in managing shared 3D object quality across multiple users with diverse preferences and limited network resources. This paper addresses the problem of allocating limited bandwidth for transmitting point cloud representations while maximizing overall user satisfaction. We propose a distributed optimization method that dynamically adjusts 3D object quality based on contextual importance, available resources, and user preferences. Our approach uses Input Convex Neural Networks (ICNN) to model user utility functions and employs the Alternating Direction Method of Multipliers (ADMM) for distributed optimization. Key advantages include scalability, adaptability, and improved quality of experience. Evaluation using open dataset demonstrates significant improvements in user satisfaction and resource utilization compared to baseline approaches. Our method achieves 93-94.6% accuracy in modeling user utility and shows up to 60% faster convergence for scenarios with 30 users, contributing to the balance between high-fidelity representation and efficient data management in hybrid-metaverses.

Locating the persons moving through an environment without the necessity of them being equipped with special devices has become vital for many applications, including security, IoT, healthcare, etc. Existing device-free indoor localization systems commonly rely on the utilization of received signal strength indicator (RSSI) and WiFi channel state information (CSI) techniques. However, the accuracy of RSSI is adversely affected by environmental factors like multipath interference and fading. Additionally, the lack of standardization in CSI necessitates the use of specialized hardware and software. In this article, we present TimeSense, a deep learning-based multiperson device-free indoor localization system that addresses these challenges. TimeSense leverages Time of Flight information acquired by the fine-time measurement protocol of IEEE 802.11-2016 standard. Specifically, the measured round trip time between the transmitter and receiver is influenced by the dynamic changes in the environment induced by human presence. TimeSense effectively detects this anomalous behavior using a stacked denoising auto-encoder model, thereby estimating the user’s location. The system incorporates a probabilistic approach on top of the deep learning model to ensure seamless tracking of the users. The evaluation of TimeSense in two realistic environments demonstrates its efficacy, achieving a median localization accuracy of 1.57 and 2.65 meters. This surpasses the performance of state-of-the-art techniques by 49% and 103% in the two testbeds.

Tracking pedestrian flow in large public areas is vital, yet ensuring privacy is paramount. Traditional visual-based tracking systems are raising concerns for potentially obtaining persistent and permanent identifiers that can compromise individual identities. Moreover, in areas such as the vicinity of restrooms, any form of data acquisition capturing human behavior should be refrained from, making it also crucial to appropriately address and complement these blind spots for a comprehensive analysis of pedestrian movement in the entire area. In this paper, we present our pedestrian tracking algorithm using distributed 3D LiDARs (Light Detection and Ranging), which capture pedestrians as 3D point clouds, omitting identifiable features. Our system bridges blind spots by leveraging historical movement data and 3D point cloud features, complemented by a generative diffusion model to predict trajectories in unseen areas. In a large-scale testbed with 70 LiDARs, the system achieved a 0.98 F-measure, highlighting its potential as a leading privacy-preserving tracking solution.

The growing demand for ride-hailing services has led to an increasing need for accurate taxi demand prediction. Existing systems are limited to specific regions, lacking generalizability to unseen areas. This paper presents a novel taxi demand forecasting system that leverages a graph neural network to capture spatial dependencies and patterns in urban environments. Additionally, the proposed system employs a region-neutral approach, enabling it to train a model that can be applied to any region, including unseen regions. To achieve this, the framework incorporates the power of Variational Autoencoder to disentangle the input features into region-specific and region-neutral components. The region-neutral features facilitate cross-region taxi demand predictions, allowing the model to generalize well across different urban areas. Experimental results demonstrate the effectiveness of the proposed system in accurately forecasting taxi demand, even in previously unobserved regions, thus showcasing its potential for optimizing taxi services and improving transportation efficiency on a broader scale.

This study proposes an adaptive message delivery mechanism for a broker in the WMO Information System 2.0 (WIS2.0), a next-generation world weather loT data exchange system, to satisfy the time constraints (deadlines) of the messages requested by subscribers in the system. Since WIS2.0 is a worldwide, heterogeneous system, each broker has to handle a variety of requests by each subscriber to receive messages (data) with different time constraints and sizes. Therefore, it needs to control the timing of message delivery to subscribers, considering publishers' delivery timing, subscribers' processing performance, network bandwidth, message types, data volume, and so on. To this end, the proposed method takes a reinforcement learning-based approach where the feasible timing to transmit messages is learned by monitoring the message arrival patterns from the publishers, message ACK delay from the subscriber, and the sensing window size, all of which are easily observable at the broker. Experiments using a network simulator showed that under the best-case scenario, the proposed method could achieve 10% higher message goodput compared to DWEDF-RL, the state-of-the-art RL-based packet scheduling method tailored to the pub/sub model.

WiFi localization systems have become increasingly popular owing to the ubiquity of WiFi-enabled devices. However, traditional WiFi networks were not originally designed for localization purposes, and changes to the Access Point locations can significantly drop localization performance. Moreover, these systems are unable to generalize models trained in one building to other buildings due to dependencies on building characteristics and their WiFi configurations. To address this challenge, we propose GlobLoc: the first global indoor positioning system that can function in any environment in a plug-and-play manner without requiring recalibration. Towards this end, we introduce a novel concept called the “virtual environment” in which users are tracked rather than the physical counterpart. This enables the localization model trained on one environment to function in another, regardless of changes occurring in the network environment, as they share the same virtual environment. Our experiments demonstrate that GlobLoc outperforms existing state-of-the-art propagation- and fingerprinting-based systems by at least 65%. This superior accuracy is due to its robustness to changes in the WiFi environment and its adaptability to different building configurations.

Augmented reality (AR) devices have gained a lot of attention in recent years due to their ability to enhance people’s abilities through 2D/3D spatial sensing and recognition functions. RGB cameras are most often used as sensors in this type of spatial recognition, and a particularly important task is the detection and tracking of objects and people in physical space. However, the camera positions and orientations on AR devices such as smartphones and smart glasses, frequently change due to the user wearing them on their head, leading to non-linear and complex motion in the video frames and reducing the accuracy of tracking people. To address this issue, the proposed method combines person re-identification based on deep metric learning with trajectory prediction to estimate the person’s sequential positions in 3D space around the camera. The experimental result shows 95.45% accuracy with our dataset.

The growing demand for ride-hailing services has led to an increasing need for accurate taxi demand prediction. However, the use of real passenger data to train predictive models raises serious privacy concerns. To address this challenge, we present a privacy-preserving taxi demand prediction system that employs a generative model to synthesize synthetic trajectory data, preserving privacy while retaining the statistical properties of the original data. The system also overcomes the challenge of location dependence of latitude-longitude values by encoding the representation into region-independent space, making it more general and applicable to different geographical areas. The system was evaluated on real-world data collected from a major taxi service provider in Japan over a period of six months. The results showed that the system can effectively defend against 98% of all attempted attacks on passenger data and against 60% of state-of-the-art attacks on the learning-based prediction models. Additionally, the proposed system ensures the prediction performance, with a barely noticeable decrease of 2.9% compared to using the original data.

The increasing use of trajectory data and machine learning has advanced our understanding of human behavior. Specifically, in situations such as team sports, in which multiple players form a group and interact with each other, predicting performance using their trajectory data as input has been attracting attention. However, understanding the role and contributions of individuals in the group by a deep understanding of the whole trajectory has not yet been well-investigated. In this study, we propose a method to quantitatively evaluate the single player’s contribution based on a deep learning model that predicts shooting success from the trajectory data of all players and a ball, leveraging official data of a professional basketball league. We use the difference of two output values by the prediction model when the trajectory data of the target player is given or not given as the model inputs. Our evaluation using the professional basketball dataset for one season confirmed that the predictive model had an accuracy of AUC=0.92. We also confirmed that the scoring contribution of each player calculated from this predictive model was significantly correlated with an existing player’s overall performance metric (R = 0.37, p < 0.001). The results suggest that our proposed method could be a new method to quantify a player's contribution to the team performance from trajectory data only instead of conventional experience-based player performance metrics.

In this demonstration paper, we present STM: a new system for securing and management of vehicle trajectory data using a generative model that balances privacy and utility. For instance, traditional methods for taxi-demand prediction pose the risk of privacy breaches from both the data and the model. To address this challenge, we deploy Spatiotemporal-GAN to generate synthetic trajectories that meet privacy regulations such as GDPR. We assess the quality of the generated data by constructing several taxi-demand prediction models. Moreover, we evaluate the privacy risk by implementing trajectory user linking attacks against the generated data and membership inference attacks against the prediction model. Our system is designed with rich interactivity and visualization, enabling the audience to use these modules. Overall, our approach demonstrates the potential of generative models in preserving privacy while maintaining data utility in the context of taxi-demand prediction.

The growing demand for intelligent environments unleashes an extraordinary cycle of privacy-aware applications that makes individuals' life more comfortable and safe. Examples of these applications include pedestrian tracking systems in large areas. Although the ubiquity of camera-based systems, they are not a preferable solution due to the vulnerability of leaking the privacy of pedestrians. In this paper, we introduce a novel privacy-preserving system for pedestrian tracking in smart environments using multiple distributed LiDARs of non-overlapping views. The system is designed to leverage LiDAR devices to track pedestrians in partially covered areas due to practical constraints, e.g., occlusion or cost. Therefore, the system uses the point cloud captured by different LiDARs to extract discriminative features that are used to train a metric learning model for pedestrian matching purposes. To boost the system's robustness, we leverage a probabilistic approach to model and adapt the dynamic mobility patterns of individuals and thus connect their sub-trajectories. We deployed the system in a large-scale testbed with 70 colorless LiDARs and conducted three different experiments. The evaluation result at the entrance hall confirms the system's ability to accurately track the pedestrians with a 0.98 F-measure even with zero-covered areas. This result highlights the promise of the proposed system as the next generation of privacy-preserving tracking means in smart environments.

Taxi-demand prediction is an important application of machine learning that enables taxi-providing facilities to optimize their operations and city planners to improve transportation infrastructure and services. However, the use of sensitive data in these systems raises concerns about privacy and security. In this paper, we propose the use of federated learning for taxi-demand prediction that allows multiple parties to train a machine learning model on their own data while keeping the data private and secure. This can enable organizations to build models on data they otherwise would not be able to access. Evaluation with real-world data collected from 16 taxi service providers in Japan over a period of six months showed that the proposed system can predict the demand level accurately within 1% error compared to a single model trained with integrated data.

In this paper, we propose KissLoc: a system that leverages onboard micro-size sensors of consumer eyewear devices for the dual purpose of activity recognition and localization. Specifically, the system trains a deep learning model for recognizing kissing activity and simultaneously identifying the timestamped location of its occurrence. Consequently, several predefined actions could be taken, including logging or controlling the smart environment. The evaluation shows that KissLoc can recognize the kissing activity with 82% accuracy while locating its occurrence with a median localization error of 1.25m.

Route Planning Systems (RPS) are a core component of autonomous personal transport systems essential for safe and efficient navigation of dynamic urban environments with the support of edge-based smart city infrastructure, but they also raise concerns about user route privacy in the context of both privately owned and commercial vehicles. Numerous high-profile data breaches in recent years have fortunately motivated research on privacy-preserving RPS, but most of them are rendered impractical by greatly increased communication and processing overhead. We address this by proposing an approach called Hierarchical Privacy-Preserving Route Planning (HPRoP), which divides and distributes the route-planning task across multiple levels and protects locations along the entire route. This is done by combining Inertial Flow partitioning, Private Information Retrieval (PIR), and Edge Computing techniques with our novel route-planning heuristic algorithm. Normalized metrics were also formulated to quantify the privacy of the source/destination points ( endpoint location privacy ) and the route itself ( route privacy ). Evaluation on a simulated road network showed that HPRoP reliably produces routes differing only by ≤ 20% in length from optimal shortest paths, with completion times within ∼ 25 seconds, which is reasonable for a PIR-based approach. On top of this, more than half of the produced routes achieved near-optimal endpoint location privacy (∼ 1.0) and good route privacy (≥ 0.8).

Taxi demand prediction is an essential process for enabling efficient taxi operations and customer satisfaction. However, most existing solutions are vulnerable to leakage of passengers' privacy or membership inference attack. In this study, we propose a privacy-preserving taxi-demand system built without real customer data. Specifically, we employ LSTM-GANs to generate synthetic trajectories reflecting the typical Spatio-temporal semantics of the original data without privacy leakage. The system evaluation was held on a real taxi service provider in Japan for six months. The results show the ability of the system to keep privacy with 86% of the cases with barely a negligible decrease in the prediction performance compared to using the original data.

The technology of 3D recognition is evolving rapidly, enabling unprecedented growth of applications towards human-centric intelligent environments. On top of these applications human segmentation is a key technology towards analyzing and understanding human mobility in those environments. However, existing segmentation techniques rely on deep learning models, which are computationally intensive and data-hungry solutions. This hinders their practical deployment on edge devices in realistic environments. In this paper, we introduce a novel micro-size LiDAR device for understanding human mobility in the surrounding environment. The device is supplied with an on-device lightweight human segmentation technique for the captured 3D point cloud data using density-based clustering. The proposed technique significantly reduces the computational complexity of the clustering algorithm by leveraging the Spatiotemporal relation between consecutive frames. We implemented and evaluated the proposed technique in a real-world environment. The results show that the proposed technique obtains a human segmentation accuracy of 99% with a drastic reduction of the processing time by 66%.

As the demand for location-based services increases, several research efforts have aimed for robust and accurate indoor localization, especially 3D localization. Due to the widespread availability of cellular networks and their support by commodity cellphones, cellular-based systems have recently been proposed as a means of achieving this. However, because of the inherent noise and instability of wireless signals, localization accuracy typically degrades and is not robust to the dynamic heterogeneity of mobile devices.In this paper, we present a CellStory, a deep learning-based floor estimation system that achieves a fine-grained and robust accuracy in the presence of noise. CellStory combines stacked denoising autoencoder learning models, and a probabilistic framework to handle noise in the received signal and capture the complex relationship between the signals detected by the mobile phone and its location. Evaluation using different Android phones in a real testbed shows that CellStory can accurately estimate the user’s floor 98.7% of the time and within one floor error 100% of the time. This accuracy demonstrates CellStory’s superiority over state-of-the-art systems as well as its robustness to heterogeneous devices.

In this paper, we demonstrate OmniCells: a cellular-based indoor localization system designed to combat the device heterogeneity problem. OmniCells is a deep learning-based system that leverages cellular measurements from one or more training devices to provide consistent performance across unseen tracking phones. In this demo, we show the effect of device heterogeneity on the received cellular signals and how this leads to performance deterioration of traditional localization systems. In particular, we show how OmniCells and its novel feature extraction methods enable learning a rich and device-invariant representation without making any assumptions about the source or target devices. The system also includes other modules to increase the deep model's generalization and resilience to unseen scenarios.

In this paper, we demonstrate a brand new design of a wearable device that enables privacy-preserving human activity recognition based on light-weight compact-size LiDAR. The device scans and represents the surrounding environment in 3D point clouds form. The system further processes this representation to define discriminative features that facilitate recognizing human activity on edge. These features are extracted using Spatio-temporal probabilistic clustering and fisher vector representations and then used to train a classification model for activity recognition purposes. Implementation and evaluation of the proposed system confirm its efficient ability to identify human activities.

Thermal comfort in indoor environments has an enormous impact on the health, well-being, and performance of occupants. Given the focus on energy efficiency and Internet of Things enabled smart buildings, machine learning (ML) is being increasingly used for data-driven thermal comfort (TC) prediction. Generally, ML-based solutions are proposed for air-conditioned or HVAC ventilated buildings and the models are primarily designed for adults. On the other hand, naturally ventilated (NV) buildings are the norm in most countries. They are also ideal for energy conservation and long-term sustainability goals. However, the indoor environment of NV buildings lacks thermal regulation and varies significantly across spatial contexts. These factors make TC prediction extremely challenging. Thus, determining the impact of building environment on the performance of TC models is important. Further, the generalization capability of TC prediction models across different NV indoor spaces needs to be studied. This work addresses these problems. Data is gathered through month-long field experiments conducted in 5 naturally ventilated school buildings, involving 512 primary school students. The impact of spatial variability on student comfort is demonstrated through variation in prediction accuracy (by as much as 71%). The influence of building environment on TC prediction is also demonstrated through variation in feature importance. Further, a comparative analysis of spatial variability in model performance is done for children (our dataset) and adults (ASHRAE-II database). Finally, the generalization capability of thermal comfort models in NV classrooms is assessed and major challenges are highlighted.

The global spread of coronavirus has sparked a considerable interest in technologies that facilitate seamless communication between users which are physically or spatially distant. Using current remote collaboration systems that utilize 3D sensing with LiDAR and depth cameras, point cloud streaming, and MR/VR devices, distant users can communicate with each other as if they did in person. However, these systems may violate users' privacy since they can share information of their entire personal space with other users. In addition, although various point cloud compression methods have been proposed, remote transmission of 3D scenes still requires significant bandwidth. This paper proposes a 3D spatial data sharing system based on the paradigm of “semantic communication”, i.e., controlling communication in the units of semantic objects. Our system understands the semantics of the scene and leverages point cloud streaming, thereby enabling users to assert fine-grained control over their privacy. Further, the system adaptively controls the size of the data frame based on network capacity and scene context. The experimental results show that the network delay can be reduced by 96%. We have also tested our system in a commercial 4G network, showing that 3-D spatial sharing with point clouds over severe networks is possible.

This paper presents the design and development of a lightweight, portable 3D spatial sensing device equipped with a compact LiDAR-type sensor. The device provides a 3D point cloud representation of the surrounding environment as captured by the sensor. Based on the acquired 3D point cloud, we propose a real-time object recognition method. The technical challenge is how to process the 3D point data in real-time while pursuing the best trade-off between the processing overhead and recognition accuracy. To answer the question, we leverage the Fisher Vector to extract spatio-temporal features of different objects enabling efficient classification of these objects using the Support Vector Machine approach. The experimental results show that the proposed method achieves mean Average Precision of 0.961. The processing speed on the device was 59.3 frames/second, indicating that object detection can be done in real-time on the device.

The realization of edge-based cyber-physical systems (CPS) poses important challenges in terms of performance, robustness, security, etc. This paper examines a novel approach to providing a user-centric adaptive route planning service over a network of Road Side Units (RSUs) in smart cities. The key idea is to adaptively select routing task parameters such as privacy-cloaked area sizes and number of retained intersections to balance processing time, privacy protection level, and route accuracy for privacy-augmented distributed route search while also handling per-query user preferences. This is formulated as an optimization problem with a set of parameters giving the best result for a set of queries given system constraints. Processing Throughput, Privacy Protection, and Travel Time Accuracy were developed as the objective functions to be balanced. A Multi-Objective Genetic Algorithm based technique (NSGA-II) is applied to recover a feasible solution. The performance of this approach was then evaluated using traffic data from Osaka, Japan. Results show good performance of the approach in balancing the aforementioned objectives based on user preferences.

Due to the widespread of dashcams, we will have more videos that capture roads/streets in driving. Consequently, a vast amount of the videos will be available and can be utilized for analyzing road safety and similar purposes. For example, suppose different dashcams can take vehicle/pedestrian traffic at a risky intersection at different timings. In that case, the collection of such videos will effectively recognize the cause of dangerous situations without surveillance camera infrastructure. However, identifying a Road Segment of Interest (RSI) in the video, such as near the intersection region, is challenging as the video frames do not usually include location tags. In this paper, we present a unique approach to attack this challenge. Assume that a video segment, called reference video, captures an RSI. We re-identify the RSI taken in another video (called test video) that captures the roads containing that RSI. By this approach, we can automatically extract the video segment corresponding to RSI from a given test video, using the reference video. We introduce AKAZE features to assess frame-level similarity and develop an algorithm to find frame-by-frame matching between reference and test videos. We have evaluated our method using 10 reference videos that correspond to 10 RSIs, each with 5 test videos. The result has shown that the average frame error distance was only 3.03 in daytime and 4.93 in nighttime, which are sufficiently low to re-identify RSI in the newly obtained test videos.

Location-based services have undergone massive improvements over the last decade. Despite intense efforts in industry and academia, a pervasive infrastructure-free localization is still elusive. Towards making this possible, cellular-based systems have recently been proposed due to the wide-spread availability of the cellular networks and their support by commodity cellphones. However, these systems only consider locating the user in a 2D single floor environment, which reduces their value when used in multi-story buildings.

Robust and accurate indoor localization has been the goal of several research efforts over the past decade. Towards achieving this goal, WiFi fingerprinting-based indoor localization systems have been proposed. However, fingerprinting involves significant effort; especially when done at high density; and needs to be repeated with any change in the deployment area. While a number of recent systems have been introduced to reduce the calibration effort, these still trade overhead with accuracy.

This paper takes a Computer-Vision (CV) and RFID fusion approach to tracking RFID-tagged objects in a room environment. Unlike similar CV-RFID fusion techniques that fuse two different estimates from CV and RFID systems to increase the certainty of locations, our system does not directly localize RFID tags. Instead, each tag-attached object's usage time by a human is detected by observation of the tag's phase variance. The time is then used to identify the pinpoint locations where the object is handed and released by the human, using the corresponding human trajectory observed by 3D depth cameras. RSSI variance caused by human movement is also leveraged to estimate rough locations of such objects that are not used by humans. Finally, the correspondence between the objects and RFID tags is automatically recognized by clustering the feature vectors of objects. Consequently, once the type of one object, such as pen, book, and cup, the other objects in the cluster can be annotated with the same object type to facilitate object management in the system. The experimental results have shown that object usage is detected with 0.984 accuracy, objects are localized with 58.3cm median error in severe NLoS environment, and ten types of objects are identified with 0.842 accuracy in a laboratory room.

In this paper, we propose a low-cost, non-invasive home activity recognition method using low-resolution power consumption data. Notably, we tackle the following two challenges. Firstly, we use only the time series of power consumption data aggregated per house and measured every few tens of seconds, which is usually used for demand monitoring by smart meters. We design a set of activities that can be recognized by such low-resolution data, and find out an appropriate feature set to train and test balanced random forest classifiers. Secondly, we consider the divergence of activity patterns seen in different households. Since supervised learning dedicated to each household is not a realistic solution, we arrange different classifiers trained by different household data in supervised learning, and present a method to automatically choose the best-fit classifier for the household of interest in the online phase. The experiment was conducted to collect aggregated power consumption data from eight real homes for 191 days. The result of activity recognition using the dataset shows that the proposed method achieved 70% recognition accuracy in identifying activities like cooking and sleeping, which is significant for non-invasive remote monitoring.

Situation awareness in the transport system has been significant for understanding the cause of traffic congestion, potential risks of accidents, and so on, and many efforts have been made to recognize surrounding situations by vehicle onboard cameras for low-cost sensing. These methods use computer vision technologies to recognize objects in the vicinity of the vehicle in the video image. However, to correctly recognize the positional relationship between the surrounding objects and the vehicle, lane detection, particularly multiple lane detection, should robustly be performed. In this paper, we propose a method for detecting multiple lanes from video images taken by onboard cameras. Since frame-by-frame spatial lane detection does not often work in a severe environment on multi-lane roads, the proposed method leverages temporal change detection, which enables complementing such lanes that are not detected in a frame and eliminating wrongly-detected ones for robust detection and tracking. In the experimental using the video images, we have achieved an accuracy of more than 90% in multi-lane environments with different conditions such as lane boundary occlusion by other vehicles, nighttime, rain, side road junctions. Besides, we achieved 96.47% accuracy to detect lanes on which vehicles are driving, while 88.70% in the comparative method.

In this paper, we propose a daily activity recognition Method using multiple low-cost infrared portable sensors and power consumption monitors, which are easy to deploy with less risk of privacy violation. It is significant to monitor daily life of elderly living alone to prevent them from having less exercise at home and from being at home with a depression tendency. Taking those sensor values as the input, our method utilizes Markov Logic Network (MLN) to express the association rules of sensors and activities as soft logical expressions. Then using the estimated activities from MLN and user feedbacks, the method trains MLN for better accuracy of recognition. We have implemented our system as a monitoring system that summarizes the activities of daily living as a timetable, which facilitates for remote family to monitor target elderly. We have deployed the system to two real houses of elderly people, and have shown the average accuracy was more than 85%.

Understanding humans and its environment is a key enabler of smart, intelligent applications and services for a future smart society. To deploy such services in our ambient environment, it is expected to fully utilize battery-less and maintenance-free IoT devices and technologies for more ambient, distributed computing. In recent years, Wi-Fi-based communications are becoming more energy-efficient, and channel state information (CSI) has the potential to sense more detailed information about the things in the real world. Besides, ambient backscatter has appeared as a promising technology for zero-energy sensing and communications. Leveraging those state-of-the-art technologies, energy harvested IoT devices for context recognition of humans and objects will be in reality. A significant challenge is how to make use of inferior, less-powerful zero-energy IoT devices to achieve processing of interest, i.e., accurate recognition of humans and objects, while a single device does not work. Therefore, we consider orchestrating distributed tiny IoT devices for both sensing and communications. Particularly, distributed machine learning in the local environment will achieve highly promising sensing in our ambient environment. In this paper, we survey the state-of-the-art technologies for zero-energy sensing and communications in the context of humans and objects sensing and recognition. Then, we address the challenges to be tackled in terms of such distributed, intelligent sensing using zero-energy devices. Finally, we introduce the concept of utilizing distributed IoT devices, followed by the statement about our ongoing work toward future zero-energy sensing and processing.

We have been working on automatic location estimation of HVACs utilizing the RSSI of BLE (Bluetooth Low Energy) to improve the time-consuming process in the HVAC system network setting where technicians need to identify the location of each HVAC one by one to assign a network address to the HVAC. This short paper particularly reports an initial attempt of training a BLE propagation prediction model adaptive to "above-ceiling" environment, which is a key enabler for automation of the assignment process. Based on the 1.3 million RSSI sample data collected in a real building for one week, we trained the model with different parameters and discovered that RMSE of RSSI prediction has the following trend: 1) it decreases with the introduction of the obstacle features -the number of beams and the number of HVAC machines, and 2) it decreases with the increase of polynomial term degree. The best model so far has an RMSE of 4.832 dBm tested by 10-fold cross-validation. The methodology and quantified result of this initial attempt provide reference not only for the network assignment process but also for building BLE-mesh networks in the above-the-ceiling environment, which can be exploited for HVACs management or sensor data aggregation.

Autonomous vehicles will bring tremendous benefits to society. However, it is expected to take a considerable period of time for their spread, and during the transition period, autonomous and manually-driven vehicles will share the same roads. Under such situations, the driving behavior of autonomous vehicles will influence manual drivers, for example, manually-driven vehicles may be stuck behind the autonomous vehicles and overtake them frequently if autonomous vehicles drive slowly for safety. In this paper, we investigate how the traffic flow and driving stress vary with autonomous vehicles by microscopic traffic simulation. We develop a microscopic traffic simulator that can reproduce traffic flow with autonomous vehicles and manually-driven vehicles. The behavior of these vehicles can be modeled by the combination of Intelligent-Driver Model (IDM) and Lane change Model with Relaxation and Synchronization (LMRS). These models can express various driver characteristics through simulation parameters such as driving speed and distance between vehicles (net distance), and we are able to create realistic scenarios like overtaking autonomous vehicles by manually-driven vehicles with faster speed than the legal speed. From the simulation results, we found that there is a desirable combination of speed and time headway that achieves both smooth traffic and less stress of drivers with a given percentage of autonomous vehicles.

By introducing the state of the art sensing technologies to the home, monitoring the daily activities of elders living alone or with husband/wife only, will be common soon to support their life from remote sites. We have developed a portable monitoring system to estimate daily activities just using simple, low-cost binary sensors and power monitors using MLN (Markov Logic Network), and installed in real elders' homes for monitoring. However, such monitoring systems usually require the Internet connection, which is not often available in depopulated areas due to the blind zone of 3G/LTE or installation cost which incurs extra investment for those people who have no interest in Internet services. In this paper, we propose a method of collecting daily activity monitoring data, which is obtained by our monitoring system working standalone at home, via LPWA (Low Power Wide Area). The idea is to utilize vehicles such as community buses and mail trucks which drive daily along almost fixed routes. An LPWA onboard receiver is mounted to collect data from each LPWA transmitter embedded into our monitoring system at home. Since not all the houses may be covered by the LPWA, LPWA-based "house clusters" are automatically formed to deliver the data in the cluster via a home contacted by the receiver vehicle. We have tested the system prototype in our university campus and installed it to the five real houses in the mountainous areas in cooperation with the Toyooka city municipal government, which has to manage broad depopulated areas.

Abstract Several attempts have been made to grasp three‐dimensional (3D) ground shape from a 3D point cloud generated by aerial vehicles, which help fast situation recognition. However, identifying such objects on the ground from a 3D point cloud, which consists of 3D coordinates and color information, is not straightforward due to the gap between the low‐level point information (coordinates and colors) and high‐level context information (objects). In this paper, we propose a ground object recognition and segmentation method from a geo‐referenced point cloud. Basically, we rely on some existing tools to generate such a point cloud from aerial images, and our method tries to give semantics to each set of clustered points. In our method, firstly, such points that correspond to the ground surface are removed using the elevation data from the Geographical Survey Institute. Next, we apply an interpoint distance‐based clustering and color‐based clustering. Then, such clusters that share some regions are merged to correctly identify a cluster that corresponds to a single object. We have evaluated our method in several experiments in real fields. We have confirmed that our method can remove the ground surface within 20 cm error and can recognize most of the objects.

While heading to a destination, we usually rely on our cognitive map constructed by audiovisual information from maps and our sight. However, errors or gaps between the real and our cognitive map often confuse us and lead to "wayfinding". In such a wayfinding state, we tend to take actions like wandering for perceiving errors and gathering information about surrounding environment. If such behavior can be detected by smartphones, we may design new applications on the smartphones, for instance, virtual "concierge" that timely helps us when we lose our ways. Also grasping spots where people are likely to lose their ways in large museums and theme parks would be useful to install or improve the signs and directions to support visitors. In this paper, we propose a method to detect individuals' wayfinding behavior from walking features by smartphone sensors. Based on the preliminary experiment, we extract sensor data features that can be collected through Android OS without privacy concerns, and build a binary classifier of user states, "normal" and "wayfinding". Through the two field experiments with 17 and 104 subjects, we have confirmed that our classifier achieved the F-measure of 0.93 and 0.85, respectively.

Understanding people flows in indoor environments such as office buildings and shopping malls is becoming more significant since such information can be useful for smart building applications. For example, crowd navigation applications can be designed for those people with big baggage to choose less crowded routes. People density and flow information is also useful for more adaptive and flexible shop advertisement. Wi-Fi RSSI-based passive measurement is a cost-effective approach as it allows to involve already-installed Wi-Fi APs as a part of the system. Nevertheless, the existing approaches have not tackled the high mobility cases where pedestrians walk with relatively high speed. In this paper, we propose a method to count the number of pedestrians waking across the line between two Wi-Fi devices. The method focuses on the RSSI fluctuations seen when pedestrians pass the device line. We extract some feature values that represent such attenuation and variance of RSSI and build a multi-class classifier to identify the number of pedestrians. The experiments were conducted with different position arrangements of Wi-Fi devices and at different locations to assess the applicability of the method. The results indicate that the proposed method can estimate the number of pedestrians with high accuracy when the Wi-Fi devices are placed obliquely against the walking direction.

Recently, several problems concerning smart and connected communities (S&CC) have been studied. In many S&CC applications such as autonomous driving, mobile crowdsourcing and crowd sensing, the mobility of vehicles and pedestrians has large impact on their performance and reliability. Many research works have used mobility generators to simulate realistic mobility, and evaluated the performance and reliability of the proposed applications and protocols. Although such mobility generators might be useful for producing typical mobility patterns, those mobility patterns are just snapshots and they cover only some part in the possible mobility patterns. Since many S&CC applications are used as social systems, their reliability and efficiency are very important. In order to improve the reliability of such mobility aware applications and accurately evaluate their performance, we need to collect many mobility patterns via simulation and/or observation from the real world, analyze their mobility influence statistically and provide adequate design platforms. In this paper, we first show the fact that many research works adopt snapshot-based mobility analyses. Then we propose a technique to reproduce a large part of possible mobility patterns, and provide a design platform to analyze their features and develop high-reliable mobility aware applications for S&CC. Some experimental results are also given.

Bluetooth Low Energy (BLE) is a promising technique for creating a cost-effective network to manage and control the BLE equipped equipment of buildings such as HVACs and lighting equipment, and the BLE network addresses can be used to identify the equipment. In order to enable the management (actuation and control) of equipment via BLE, each BLE network address must be linked with the physical location of equipment. However, the manual location identification effort is labour-intensive and time-consuming. In this paper, we introduce our basic idea to handle this issue and show a preliminary experimental result. The idea is to design semi-automatic position estimation technique by leveraging the change in RSSI of BLE devices during walking. Our method requires only one tester to carry a smartphone and walk around the building to collect the RSSI. We deployed 10 BLE devices to conduct the preliminary experiment in a single room. The result shows that the locations of 80% BLE devices are correctly identified.

In an era where billions of IoT devices are deployed, edge/fog computing paradigms are attracting attention for their ability to reduce processing delays and mitigate waste of communication resources. However, since the computing system assumed by edge/fog paradigms have heterogeneity (in terms of the computing power of devices, network performance between devices, device density, etc.), provisioning computational resources according to computational demand becomes a challenging constrained optimization problem. In this paper, we propose in-situ resource provisioning method consisting of insitu resource area selection with adaptive scale out and in-situ task scheduling based on tabu search algorithm. We conducted a simulation study in a target regional area where 2,000 IoT devices and 10 IoT services are deployed to evaluate the effectiveness of the proposed algorithm. The simulation results show that our proposed algorithm can obtain higher user QoS compared to conventional resource provisioning algorithms.

Due to the nature of smartphones' portability and mobility, many mobile apps are usually utilized in the real field environment using GPS, Wi-Fi and embedded sensors. For example, any navigation app uses GPS and Wi-Fi to locate the user in the map, and streaming apps may be used in cafeteria or even outside to satisfy the users' demand to watch soccer games anywhere and anytime. To test the usability and performance of such mobile apps in in-situ environment, we need to bring the apps to such physical world and run (a number of) test scenarios, which is often cost-inefficient depending on the size, apps and situations assumed in those scenarios. In this paper, we design and develop a testbed to test mobile apps in VR space. The system allows developers to use a real smartphone in VR and to test and evaluate their apps at the interested locations, with various network environment. The system builds and reproduces the real world environment of 3D space and real networks in the VR environment, using the existing 3D city models and our original Wi-Fi database. Then it enables to real-timely integrate the screen of the VR user's smartphone in the VR space. The user can operate the app via the VR view, and test the usability and performance of the app in such an emulated environment. The experimental result shows our architecture could achieve such cyber-physical integration with 695.5 ms delay, which is negligible in many semi-realtime services such as navigations.

We present CrowdMeter: a participatory system that leverages the sensed data collected from users' phones during their daily train commutes to gauge the real-time congestion level in railway stations. CrowdMeter tracks the passenger's position in the station as well as identifies her context (e.g., waiting for a train, buying a ticket) along her trajectory from the station's entrance to the train. Therefrom, CrowdMeter extracts novel features, based on the user's location and context, from the phone sensors. These features capture the passenger's behavior (e.g., the walking pattern) and the ambient environment characteristics (e.g., the ambient sound) that can indicate the surrounding congestion level along the passenger's route in a railway station. Finally, the system highlights each area of the station with a specific color (green, amber, red) that corresponds to one of a three congestion levels (low, medium, high).Evaluation of CrowdMeter through a field experiment in 10 different train stations in Japan shows that it can infer the congestion levels accurately, highlighting its promise as a ubiquitous travel-support service.

The conventional wireless sensor networks are designed to gather information from sensors, and pervasive computing systems involve such wireless sensor networks that consist of a number of tiny-/micro-sensors to sense the real-world objects and phenomena. In addition, they also need to conduct training tasks for highly-intelligent processing like pattern recognition and anomaly detection, but doing such tasks in cloud servers may often be impossible or expensive due to lack of communication facility, data transfer overhead, privacy concerns and so on. In this paper, we propose a unique approach to carry out deep learning (both training/testing) among wireless sensors nodes for in-situ, edge-heavy processing. The idea is to appropriately assign neurons of CNN (Convolutional Neural Network) to wireless nodes, each of which has limited processing capability but can have some power when they are united. For this purpose, we have designed a distributed, restricted version of CNNs. We conducted two experiments using real data; one is for anomaly detection of temperature in an over-1,400m2lounge space using 50 temperature sensors to confirm the learning capability as well as communication overhead, and another is for activity recognition using a 6x6 array of thin-, energy-efficient film-type infra-red sensors with micro-processors to demonstrate our concept.

This paper presents a new algorithm for building an indoor map by integrating point clouds of 2D light detection and ranging (LIDAR) scanners in indoor environments. Iterative closest point (ICP) algorithm is one of the well-known methods for such purpose and often used for mobile robot SLAM. However, the algorithm is designed based on dense (or continuous) measurement of the same space with known relative positions of measurement points and angles, and it does not often work efficiently if the measurement is sparse and/or LIDAR locations are unknown. Such situations are seen when some installed LIDARs in a room are used to build a background indoor map for object tracking, or mobile LIDARs are used by technicians to build a digital indoor map, where each space is captured only at a few locations with different angles. To tackle this issue, our method extracts line segments and edge points as features from given LIDAR point clouds and finds shape coincidences commonly contained in a pair of given point clouds to identify positional relationships between LIDARs. By this information, these point clouds can be integrated into common 2D coordinates. Indoor map generation is realized by sequentially applying this integration procedure to every pair of point clouds. Also, the experiments on real data show that our method can identify the relative positions of LIDARs with 10cm-order errors in average, and by sequentially applying the point-cloud integration, the generated maps have only 3% errors.

Delay-tolerant networking (DTN) architecture has emerged to cope up with loss of continuous connectivity issue that may occur under some environmental conditions such as infrastructure down due to disaster. Bundle protocols generally enhance network delay and reliability by storing and forwarding data-packed bundles on multiple nodes in the same time through flooding algorithm. On the other hand, the flooding mechanism usually incurs a consumption of resources. In this paper, we propose a robust routing protocol called BALANCE for self-organized civilian DTN. In civilian contexts, a resource constraint is a primary issue. So, our BALANCE sets an objective function to minimize the number of replications over networks by using a regression model for predicting future spreading effect considering the delivering probability. Since the device owners (civilian) could be selfish and decide not to store or forward any bundles, we further introduce a selfish factor applied together with our newly formulated SINR-based function to achieve the delivering probability. We conduct evaluations in stationary and mobile scenarios regarding disaster situations. The results show the highest traded-off values of delivering performance and resource consumption of the proposed protocol compared to widely-used protocols and present our superior in restricted conditions like limited buffer size or selfish participants.

Indoor localization is the key infrastructure for indoor location-aware applications. In this paper, we consider an emergency scenario, where a team of soldiers or first responders perform time-critical missions in a large and complex building. In particular, we consider the case where infrastructure-based localization is not feasible for various reasons such as installation/management costs, a power outage, and terrorist attacks. We design a novel algorithm called the collaborative indoor positioning scheme (CLIPS), which does not require any pre-existing indoor infrastructure. Given that users are equipped with a signal strength map for the intended area for reference, CLIPS uses this map to compare and extract a set of feasible positions from all positions on the map when the device measures signal strength values at run time. Dead reckoning is then performed to remove invalid candidate coordinates, eventually leading to only correct positions. The main departure from existing peer-assisted localization algorithms is that our approach does not require any infrastructure or manual configuration. We perform testbed experiments and extensive simulations, and our results verify that our proposed scheme converges to an accurate set of positions much faster than existing noncollaborative solutions.

In this paper, we propose a general complex event processing (CEP) engine aiming for accomplishing at smart IoT edge devices in a fully distributed manner. We introduce a pseudo-source mechanism to cover a wide range of processing and obsolete prerequisite of source-specification at the same time, along with a brand-new event specification language defined to support relation-based processing. Against cloud-based approaches, our behind-edge approach can prevent data overflow and privacy issues, and fully distributed processing can draw the power of the edge devices. To achieve that in a resource-limited edge environment, we formulate an optimization problem of processing task assignment and stream delivery, and propose a fully-autonomous workload distribution mechanism. A large-scale simulation with a realistic smart-building scenario shows that our proposed method achieves about 6.6 times smaller flow volume and 2 times lower loss rate compared to centralization and is relatively superior to a hop-based distribution approach. Notably, a prototype engine is successfully deployed over an ad-hoc wireless sensor and actuator network through Intel Edison modules in the real environment.

Recently, several researches concerning with smart and connected communities have been studied. Soon the 4G / 5G technology becomes popular, and cellular base stations will be located densely in the urban space. They may offer intelligent services for autonomous driving, urban environment improvement, disaster mitigation, elderly/disabled people support and so on. Such infrastructure might function as edge servers for disaster support base. In this paper, we enumerate several research issues to be developed in the ICDCS community in the next decade in order for building safe, smart cities resistant to disasters. In particular, we focus on (A) up-to-date urban crowd mobility prediction and (B) resilient disaster information gathering mechanisms based on the edge computing paradigm. We investigate recent related works and projects, and introduce our on-going research work and insight for disaster mitigation.

Passive infra-red (PIR) motion detection binary sensors are reasonable for human tracking in household as they can be portable operated by small batteries and be attached onto walls, shelves, floors or ceilings. However, such flexibility of sensor location causes difficulty since the location information of each sensor should be managed and given to the tracking system by the residents. To mitigate human errors, we propose a sensor localization method to automatically identify the location of multiple binary motion sensors in a house from the observed sequence by binary sensors. The experimental results show that the accuracy of our approach is above 80% after 5 days observation.

In surface mine environment, the size of mining equipment is typically very huge. Therefore, the sight of equip- ment operators in high- mounted cockpits is severely limited and those operators are often unaware of the approaching or nearby vehicles. It is reported that this unawareness causes serious accidents, which motivates us for proximity detection of those vehicles. Inter-Vehicle Communication (IVC) is a promising approach in the road transport environment, but is not directly applicable to mining vehicles due to different features of radio propagation affected by huge bodies and terrain. In this paper, we propose a method that leverages IVC to support drivers' awareness in surface mine. The idea is that we build a radio propagation model for haul trucks, which are the most typical vehicles in surface mine, by accurate ray-tracing. The model is validated in the field experiment, where actual radio signal propagation is measured and compared with the model. Then a mining environment, where steep cliffs due to excavation are often seen, is modeled and we estimate the received signal strength based on the locations of haul trucks and antenna locations. As a result, we point out several blind spots in such environment, which is significant for V2V proximity detection in surface mine.

EdgeComputing and Fog Computing are new paradigms where data processing is executed in or on the edge of networks to mitigate cloud server load. However, EdgeComputing and Fog Computing still need powerful servers on the edge of networks which impose additional costs for deployments. We proposed a platform called IFoT (Information Flow of Things) that efficiently performs distributed processing as well as distribution and analysis of data streams near their sources based on "Process On Our Own (PO3)" concept. In IFoT, processing of tasks for cloud servers is delegated to an ad-hoc distributed system consisting of proximity IoT devices for distributed real-time stream processing. In this demonstration, we show a face recognition system for person tracking developed on top of IFoT middleware which locally processes video streams in real-time and in a distributed manner by using computational resources of IoT devices.

No abstract available.

Increasing availability and usability enhancement of Wi-Fi in public areas has become more active. However, due to the dense deployment of Wi-Fi access points (APs), there is a chaotic and disorderly environment in urban areas. In our previous work, we have designed a function that predicts the network performance at each Wi-Fi AP according to the measurement of IEEE802.11 MAC frames sensed in each Wi-Fi channel. However, it was not examined in such scenarios assuming urban environment. We should understand the situations of current Wi-Fi AP deployment and traffic conditions, and should confirm the effectiveness of channel migration in such realistic environment. In this study, we proposed urban Wi-Fi channel utilization model based on real urban Wi-Fi measurement. We show that our method can predict the best channels and APs can migrate to them in the urban scenario.

We present TransitLabel, a crowd-sensing system for automatic enrichment of transit stations indoor floorplans with different semantics like ticket vending machines, entrance gates, drink vending machines, platforms, cars' waiting lines, restrooms, lockers, waiting (sitting) areas, among others. Our key observations show that certain passengers' activities (e.g., purchasing tickets, crossing entrance gates, etc) present identifiable signatures on one or more cell-phone sensors. TransitLabel leverages this fact to automatically and unobtrusively recognize different passengers' activities, which in turn are mined to infer their uniquely associated stations semantics. Furthermore, the locations of the discovered semantics are automatically estimated from the inaccurate passengers' positions when these semantics are identified. We evaluate TransitLabel through a field experiment in eight different train stations in Japan. Our results show that TransitLabel can detect the fine-grained stations semantics accurately with 7.7% false positive rate and 7.5% false negative rate on average. In addition, it can consistently detect the location of discovered semantics accurately, achieving an error within 2.5m on average for all semantics. Finally, we show that TransitLabel has a small energy footprint on cell-phones, could be generalized to other stations, and is robust to different phone placements; highlighting its promise as a ubiquitous indoor maps enriching service.

In recent years, several advanced safety techniques have been put to practical use for supporting safe driving, ASV (Advanced Safety Vehicle) can detect the presence of surrounding vehicles and pedestrians autonomously. On the other hand, cooperative techniques such as cooperative vehicle-to-vehicle awareness have attracted a lot of attentions in the field of ITS. In this paper, we propose a method for identifying positional relationship of vehicle groups. In the proposed method, connected vehicles share surrounding vehicles' positions detected by installed range sensors via inter-vehicle communication. By composing those pieces of information into a single view, each connected vehicle can recognize the vehicle group ahead, which cannot be recognized standalone. By utilizing the distance information from two different information sources and executing the matching process, the correct positional relationship among neighboring vehicles can obtain. We have conducted a simulation-based experiment using two types of data to evaluate the effectiveness of the approach, and we have confirmed to achieve the correct identification with high probability by using shared velocity information.

We present PassActiv, a mobile sensing system for the automatic activity recognition of railway passengers. Our key observations show that certain passengers' activities (e.g., purchasing tickets, etc) present identifiable signatures on one or more cell-phone sensors which can be leveraged to automatically recognize those activities. Evaluation of PassActiv through a field experiment in major train and subway stations in Japan shows that PassActiv can detect different activities accurately with at most 3% false positive rate and 4% false negative rate for all types of passengers' activities.

A system to improve the quality of human life is developed and proposed. A model-based approach is used where smart home residents, appliances, energy sources and correlations among them are comprehensively modeled. The model was integrated with activity recognition information that enables the system to suggest smart life tips that provide advice to residents in a non-intrusive way. A crowd-sourced large-scale survey of 1,000 subjects was conducted that enabled important tips for improving the quality of human life to be quantified. On the basis of the survey results, quantitative metrics and strategies were designed for presenting suitable tips in a timely manner depending on the lifestyle of subjects. The system was evaluated by (1) 34 actual subjects in virtual smart homes and (2) family members in an actual house in an experiment lasting more than one month in which actual sensors were deployed.

This paper proposes a method of aggregating tempo-spatial data generated by sensors deployed in buildings or houses. The size of each sensor data such as temperature is usually small, but it often involves many additional data to represent its attribute values like time, location, data type and data precision. This would often increase the traffic volume between sensor gateway at building/home side and service providers at server side. In our method, such sensor data are packed into multidimensional matrices indexed by those attribute values for more compact representation, and the compressed sensing technique is adaptively applied to further reduce the data size. The method was applied to a field trial with KDDI corporation to collect data from 29 community facilities, and the traffic volume was reduced to 50% with reasonable precision of data restoration.

In this paper, we design a mobile augmented reality (AR) system called TweetGlue, which overlays text messages (i.e., tweets) that are posted to a local social networking service onto live view images from cameras in mobile/wearable devices. By displaying the tweets at the current positions of the users who posted them, it supports social interaction between the users. Accurate pose tracking of mobile devices is an essential building block of such mobile AR applications. While visual features that are extracted from scenes are commonly used for vision-based pose estimation, it would not be suitable for such a mobile social AR application because the features may be often occluded by human bodies nearby. To cope with the problem, we leverage an external pedestrian tracking system using a small number of laser-based distance measurement sensors (i.e., LRS sensors) to utilize the surrounding human bodies as virtual markers for pose estimation. The mobile devices periodically analyze images from the embedded camera sensor to estimate relative positions of pedestrians in the images. By matching the estimated relative positions with accurate human location measurements by the LRS sensors, the system robustly identifies location and horizontal orientation of the devices. Through simulation experiments, we show that the TweetGlue system can accurately identify pose of mobile devices in 83% of the simulated scenarios.

We propose a method to estimate car-level train congestion using Bluetooth RSSI observed by passengers' mobile phones. Our approach employs a two-stage algorithm where car-level location of passengers is estimated to infer car-level train congestion. We have learned Bluetooth signals attenuate due to passengers' bodies, distance and doors between cars through the analysis of over 50,000 Bluetooth real samples. Based on this prior knowledge, our algorithm is designed as a Bayesian-based likelihood estimator, and is robust to the change of both passengers and congestion at stations. The car-level positions are useful for passengers' personal navigation inside stations and car-level train congestion information helps determine better strategies of taking trains. Through a field experiment, we have confirmed the algorithm can estimate the location of 16 passengers with 83% accuracy and also estimate train congestion with 0.82 F-measure value in average.

Data collection from a crowd of mobile devices is an essential building block of emerging mobile sensing systems. In this paper, we propose an efficient information diffusion protocol for sensor data collection via an opportunistic network. The proposed method detects groups of pedestrians based on the history of radio connectivity between the nodes and maintains a local network (i.e., a cluster) among the detected group members. By collaboratively performing neighbor discovery and link management with the cluster members, it enhances energy-efficiency of the neighbor discovery and minimizes the information delivery delay. Simulation results show that the proposed method can improve the message delivery performance by 16%–83% with equivalent contact probing intervals.

This paper presents a strategy to choose WiFi channels in urban areas. We consider (i) inter-channel interference where adjacent channels interfere with each other in WiFi systems and (ii) urban situations where many APs in different systems are deployed in an uncoordinated way. As it is often hard to identify the channel with less interference in such a situation, we present a channel scoring function that estimates the performance level of each channel. To build the scoring function, we have conducted exhaustive simulations with a large number of scenarios, and multiple regression analysis has been applied where channel occupancy patterns, traffic volumes and RSS in those channels are used as explanatory variables. To evaluate our method, this scoring function was examined in a realistic scenario where several APs interfere with the AP of interest. We have confirmed that the scores and the actual performance are well-matched where the Spearman's rank correlation coefficient was sufficiently high and can identify the top-ranked channel as well.

Despite recent advances in localization technology for mobile devices, to provide real-time position information to people indoors is still a big challenge; usually there is a trade-off between localization accuracy and infrastructural costs (e.g., dense anchor deployment). A possible solution would be employing cooperative approaches which utilize estimated positions of surrounding mobile nodes to complement a small number of anchors. However, it often results in poor estimation accuracy since a temporary large position error due to node mobility easily propagates to neighbor nodes. This paper presents a novel cooperative localization algorithm that addresses this problem by focusing on “stop-and-go behavior” of indoor pedestrians. The key idea is to collaboratively find movement state (moving or static) of each node based on peer-to-peer distance measurement which is inherently necessary for cooperative localization, and use only static nodes as reference points for localization to avoid potential accuracy deterioration. Also, nodes in static state can reduce localization frequency to conserve battery power, keeping the tracking quality. Through extensive simulations, we have demonstrated the performance of our method in terms of accuracy and energy efficiency. The effectiveness in a real application scenario has been also confirmed using a measurement-based sensor model and real mobility traces.

This paper presents a scheduling algorithm for a set of wireless stations such as road-side access points for vehicular networks and outdoor WiFi stations, which are deployed in wide urban areas and may compete with each other for limited wireless resources. Different from a number of conventional approaches most of which consider detailed information on individual stations and signal interference among them, we focus more on geography of the areas of interest, and provide a novel algorithm that pursues the best balance among (i) optimality of resource utilization, (ii) robustness to new station installation and traffic demand changes, and (iii) scalability to the population of stations and area size. We have confirmed the performance by experimental simulations with several scenarios, and the applicability of approach has been testified by a case study on a scheduling problem for roadside access points of vehicular networks in cooperation with a manufacturing corporation.

Indoor localization has attracted much attention recently due to its potential for realizing indoor location-aware application services. This paper considers a time-critical scenario with a team of soldiers or first responders conducting emergency mission operations in a large building in which infrastructure-based localization is not feasible (e.g., due to management/installation costs, power outage, terrorist attacks). To this end, we design and implement a collaborative indoor positioning scheme (CLIPS) that requires no preexisting indoor infrastructure. We assume that each user has a received signal strength map for the area in reference. This is used by the application to compare and select a set of feasible positions, when the device receives actual signal strength values at run time. Then, dead reckoning is performed to remove invalid candidate coordinates eventually leaving only the correct one which can be shared amongst the team. Our evaluation results from an Android-based testbed show that CLIPS converges to an accurate set of coordinates much faster than existing noncollaborative schemes (more than 50% improvement under the considered scenarios).

In this paper, we propose a novel approach to positioning mobile phone users in a crowd of people. We assume such a situation that people in specific space (e.g., an exhibition hall) have mobile phones and they periodically probe proximity between the surrounding users via short-range wireless communication (e.g., Bluetooth). We also assume that laser range scanners (LRSs) are deployed in the space to track positions of those mobile phone users. By fusing the accurate, but anonymous trace information obtained by LRSs and the identified phone-to-phone proximity information, our method derives accurate identified traces of each mobile phone user, which are essential for pedestrian navigation and other emerging location/situation-aware applications. Through extensive simulations, we show that the tracking error quickly converges below 1m in most cases.

We propose an opportunistic ad hoc localization algorithm called Urban Pedestrians Localization (UPL), for estimating locations of mobile nodes in urban districts. The design principles of UPL are twofold. First, we assume that location landmarks are deployed sparsely due to deployment-cost constraints. Thus, most mobile nodes cannot expect to meet these location landmarks frequently. Each mobile node in UPL relies on location information received from its neighboring mobile nodes instead in order to estimate its area of presence in which the node is expected to exist. Although the area of presence of each mobile node becomes inexact as it moves, it can be used to reduce the areas of presence of the others. Second, we employ information about obstacles such as walls, and present an algorithm to calculate the movable areas of mobile nodes considering obstacles for predicting the area of presence of mobile nodes accurately under mobility. This also helps to reduce each node's area of presence. The experimental results have shown that UPL could be limited to 0.7r positioning error in average, where r denotes the radio range by the above two ideas.

In this paper, we design an efficient method of simulating wireless networks that use CSMA/CA-based protocols in the MAC layer. In the method, a stochastic model to estimate the CSMA/CA frame transmission delay is naturally incorporated into the conventional fully event-based model. The stochastic model can simplify the interactions between a frame transmitter and its surrounding nodes, which alleviates the event scheduling overhead in simulation. The important feature is that the stochastic model can be applied in “per-node” and “time” basis, i.e. we may simulate the behavior of some intended nodes precisely while the others are simplified by the stochastic mode to save computational resources. To the best of our knowledge, this is the first approach to coexistence of the stochastic and event-based models in wireless multi-hop network simulation. We have implemented this scheme in a commercial network simulator and conducted several experiments. From the results, it is confirmed that the proposed method could perform simulation of frame transmission much faster than the fully event-based simulation achieving the same accuracy as the conventional model.

The electronic triage system developed by our research group is a wireless sensor network, constructed by electronic triage tags composed of vital sign sensors and ZigBee modules. The system assumes a network of hundreds of patients, simultaneously transmitting data to a remote sink. We have investigated the performance of a simple data collection mechanism in real scenarios, based on IEEE 802.15.4 in a ZigBee device like SunSPOT, and have examined some challenges in such scenarios. The data delivery ratio of 98% has been recorded while sending data equivalent to that of 100 devices in a triage tent scenario, where every device is one hop away from the sink. Adjustment of hello packet interval according to the scenarios has been observed to be an important factor. The results are found to be useful in adjusting parameters in our future work.

In order to analyze important properties of network protocols, such as robustness and applicability, we may need exhaustive tests to observe effects of various factors under different settings. For recent protocols on dynamic, large-scale and environment-aware networks such as wireless sensor networks and mobile ad-hoc networks, we should consider many factors due to diversity of hardware profiles, upper/lower layer protocols and physical environments. In this paper, we propose a comprehensive test method to analyze the effect of (usually 10 or more) factors on such network systems. Our method takes a set of factors to be considered and their (representative) domain values as inputs, analyzes their effects on the systems, and determines dominant factors that have impact on the performance and their interactions. Instead of applying exhaustive tests that require all the combinations of domain values, we take a step-wise approach that examines step-by-step suspected sets of factors, which requires fewer combinations. We also justify this approach based on a reasonable fault model. The approach also contains an analytical method to identify the performance characteristics. Through realistic case studies, we show that we could find sets of dominant factors in wireless networks systematically.

In mobile wireless networks such as WSNs, WMNs and MANETs, movement of sensor nodes, clients and relay nodes has a great impact on the performance. Nevertheless, geography is too simplified in random-based mobility models such as RWP, while it is unrealistic to prepare trace-based mobility patterns for potential combinations of geography and mobility. To fulfill the gap, this paper provides a new method to automatically generate natural mobility patterns realizing designated node distributions. The goal of this work is to synthesize the movement patterns that can capture real (or intentional) node distributions. The method determines the probabilities of choosing waypoints from the subregions, satisfying the given node distributions. For this purpose, the relationship between the probabilities and node distributions is analyzed. Based on the analysis, the problem is formulated as an optimization problem of minimizing the error from the designated node distribution. Since the problem has non-linear constraints, a heuristic algorithm is designed to derive the near-optimal solutions. Several experiments have been conducted to show that a variety of node distributions could be realized in the proposed mobility model where the maximum error from the given node distributions was around 0.5%. Additionally, a case study has been conducted to show the applicability of the method.

This paper presents a cooperative localization approach for mobile nodes using wireless and ranging devices. We consider scenarios where node mobility follows stop-and-go behavior; we can then utilize the different movement states of nodes as an input to our localization approach. In the proposed method, each node autonomously finds among its surrounding nodes the ones that do not seem to move, and treats them as static nodes. Only nodes that are deemed static are then used as reference points for position estimation. Furthermore, each node adjusts its localization frequency automatically according to its estimated velocity. Performance evaluation results based on a realistic sensor model and actual mobility traces show that our method could achieve sufficient accuracy and efficiency for an exhibition scenario where people need to be tracked.

This paper presents a vehicular positioning system in which multiple vehicles cooperatively calibrate their positions and recognize surrounding vehicles with their GPS receivers and ranging sensors. The proposed system operates in a distributed manner and works even if all vehicles nearby do not or cannot participate in the system. Each vehicle acquires various pieces of positioning information with different degrees of accuracies depending on the sources and recency of information, and compiles them based on likelihood derived from estimated accuracies to minimize estimation errors. A simulation based performance evaluation given in the paper shows that the proposed system improves the estimation accuracy by 85% on average with respect to the standalone GPS receiver, and recognizes about 70% surrounding vehicles with an error of 1m.

In this paper, we propose an algorithm to estimate 2D shapes and positions of obstacles such as buildings using GPS and wireless communication history of mobile nodes. Our algorithm enables quick recognition of geography, which is required in broader types of activities such as rescue activities in emergency situations. Nevertheless, detailed building maps might not be immediately available in private regions such as large factories, warehouses and universities, or prepared maps might not be effective due to collapse of buildings or roads in disaster situations. Some methodologies adopt range measurement sensors like infra-red and laser sensors or cameras. However, they require dedicated hardware and actions for the measurement. Meanwhile, the proposed method can create a rough 2D view of buildings and roads using only wireless communication history between mobile nodes and position history from GPS receivers. The results from the experiment conducted in 150m×190m region on our university campus assuming rescue and treatment actions by 15 members have shown that our method could generate a local map with 85% accuracy within 350 seconds. We have also validated the performance of our algorithm by simulations with various settings.

In this paper, we propose a new trajectory estimation method named TRADE (TRAjectory estimation in DEcentralized way). TRADE is a range-free localization algorithm in fully decentralized mobile ad hoc networks. In TRADE, each mobile node periodically transmits messages containing its estimated trajectory information, and re-computes its own trajectory using those from its neighbors. This information exchange considerably contributes to improvement of the position accuracy. Furthermore, we give the optimal design of the protocol based on the analysis of the algorithm property. Through the analysis, we consider how much trajectory information should be exchanged among nodes to estimate the position within a certain error range in the protocol design. We have evaluated the position accuracy under various settings, and have shown the effectiveness of the protocol in the real world through two realistic application examples.

In this paper, we propose a new overlay multicast protocol designed for inter-active multimedia streaming applications. The protocol considers the heterogeneity of end-hosts and tries to minimize the negative impact (data outage) of end-hosts' unannounced departures. For this purpose, it concentrates on end- hosts' reliability (lifetime for instance) and constructs a shared tree called ms-DDBMSST (multiple-source Degree and Delay Bounded Maximum Stability Spanning Tree) as an overlay network that involves all the participants of the application, in a distributed manner. For a given set of nodes where some of them are senders, ms-DDBMSST is a spanning tree where the receive path stability of the entire tree is maximized while satisfying the delay-from-source constraint and degree constraint for each node. We believe that this is the first approach that defines ms-DDBMSST construction problem and presents a distributed protocol for the purpose. Our performance evaluation is based on experiments in both simulated networks and PlanetLab that strongly shows the efficiency and usefulness of the proposed protocol.

In this paper, we propose an off-line algorithm called TRACKIE to estimate trajectories of mobile nodes based on encounter information. This method only assumes reasonable number of landmarks and ad-hoc wireless communication facility of mobile nodes, and does not rely on multi-hop ad-hoc networks nor global positioning system. The method achieves low-cost estimation of trajectories and provides accurate solution (the average estimation error was less than 40% of the wireless range in simulations). We have evaluated TRACKIE with Micaz Mote and shown that estimation error is about 2 m in real environments where wireless range is about 3 m.

In this paper, we present a range-free ad-hoc localization algorithm called UPL (Urban Pedestrians Localization), for positioning mobile nodes in urban district. The design principle of UPL is two-fold. (1) We assume that location seeds are deployed sparsely due to deployment-cost constraints. Thus most mobile nodes cannot expect to meet these location seeds frequently. Therefore, each mobile node in UPL relies on location information received from its neighboring mobile nodes in order to estimate its area of presence. The area of presence of each mobile node becomes inexact as it moves, but it is helpful to reduce the areas of presence of the other mobile nodes. (2) To predict the area of presence of mobile nodes accurately under mobility, we employ information about obstacles such as walls, and present an algorithm to calculate the movable areas of mobile nodes considering obstacles. This also helps to reduce each node's area of presence. The experimental results have shown that by the above two ideas UPL could achieve 8mpositioning error in average with 10mof radio range.

We design a mobility-aware data management (MoDA) scheme for mobile ad hoc networks (MANETs) composed by mobile nodes such as urban pedestrians and vehicles. By fully utilizing the knowledge about the trajectories of mobile nodes, MoDA determines how replicas of data are copied and transferred among mobile nodes to provide the required data accessibility. Experimental results have shown that MoDA could achieve the small number of data transfers among mobile nodes while keeping reasonable accessibility.

In this paper, we propose a routing protocol for MANETs called Contact-based Hybrid Routing (CHR) protocol where each node maintains potential routes to the nodes which it encountered. Only one route request message is forwarded along the potential route to the destination maintained by the source, and if a gap between two intermediate nodes is found on the potential route, it is filled by the potential routes maintained by the two nodes. Based on this idea, our goal is to reduce the number of route request messages by maintaining small amount of information at nodes. The experimental results in the RWP mobility and the disaster evacuation mobility have shown that CHR could reduce the number of messages while keeping reasonable reachability to the destinations.

In this paper, we present a QoS routing protocol called GVGrid for multi-hop mobile ad hoc networks constructed by vehicles, i.e., vehicular ad hoc networks (VANETs). GVGrid constructs a route on demand from a source (a fixed node or a base station) to vehicles that reside in or drive through a specified geographic region. The goal of GVGrid is to maintain a high quality route, i.e. a robust route for the vehicles movement. Such a route can be used for high quality communication and data transmission between roadsides and vehicles, or between vehicles. The experimental results have shown that GVGrid could provide routes with longer lifetime, compared with an existing routing protocol for VANETs

We have proposed realistic mobility models and developed a network simulator called MobiREAL for performance evaluation of MANET applications. Simulator users can easily generate a mobility scenario on any city section with the support tool of MobiREAL and can easily analyze the impact of mobility through the simulation visualization tool of MobiREAL. In this demonstration, we will show the usefulness of MobiREAL and importance to simulate MANET applications in realistic environments.

In this paper, we propose an application layer multicast protocol called Emma/QoS. Emma/QoS manages multiple multicast video streaming flows in a fully distributed manner so that the user gain in receiving those flows can be maximized. Emma/QoS assumes that each end-host is capable of filtering streaming flows when relaying them to the other end-hosts. Using this functionality, Emma/QoS (a) accommodates a request of a new video source by degrading the bit-rates of the existing video streams, and (b) adjusts their transmission rates according to given user benefit functions, which can be application-specific. Simulation results have shown the efficiency of Emma/QoS.

In this paper, we evaluate the performance of real-time data transmission by our routing protocol GVGrid. In Ref. [5], we have presented GVGrid, a position-based routing protocol for multi-hop mobile ad hoc networks constructed by vehicles. GVGrid constructs a route on demand from a source (a fixed node or a station) to vehicles that exist in a destination region. The experimental results have shown that GVGrid could achieve high quality data transmission in a region of a medium and high density of vehicles.

In this paper, we propose a new protocol called shared tree streaming (or STS in short) protocol that is designed for interactive multimedia streaming applications. STS is a decentralized protocol that constructs a shared tree called s-DBMDT (sender-dependent degree-bounded minimum diameter tree) as an overlay network that involves all the participants of the application. For a given set of nodes where some of them are senders, s-DBMDT is a spanning tree where the maximum delay on the tree from those senders is minimized and the degree constraint on each node is held. We believe that this is the first approach that defines s-DBMDT construction problem and presents a distributed protocol for the purpose. Our performance evaluation is based on experiments in both simulated networks and real networks that strongly shows the efficiency and usefulness of STS protocol.

In this paper, we propose a probabilistic rule-based model to describe behavior of mobile nodes for accurately evaluating performance of MANET applications. The proposed model allows us to describe how mobile nodes change their destinations, routes and speeds/directions based on their positions, surroundings (e.g. neighboring nodes), information obtained from applications, and so on. We have designed and developed a network simulator called MobiREAL based on the proposed methodology. Through experiments, we show importance to simulate MANET applications in realistic environments.

In order for precise evaluation of MANET applications, more realistic mobility models are needed in wireless network simulations. In this paper, we focus on the behavior of pedestrians in urban areas and propose a new method to generate a mobility scenario called Urban Pedestrian Flows (UPF). In the proposed method, we classify pedestrians in a simulation field into multiple groups by their similar behavior patterns (simply called flows hereafter, which indicate how they move around geographic points). Given the observed road density in the target field, we derive using linear programming techniques how many pedestrians per minute follow each flow. Using the derived flows, we generate a UPF scenario which can be used in network simulators. In particular, we have enhanced a network simulator called MobiREAL, which has been developed in our research group, so that we can generate and use the UPF scenario. MobiREAL simulator has three main facilities: the behavior simulator, network simulator and animator. The behavior simulator can generate/delete mobile nodes according to the UPF scenario. The network simulator can simulate MANET protocols and applications. The animator offers elegant visualization of simulation traces as well as graphical user interfaces for facilitating derivation of UPF scenarios. Through several case studies, we show similarity of the derived flows to the observed ones, as well as the metrics that characterize the mobility of the scenario.

Service overlay networks have recently attracted tremendous interests. In this paper, we propose a new integrated framework for specifying services composed of service components running on different service nodes and for executing the services considering efficient utilization of overlay network resources. For a given service description written in an extended Petri net model, our method automatically derives a set of descriptions of service nodes' behavior which specifies how service nodes on an overlay network collaborate to provide the specified services. The derived descriptions minimize channel utilization, total response time or load of service nodes based on a given cost criterion. The experimental results show that a multimedia service for decorating and transcoding video contents can be well specified and implemented.

Here, we present a protocol for dynamically maintaining a degree-bounded delay sensitive spanning tree in a decentralized way on overlay networks. The protocol aims at repairing the spanning tree autonomously even if multiple node's leave operations or failures (disappearances) occur simultaneously or continuously in a specified period. It also aims at maintaining the diameter (maximum delay) of the tree as small as possible. The simulation results using ns-2 have shown that the protocol could keep reasonable diameters compared with the existing centralized static algorithm even if many node's participations and disappearances occur frequently.

In this paper, we propose a location-aware multicast protocol on MANET called MgCast (multiCAST or multiple geographical regions). Given a source node and a set of geographical destination regions, MgCast constructs and maintains a routing tree from the source to nodes which reside in the regions, in a decentralized manner. Our aim is to pursue trade-off between the route discovery ratio, the number of route request messages and the number of links of the tree. The experimental results have shown that MgCast could achieve a good balance between these metrics.

In this paper, we present a new application level multicast protocol called Emma (end-user multicast for multi-party applications) suitable for communication systems where multiple video sources are exchanged in real-time among end-hosts, such as video-conferencing. The primal feature of Emma is that video sources with the higher priority given by users are prioritized among others for the provision of quality of service at the user level and that all the operations in Emma are done in a distributed manner. Our experimental results have shown that Emma achieves reasonable performance on overlay networks with high user satisfaction.

In this paper, we propose a selection technique for replicated multicast video servers. We assume that each replicated video server transmits the same video source as different quality levels of multicast streams. Using an IGMP facility like m-trace, each receiver monitors packet count information of those streams on routers and periodically selects the one which is expected to provide low loss rate and to be suitable for the current available bandwidth of receivers. Moreover, collection of packet count information is done in a scalable and efficient manner by sharing the collected information across receivers. Our experimental results using the network simulator have shown that our method could achieve much higher quality satisfaction of receivers, under the reasonable amount of tracing traffic.

In this paper, we present a method for the synthesis of extended Petri net-based distributed specifications. Our method finds an optimal allocation of resources (computational data) that optimizes the derived distributed specification, based on some reasonable communication-cost criteria.

In this paper, we propose a receiver-cooperative bandwidth management method for layered multicast streams, considering not only bandwidth requirements but also receivers' preference. In the method, we assume that each receiver has a preference value for each layer of streams. When a receiver requests an additional layer and required bandwidth is not available on links, it can let other receivers release a part of the layers of streams which they receive if it increases the sum of the satisfied preference values of all receivers as a total. We give an algorithm to calculate an optimal way of releasing layers in a polynomial time under some assumption. We did an experiment to transmit JPEG video over a private IP network and confirmed that the method could control bandwidth among the video streams within a second.

In this paper, we propose a new bandwidth allocation technique where a new stream can preempt an appropriate amount of bandwidth from other existing streams, considering both quality and priority requirements of users. The existing preemption-based technique has focused on the preemption among the streams with the same user. However, in multi-party multimedia applications, there would be a case where we would like to raise the quality of a specific stream (e.g. in a video conference, the video stream of a new chair may become more important than those of the existing participants). Therefore the preemption should be allowed among streams with different users and the best preemption in terms of their requirements should be provided. In our technique, using the algorithm for solving the minimum flow cost problem, the preemption is calculated so that the total loss of quality and priority of the existing streams can be minimized. We have implemented the proposed technique and evaluated it through the experiment using MPEG1 video streams, and have confirmed that our technique can keep a high frame rate for each existing MPEG1 video stream even when accommodating a lot of new streams.

Some methods for deriving protocol specifications from given service specifications with time constraints have been proposed. However, existing methods cannot treat the class of service specifications with both parallel synchronization and data values. They also assume that all clocks in the distributed system are synchronized. We propose an algorithm to derive a correct protocol specification automatically from a given service specification described in an extended model of time Petri nets where the above restrictions are eliminated. Using our method, we will be free from considering the details of communication delays on the design of real-time distributed systems.

In general, the services of a distributed system are provided by some cooperative protocol entities. The protocol entities must exchange some data values and synchronization messages in order to ensure the temporal ordering of the events which are described in a service specification of the distributed system. It is desirable that a correct protocol entity specification for each node can be derived automatically from a given service specification. In this paper, we propose an algorithm which synthesizes a correct protocol entity specification automatically from a service specification in a Petri Net model with Registers called PNR model. In our model, parallel events and selective operations can be described naturally. The control flow of a service specification must be described as a free-choice net in order to simplify the derivation algorithm, however, many practical systems can be described in this class. In our approach, since each protocol entity specification is also described in our PNR model, we can easily understand what events can be executed in parallel at each protocol entity.