Efficient LLM inference on resource-constrained devices presents significant challenges in compute and memory utilization. Due to limited GPU memory, existing systems offload model weights to CPU memory, incurring substantial I/O overhead between the CPU and GPU. This leads to two major inefficiencies: (1) GPU cores are underutilized, often remaining idle while waiting for data to be loaded; and (2) GPU memory has low impact on performance, as reducing its capacity has minimal effect on overall throughput.In this paper, we propose SpecOffload, a high-throughput inference engine that embeds speculative decoding into offloading. Our key idea is to unlock latent GPU resources for storing and executing a draft model used for speculative decoding, thus accelerating inference at near-zero additional cost. To support this, we carefully orchestrate the interleaved execution of target and draft models in speculative decoding within the offloading pipeline, and propose a planner to manage tensor placement and select optimal parameters. Compared to the best baseline, SpecOffload improves GPU core utilization by 4.49x and boosts inference throughput by 2.54x. Our code is available at https://github.com/MobiSense/SpecOffload-public .

Grounding natural language instructions to visual observations is fundamental for embodied agents operating in open-world environments. Recent advances in visual-language mapping have enabled generalizable semantic representations by leveraging visionlanguage models (VLMs). However, these methods often fall short in aligning free-form language commands with specific scene instances, due to limitations in both instance-level semantic consistency and instruction interpretation. We present OpenMap, a zero-shot open-vocabulary visual-language map designed for accurate instruction grounding in navigation tasks. To address semantic inconsistencies across views, we introduce a Structural-Semantic Consensus constraint that jointly considers global geometric structure and vision-language similarity to guide robust 3D instancelevel aggregation. To improve instruction interpretation, we propose an LLM-assisted Instruction-to-Instance Grounding module that enables fine-grained instance selection by incorporating spatial context and expressive target descriptions. We evaluate OpenMap on ScanNet200 and Matterport3D, covering both semantic mapping and instruction-to-target retrieval tasks. Experimental results show that OpenMap outperforms state-of-the-art baselines in zero-shot settings, demonstrating the effectiveness of our method in bridging free-form language and 3D perception for embodied navigation.

Optical motion capture is a foundational technology driving advancements in cutting-edge fields such as virtual reality and film production. However, system performance suffers severely under large-scale marker occlusions common in real-world applications. An in-depth analysis identifies two primary limitations of current models: (i) the lack of training datasets accurately reflecting realistic marker occlusion patterns, and (ii) the absence of training strategies designed to capture long-range dependencies among markers. To tackle these challenges, we introduce the CMU-Occlu dataset, which incorporates ray tracing techniques to realistically simulate practical marker occlusion patterns. Furthermore, we propose OpenMoCap, a novel motion-solving model designed specifically for robust motion capture in environments with significant occlusions. Leveraging a marker-joint chain inference mechanism, OpenMoCap enables simultaneous optimization and construction of deep constraints between markers and joints. Extensive comparative experiments demonstrate that OpenMoCap consistently outperforms competing methods across diverse scenarios, while the CMU-Occlu dataset opens the door for future studies in robust motion solving. The proposed OpenMoCap is integrated into the MoSen MoCap system for practical deployment. The code is released at: https://github.com/qianchen214/OpenMoCap.

Flexible manufacturing is one of the core goals of Industry 4.0 and brings new challenges to current industrial control systems. Our detailed field study on auto glass industry revealed that existing production lines are laborious to reconfigure, difficult to upscale, and costly to upgrade during production switching. Such inflexibility arises from the tight coupling of devices, controllers, and control tasks. In this work, we propose a new architecture for industrial control systems named Control-as-a-Service (CaaS). CaaS transfers and distributes control tasks from dedicated controllers into network switches. By combining control and transmission functions in switches, CaaS virtualizes the whole industrial network to one Programmable Logic Controller (PLC). We propose a set of techniques that realize end-to-end determinism for in-network industrial control and a joint task and traffic scheduling algorithm. We evaluate the performance of CaaS on testbeds based on real-world networked control systems. The results show that the idea of CaaS is feasible and effective, and CaaS achieves absolute packet delivery, 42-45% lower latency, and three orders of magnitude lower jitter. We believe CaaS is a meaningful step towards the distribution, virtualization, and servitization of industrial control.

Time-Sensitive Networking (TSN) is foreseen as a foundational technology that enables Industry 4.0. It offers deterministic data transmission over Ethernet for critical applications such as industrial control and automotive systems. However, TSN is susceptible to hardware and software errors, necessitating an effective diagnostic system. Traditional network diagnostic tools are inadequate for TSN fault localization and classification due to the tightly coupled traffic and high precision requirements in TSN. In response, this paper presents TSNCard, a cross-cycle postcard-based diagnostic system tailored for Time-Aware Shaper (IEEE 802.1 Qbv) in TSN. TSNCard introduces a novel telemetry protocol that leverages the cyclical nature of TSN networks for data collection at each node. This protocol, coupled with dedicated analytic algorithms and hardware innovations within switches, forms a comprehensive system for TSN monitoring, fault localization and classification. Extensive experiments on both simulation and physical testbeds show that TSNCard can 100% detect fault location and type of the TSN misbehavior while adhering to industrial bandwidth restrictions. TSNCard not only bridges the gap in the TSN protocol stack, but also serves as a versatile toolkit for time-synchronized network analysis, paving the way for future research. The code is available athttps://github.com/MobiSense/TSNCard

Recent years have witnessed an increasing demand for human fall detection systems. Among all existing methods, Wi-Fi-based fall detection has become one of the most promising solutions due to its pervasiveness. However, when applied to a new domain, existing Wi-Fi-based solutions suffer from severe performance degradation caused by low generalizability. In this paper, we propose XFall, a domain-adaptive fall detection system based on Wi-Fi. XFall overcomes the generalization problem from three aspects. To advance cross-environment sensing, XFall exploits an environment-independent feature called speed distribution profile, which is irrelevant to indoor layout and device deployment. To ensure sensitivity across all fall types, an attention-based encoder is designed to extract the general fall representation by associating both the spatial and temporal dimensions of the input. To train a large model with limited amounts of Wi-Fi data, we design a cross-modal learning framework, adopting a pre-trained visual model for supervision during the training process. We implement and evaluate XFall on one of the latest commercial wireless products through a year-long deployment in real-world settings. The result shows XFall achieves an overall accuracy of 96.8%, with a miss alarm rate of 3.1% and a false alarm rate of 3.3%, outperforming the state-of-the-art solutions in both in-domain and cross-domain evaluation.

Environment depth estimation by fusing camera and radar enables a broad spectrum of applications such as autonomous driving, environmental perception, context-aware localization and navigation. Various pioneering approaches have been proposed to achieve accurate and dense depth estimation by integrating vision and LiDAR through deep learning. However, due to the challenges of sparse sampling of in-vehicle LiDARs, high ground-truth annotation overhead, and severe dynamics in real environments, existing solutions have not yet achieved widespread deployment on commercial autonomous vehicles. In this paper, we propose LeoVR, a motion-inspired self-supervised visual-LiDAR fusion approach that enables accurate environment depth estimation. Leveraging the vehicle motion information, LeoVR employs two effective system frameworks to$(i)$optimize the depth estimation results, and$(ii)$provide supervision signals for DNN training. We fully implemented LeoVR on both a robotic testbed and a commercial vehicle and conducted extensive experiments over an 8-month period. The results demonstrate that LeoVR achieves remarkable performance with an average depth estimation error of 0.17$m$, outperforming existing state-of-the-art solutions by$\gt $45.9%. Besides, even cold-start in real environments by self-supervised training, LeoVR still achieves an average error of 0.2$m$, outperforming the related works by$\gt $47.8% and comparable to supervised training methods.

Edge-assisted visual SLAM plays a crucial role in enabling innovative mobile applications, such as autonomous swarm inspection, search-and-rescue, and smart logistics. Constrained by the computational capacities of lightweight mobile devices, current approaches delegate lightweight, time-sensitive tracking tasks to the mobile end while offloading resource-intensive, latency-tolerant map optimization tasks to the edge. However, our pilot study reveals several limitations of the tracking-optimization decoupled paradigm, stemming from the disruption of inter-dependencies between the two tasks. In this paper, we design and implement edgeSLAM2, an innovative system that reshapes the edge-assisted visual SLAM paradigm by tightly integrating tracking and partial-yet-crucial optimization on mobile. edgeSLAM2 harnesses the heterogeneous computing units offered by the commercial systems-on-chip (SoCs) to enhance the computational capacity of mobile devices, which in turn, allows edgeSLAM2 to design a suit of novel algorithms for map sync, optimization, and tracking that accommodate such architectural upgrade. By capitalizing on the full potential of on-chip intelligence, edgeSLAM2 supports both solitary and collaborative SLAM with accuracy and immediacy, underpinned by a cohesive software-hardware co-design. We deploy edgeSLAM2 on drones for industrial inspection. Comprehensive experiments in one of the world’s largest oil fields over three months demonstrate its superior performance.

Most existing multi-user Augmented Reality (AR) systems only support multiple co-located users to view a common set of virtual objects but lack the ability to enable each user to directly interact with other users appearing in his/her view. Such multi-user AR systems should be able to detect the human keypoints and estimate device poses (for identifying different users) in the meantime. However, due to the stringent low latency requirements and the intensive computation of the preceding two capabilities, previous research only enables either of the two capabilities for mobile devices even with the aid of the edge server. Integrating the two capabilities is promising but non-trivial in terms of latency, accuracy, and matching. To fill this gap, we propose DiTing to achieve real-time ID-aware multi-device visual interaction for multi-user AR applications, which contains three key innovations: Shared On-device Tracking to merge the similar computation for optimized latency, Tightly Coupled Dual Pipeline to enhance the accuracy of each task through mutual assistance, and Body Affinity Particle Filter to precisely match device poses with human bodies. We implement DiTing on four types of mobile AR devices and develop a multi-user AR game as a case study. Extensive experiments show that DiTing can provide high-quality human keypoint detection and pose estimation in real time (30fps) for ID-aware multi-device interaction and outperform the state-of-the-art baseline approaches.

This article introduces LinkStream, a liquidity analysis system based on multiple video streams designed and implemented for oilfield. LinkStream combines a variety of technologies to solve several problems in computing power and network latency. First, the system adopts an edge-central architecture and tailoring based on spatio-temporal correlation, which greatly reduces computing power requirements and network costs, and enables real-time analysis of large-scale video stream on limited edge devices. Second, it designed a set of liquidity information to describe the liquidity status in the oilfield. Finally, it uses object tracking technology to design a counting algorithm for the unique tubing object in the oilfield. We have deployed LinkStream in an oilfield in Iraq. LinkStream can perform real-time inference on over 200 video streams with acceptable resource overhead.

LoRa and its enabled LoRa wide-area network (LoRaWAN) have been seen as an important part of the next-generation network for massive Internet-of-Things (IoT). Due to LoRa's low-power and long-range nature, LoRa signals are much weaker than the noise floor, particularly in complex urban or semi-indoor environments. Therefore, weak signal decoding is critical to achieve the desired wide-area coverage in general. Existing work has shown the advantages of exploring deep neural networks (DNN) for weak signal decoding. However, the existing single-gateway based DNN decoder is hard to fully leverage the spatial information in multi-gateway scenarios. In this paper, we propose SRLoRa, an efficient DNN LoRa decoder that fully utilizes the spatial information from multiple gateways to decode extremely weak LoRa signals. Specifically, we design interleaving denoising and merging layers to improve signal quality at ultra-low SNR. We develop efficient merging on feature maps extracted by denoising DNNs to tolerate time misalignments among different signals. We define max and min operations in the merging layer to efficiently extract salient features and reduce noise, merging the features extracted from multiple gateways to guide future DNN layers to gradually improve signal quality. We implement SRLoRa with USPR N210 and commercial LoRa nodes and evaluate its performance indoors and outdoors. The results show that with four gateways, SRLoRa achieves SNR gain at 4.53--4.82 dB, which is 2.51× of Charm, leading to a 1.84× coverage area compared to standard LoRa in an urban deployment.

Indoor navigation is essential to a wide spectrum of applications in the era of mobile computing. Existing vision-based technologies suffer from both start-up costs and the absence of semantic information for navigation. We observe an opportunity to leverage pervasively deployed surveillance cameras to deal with the above drawbacks and revisit the problem of indoor navigation with a fresh perspective. In this paper, we proposeiSAT, a system that enables public surveillance cameras, as indoor navigating satellites, to locate users on the floorplan, tell users with semantic information about the surrounding environment, and guide users with navigation instructions. However, enabling public cameras to navigate is non-trivial due to 3 factors: absence of real scale, disparity of camera perspective, and lack of semantic information. To overcome these challenges,iSATleverages POI-assisted framework and adopts a novel coordinate transformation algorithm to associate public and mobile cameras, and further attaches semantic information to user location. Extensive experiments in 4 different scenarios show thatiSATachieves a localization accuracy of 0.48m and a navigation success rate of 90.5 percent, outperforming the state-of-th-art systems by$> 30\%$. Benefiting from our solution, all areas with public cameras can upgrade to smart spaces with visual navigation services.

CNN based real-time object detection can facilitate various AI applications that need to understand the surroundings via camera, such as autonomous package delivery robots, augmented reality, and intelligent drone applications. Currently, due to the high computation cost of CNN, accurate real-time object detection is only possible when mobile devices can upload video frames to powerful edge servers through high-speed wireless networks like WiFi. However, for many open-air AI applications, the network conditions (such as cellular networks) are usually unfavorable, far from satisfying the network demands of state-of-the-art systems. In this paper, we focus on the challenges incurred by mobile communication networks and propose WAVEcontaining three novel techniques, which areDeep RoI Encoding,Prioritized Parallel OffloadingandFine-grained Offloading Strategy, to realizereal-time,robustandlow-costobject detection for open-air AI applications. The experimental results show that under LTE networks, WAVE realizes high-accuracy real-time object detection and face recognition and significantly outperforms state-of-the-art systems.

Real-time mobile video analysis like object detection and tracking is key to various household applications such as AR, cognitive assistance and smart home. Such applications rely on heavy DNN models, which are not suitable for mobile devices due to resource limitation. The long latency of cloud offloading is unacceptable for the real-time requirements, and the direct edge offloading relies on powerful edge servers, which is impractical for household scenarios. To solve this challenge, we take advantage of the computing devices that are low-cost or already exist in our lives, and propose Trine, a cloud-edge-device cooperated framework, in which complicated computation tasks are offloaded from the device to the cloud with the edge as the key bond to coordinate. In addition, due to the heterogeneity of edge devices, which leads to no one-fits-all algorithm that is optimal in all situations, we propose a profile-based algorithm to customize trackers for various edge devices. We implemented Trine on an android phone and three edge devices. The experiments demonstrate that Trine achieves 8-36% higher real-time accuracy and 25-89% higher robustness than state-of-the-art.

Accurate, real-time object detection on resource-constrained devices enables autonomous mobile vision applications such as traffic surveillance, situational awareness, and safety inspection, where it is crucial to detect both small and large objects in crowded scenes. Prior studies either perform object detection locally on-board or offload the task to the edge/cloud. Local object detection yields low accuracy on small objects since it operates on low-resolution videos to fit in mobile memory. Offloaded object detection incurs high latency due to uploading high-resolution videos to the edge/cloud. Rather than either pure local processing or offloading, we propose to detect large objects locally while offloading small object detection to the edge. The key challenge is to reduce the latency of small object detection. Accordingly, we develop EdgeDuet, the first edge-device collaborative framework for enhancing small object detection with tile-level parallelism. It optimizes the offloaded detection pipeline in tiles rather than the entire frame for high accuracy and low latency. Evaluations on drone vision datasets under LTE, WiFi 2.4GHz, WiFi 5GHz show that EdgeDuet outperforms local object detection in small object detection accuracy by 233.0%. It also improves the detection accuracy by 44.7% and latency by 34.2% over the state-of-the-art offloading schemes.

For energy constraint wireless IoT nodes, their radios usually operate in duty cycle mode. With low maintenance and negotiation cost, asynchronous duty cycle radio management is widely adopted. To achieve fast network flooding is challenging in asynchronous duty cycle networks. Recently, concurrent flooding, which allows a set of nodes (called concurrent senders ) to immediately broadcast the received packet without any backoff, is a promising approach to improve the flooding speed. We observe that selecting either large or small number of concurrent senders cannot achieve the optimal flooding speed in different deployments. There is a tradeoff between the degradation of concurrent broadcast efficiency and the missing of early receiving chance. In this article, we propose COFlood (Concurrent Opportunistic Flooding), a practical and efficient concurrent flooding protocol in asynchronous duty cycle networks. First, COFlood constructs a concurrent flooding tree in distributed manner. The non-leaf nodes are selected as concurrent senders and they can cover the entire network while reserving the most capacity of concurrent broadcast for later added opportunistic concurrent senders. Moreover, we find that exploiting both early wake-up nodes and long lossy links can speed up the concurrent flooding tree-based network flooding by increasing the early receiving chances. Then, COFlood develops a lightweight method to select the nodes that meet the conditions of these two opportunities as opportunistic concurrent senders. We implement COFlood in TinyOS and evaluate it on two real testbeds. In comparison with state-of-the-art concurrent flooding protocol, completion time and energy consumption can be reduced by up to 35.3% and 26.6%.

Accurate and real-time instance segmentation on mobile devices enables a wide spectrum of applications such as augmented reality, context-aware inspection and environ-mental cognition. However, the computation resource demanded by instance segmentation impedes its deployment on resource-constrained commercial mobile devices. Prior studies enable smartphones to conduct computational-intensive tasks in real-time with the assistance of an edge server. However, simply applying an edge-assisted framework hardly achieves delightful segmentation performance due to the movements of devices and targets, pixel-level precision requirements, and huge computational overhead even for edge nodes. This work proposes edgeIS, an edge-assisted system that enables real-time and accurate instance segmentation on mobile devices. edgeIS embraces the mobile device sensing ability of surroundings and its own motion, and redesigns an innovative mobile-edge collaboration paradigm suitable for segmentation tasks. We implement edgeIS on a lightweight edge node and different mobile devices. Extensive experiments are conducted under four datasets. The results show that edgeIS can run on mobile devices in real-time and achieve a 0.92 segmentation IoU, outperforming existing state-of-the-art solutions. We further embed edgeIS in an AR-based inspection system deployed in an oil field and the performance of edgeIS meets the demand of the industrial scenario.

Most existing multi-user Augmented Reality (AR) systems only support multiple co-located users to view a common set of virtual objects but lack the ability to enable each user to directly interact with other users appearing in his/her view. Such multi-user AR systems should be able to detect the human keypoints and share device poses (for identifying different users) in the meanwhile. However, due to the stringent low latency requirements and the intensive computation of the above two capabilities, previous research only enables either of the two capabilities for mobile devices even with the aid of the edge server. Integrating the above two capabilities is promising but non-trivial in terms of latency, accuracy, and matching. To fill this gap, we propose DiTing to achieve real-time ID-aware multi-device visual interaction for multi-user AR applications, which contains three key innovations: Shared On-device Tracking to merge the similar computation for optimized latency, Tightly Coupled Dual Pipeline to enhance the accuracy of each task through mutual assistance, Body Affinity Particle Filter to precisely match device poses with human bodies. We implement DiTing on four types of mobile AR devices and develop a multi-user AR game as a case study. Extensive experiments show that DiTing can provide high-quality human keypoint detection and pose estimation in real-time (30fps) for ID-aware multi-device interaction and outperform the SOTA baseline approaches.

Environment depth estimation by fusing camera and radar enables a broad spectrum of applications such as autonomous driving, environmental perception, context-aware localization and navigation. Various pioneering approaches have been proposed to achieve accurate and dense depth estimation by integrating vision and LiDAR through deep learning. However, due to the challenges of sparse sampling of in-vehicle LiDARs, high ground-truth annotation overhead, and severe dynamics in real environments, existing solutions have not yet achieved widespread deployment on commercial autonomous vehicles. In this paper, we propose LeoVR, a visual-LiDAR fusion based self-supervised approach that enables accurate environment depth estimation. LeoVR digs into the vehicle's motion information and designs two effective system frameworks based on it to (i) optimize the depth estimation results, and (ii) provide supervision signals to train a DNN. We fully implement LeoVR on a robotic testbed and commercial vehicle to conduct extensive experiments across 6 months. The results demonstrate that LeoVR achieves remarkable performance with an average depth estimation error of 0.17m, outperforming existing state-of-the-art solutions by > 43%. Besides, even cold-start in real environments by self-supervised training, LeoVR still achieves an average error of 0.21m, outperforming the related works by > 45% and comparable to those supervised training methods.

Intelligent personal and home applications demand multiple deep neural networks (DNNs) running on resourceconstrained platforms for compound inference tasks, known as multitask inference. To fit multiple DNNs into low-resource devices, emerging techniques resort to weight sharing among DNNs to reduce their storage. However, such reduction in storage fails to translate into efficient execution on common accelerators such as GPUs. Most DNN graph rewriters are blind for multi-DNN optimization, while GPU vendors provide inefficient APIs for parallel multi-DNN execution at runtime. A few prior graph rewriters suggest cross-model graph fusion for low-latency multi-DNN execution. Yet they request duplication of the shared weights, erasing the memory saving of weight-shared DNNs. In this paper, we propose MTS, a novel graph rewriter for efficient multitask inference with weight-shared DNNs. MTS adopts a model stitching algorithm which outputs a single computational graph for weight-shared DNNs without duplicating any shared weight. MTS also utilizes a model grouping strategy to avoid overwhelming the GPU when co-running tens of DNNs. Extensive experiments show that MTS accelerates multitask inference by up to 6.0× compared to sequentially executing multiple weightshared DNNs. MTS also yields up to 2.5× lower latency and 3.7× less memory usage compared with NETFUSE, a state-of-the-art multi-DNN graph rewriter.

Mobile video streaming occupies three-quarters of today's cellular network traffic. The quality of mobile videos becomes increasingly important for video providers to attract more users. For example, they invest in network bandwidth resources and conduct adaptive bitrate techniques to improve video quality. Prior adaptive bitrate (ABR) algorithms perform well under given throughput traces on broadband and WiFi networks. They may perform poorly for mobile video streaming due to the high network dynamics of cellular networks. To study the properties of throughput traces under cellular networks, we collect 4G network throughput traces for over four months in two large cities, Beijing and Suzhou in China. We derive the environment-specific Markov property of throughputs in the dataset. Accordingly, we propose NEIVA, an environment identification based technique to adaptively predict future throughput for different types of environments. We also implement NEIVA and integrate it with the state-of-the-art ABR algorithm, model predictive control (MPC) approach in our testbed for experiments. By emulating mobile video streaming under throughput traces in our dataset, NEIVA achieves 20 - 25 percent improvement on throughput prediction accuracy comparing to baseline predictors. Meanwhile, NEIVA achieves 11 - 20 percent user QoE improvement over MPC with baseline predictors.

Motivated by safety challenges resulting from distracted pedestrians, we design and implement VSpace, a system that achieves spatially-aware sensing of oncoming vehicles based on the Doppler Effect of vehicle sound. VSpace can figure out the natural velocity, direction, and minimum distance of an oncoming vehicle by acoustic Doppler profiling online. VSpace works in four major steps: (1)Vehicle acoustic signal processing; (2)Vehicle classification; (3)Vehicle sound Doppler profiling and (4)Vehicle spatial status reconstruction. All implementation of VSpace is based on a COTS smartphone with a microphone. We evaluate VSpace with comprehensive experiments of 20 different vehicles in 240 practical cases, in which VSpace achieves an accuracy of 91.67% in vehicle classification and 10.85%, 13.81%, 11.59% relative error for natural velocity, direction, and minimum distance calculation.

This article introduces LinkStream, a liquidity modeling system based on multiple video streams designed and implemented for oilfield. LinkStream combines a variety of technologies to solve several problems in computing power and network latency. First, the system adopts an edge-central architecture and tailoring based on spatio-temporal correlation, which greatly reduces computing power requirements and network costs, and enables real-time analysis of large-scale video stream on limited edge devices. Second, it designed a set of liquidity models to describe the liquidity status in the oilfield. Finally, it uses object tracking technology to design a counting algorithm for the unique tubing object in the oilfield. We have deployed LinkStream in an oilfield in Iraq. LinkStream can perform real-time inference on over 200 video streams with acceptable resource overhead.

Existing smartphone-based Augmented Reality (AR) systems are able to render virtual effects on static anchors. However, today's solutions lack the ability to render follow-up effects attached to moving anchors since they fail to track the 6 degrees of freedom (6-DoF) poses of them. We find an opportunity to accomplish the task by leveraging sensors capable of generating sparse point clouds on smartphones and fusing them with vision-based technologies. However, realizing this vision is non-trivial due to challenges in modeling radar error distributions and fusing heterogeneous sensor data. This study proposes FollowUpAR, a framework that integrates vision and sparse measurements to track object 6-DoF pose on smartphones. We derive a physical-level theoretical radar error distribution model based on an in-depth understanding of its hardware-level working principles and design a novel factor graph competent in fusing heterogeneous data. By doing so, FollowUpAR enables mobile devices to track anchor's pose accurately. We implement FollowUpAR on commodity smartphones and validate its performance with 800,000 frames in a total duration of 15 hours. The results show that FollowUpAR achieves a remarkable rotation tracking accuracy of 2.3° with a translation accuracy of 2.9mm, outperforming most existing tracking systems and comparable to state-of-the-art learning-based solutions. FollowUpAR can be integrated into ARCore and enable smartphones to render follow-up AR effects to moving objects.

For energy constrained wireless IoT nodes, their radios usually operate in duty cycle mode. With low maintenance and negotiation cost, asynchronous duty cycle radio management is widely adopted. To achieve fast network flooding is challenging in asynchronous duty cycle networks. Recently, concurrent flooding is a promising approach to improve the performance of network flooding. In concurrent flooding, a key challenge is how to select a set of concurrent senders to improve both flooding speed and energy efficiency. We observe that selecting neither large nor small number of concurrent senders can achieve the optimal performance in different deployments. In this paper, we propose COFlood (Concurrent Opportunistic Flooding), a practical and effective concurrent flooding protocol in asynchronous duty cycle networks. The basic idea is based on an energy-efficient flooding tree, COFlood opportunistically selects extra concurrent senders that can speed up network flooding. First, COFlood constructs an energy-efficient flooding tree in distributed manner. The non-leaf nodes are selected as senders and they can cover the entire network with low energy consumption. Moreover, we find that exploiting both early wakeup nodes and long lossy links can speed up the flooding tree based network flooding. Then, COFlood develops a light-weight method to select the nodes that meet the conditions of these two opportunities as opportunistic senders. We implement COFlood in TinyOS and evaluate it on two real testbeds. In comparison with state-of-the-art concurrent flooding protocol, completion time and energy consumption can be reduced by up to 35.3% and 26.6%.

Due to limited energy supply on many Internet of Things (IoT) devices, asynchronous duty cycle radio management is widely adopted to save energy. Flooding is a critical way to disseminate messages through the whole network. Capture effect enabled concurrent broadcast is appealing to accelerate network flooding in asynchronous duty cycle networks. However, when the flooding payload's size is large, the concurrent broadcast performance is far from efficient due to the frequently unsatisfied capture effect. Intuitively, senders can send a short packet containing partial flooding payload to keep concurrent broadcast efficiency. In practice, we still face two challenges. Considering packet loss, a receiver needs an effective way to recover the entire flooding payload from several received packets as soon as possible. Moreover, considering different channel states of different senders, how a sender chooses the optimal packet length to guarantee high channel utilization is not easy. In this paper, we propose Chase++ a Fountain-code based concurrent broadcast control layer to enable fast flooding in asynchronous duty cycle networks. Chase++ uses Fountain code to alleviate the negative influence of a certain part of the flooding payload's continuous loss. Moreover, Chase++ adaptively selects packet length with the local estimation of channel utilization. Specifically, Chase++ partitions long payload into several short payload blocks, further encoded into many encoded payload blocks by Fountain-code. Then, with temporal and spatial features of the sampled RSS (received signal strength) sequence, a sender estimates the number of concurrent senders. Finally, according to the estimated number of concurrent senders, the sender determines the optimal number of encoded payload blocks in a packet and assembles the encoded payload blocks as lots of packets. Then, the concurrent broadcast layer continuously transmits these packets. Receivers can recover the original flooding payload after several independent encoded payload blocks are collected. We implement Chase++ in TinyOS with TelosB nodes. We further evaluate Chase++ on Local testbed with 50 nodes and Indriya testbed with 95 nodes. The improvement of network flooding speed can reach 23.6% and 13.4%, respectively.

Smartphone localization is essential to a wide spectrum of applications in the era of mobile computing. The ubiquity of smartphone mobile cameras and surveillance ambient cameras holds promise for offering sub-meter accuracy localization services thanks to the maturity of computer vision techniques. In general, ambient-camera-based solutions are able to localize pedestrians in video frames at fine-grained, but the tracking performance under dynamic environments remains unreliable. On the contrary, mobile-camera-based solutions are capable of continuously tracking pedestrians; however, they usually involve constructing a large volume of image database, a labor-intensive overhead for practical deployment. We observe an opportunity of integrating these two most promising approaches to overcome above limitations and revisit the problem of smartphone localization with a fresh perspective. However, fusing mobile-camera-based and ambient-camera-based systems is non-trivial due to disparity of camera in terms of perspectives, parameters and incorrespondence of localization results. In this article, we propose iMAC, an integrated mobile cameras and ambient cameras based localization system that achieves sub-meter accuracy and enhanced robustness with zero-human start-up effort. The key innovation of iMAC is a well-designed fusing frame to eliminate disparity of cameras including a construction of projection map function to automatically calibrate ambient cameras, an instant crowd fingerprints model to describe user motion patterns, and a confidence-aware matching algorithm to associate results from two sub-systems. We fully implement iMAC on commodity smartphones and validate its performance in five different scenarios. The results show that iMAC achieves a remarkable localization accuracy of 0.68 m, outperforming the state-of-the-art systems by >75%.

In a large-scale wireless sensor network, hundreds and thousands of sensors sample and forward data back to the sink periodically. In two real outdoor deployments GreenOrbs and CitySee, we observe that some bottleneck nodes strongly impact other nodes’ data collection and thus degrade the whole network performance. To figure out the importance of a node in the process of data collection, system manager is required to understand interactive behaviors among the parent and child nodes. So we present a management tool BOND (BOttleneck Node Detector), which explains the concept of Node Dependence to characterize how much a node relies on each of its parent nodes, and also models the routing process as a Hidden Markov Model and then uses a machine learning approach to learn the state transition probabilities in this model. Moreover, BOND can predict the network dataflow if some nodes are added or removed to avoid data loss and flow congestion in network redeployment. We implement BOND on real hardware and deploy it in an outdoor network system. The extensive experiments show that Node Dependence indeed help to explore the hidden bottleneck nodes in the network, and BOND infers the Node Dependence with an average accuracy of more than 85%.

Existing indoor navigation solutions usually require pre-deployed comprehensive location services with precise indoor maps and, more importantly, all rely on dedicatedly installed or existing infrastructure. In this article, we present Pair-Navi, an infrastructure-free indoor navigation system that circumvents all these requirements by reusing a previous traveler’s (i.e., leader) trace experience to navigate future users (i.e., followers) in a Peer-to-Peer mode. Our system leverages the advances of visual simultaneous localization and mapping ( SLAM ) on commercial smartphones. Visual SLAM systems, however, are vulnerable to environmental dynamics in the precision and robustness and involve intensive computation that prohibits real-time applications. To combat environmental changes, we propose to cull non-rigid contexts and keep only the static and rigid contents in use. To enable real-time navigation on mobiles, we decouple and reorganize the highly coupled SLAM modules for leaders and followers. We implement Pair-Navi on commodity smartphones and validate its performance in three diverse buildings and two standard datasets (TUM and KITTI). Our results show that Pair-Navi achieves an immediate navigation success rate of 98.6%, which maintains as 83.4% even after 2 weeks since the leaders’ traces were collected, outperforming the state-of-the-art solutions by >50%. Being truly infrastructure-free, Pair-Navi sheds lights on practical indoor navigations for mobile users.

Smartphone localization is essential to a wide spectrum of applications in the era of mobile computing. The ubiquity of smartphone mobile cameras and surveillance ambient cameras holds promise for offering sub-meter accuracy localization services thanks to the maturity of computer vision techniques. In general, ambient-camera-based solutions are able to localize pedestrians in video frames at fine-grained, but the tracking performance under dynamic environments remains unreliable. On the contrary, mobile-camera-based solutions are capable of continuously tracking pedestrians, however, they usually involve constructing a large volume of image database, a labor-intensive overhead for practical deployment. We observe an opportunity of integrating these two most promising approaches to overcome above limitations and revisit the problem of smartphone localization with a fresh perspective. However, fusing mobile-camera-based and ambient-camera-based systems is non-trivial due to disparity of camera in terms of perspectives, parameters and incorrespondence of localization results. In this paper, we propose iMAC, an integrated mobile cameras and ambient cameras based localization system that achieves sub-meter accuracy and enhanced robustness with zero-human start-up effort. The key innovation of iMAC is a well-designed fusing frame to eliminate disparity of cameras including a construction of projection map function to automatically calibrate ambient cameras, an instant crowd fingerprints model to describe user motion patterns, and a confidence-aware matching algorithm to associate results from two sub-systems. We fully implement iMAC on commodity smart-phones and validate its performance in five different scenarios. The results show that iMAC achieves a remarkable localization accuracy of 0.68m, outperforming the state-of-the-art systems by > 75%.

Visual peer-to-peer navigation is a suitable solution for indoor navigation for it relieves the labor of site-survey and eliminates infrastructure dependence. However, a major drawback hampers its application, as the peer-to-peer mode suffers from a deficiency of paths in large indoor scenarios with multifarious places-of-interest. Nevertheless, we propose one with a profound crowdsourcing scheme that addresses the drawback by merging the paths of different leaders' into a global map. To realize the idea, we further deal with entailed challenges, namely the unidirectional disadvantage, the scale ambiguity, and large computational overhead. We design a navigation strategy to solve the unidirectional problem and turn to VIO to tackle scale ambiguity. We devise a mobile-edge architecture to enable real-time navigation (30fps, 100ms end-to-end delay) and lighten the burden of smartphones (35% battery life for 2h35min) while assuring the accuracy of localization and map construction. Through experimental validations, we show that P2P navigation, previously relying on the abundance of independent paths, can enjoy a sufficiency of navigation paths with a crowdsourced global map. The experiments demonstrate a navigation success rate of 100% and spatial offset of less than 3.2m, better than existing works.

Taxi-sharing allows occupied taxis to pick up new passengers on the fly, promising to reduce waiting time for taxi riders and increase productivity for drivers. However, it becomes more difficult to strike the balance between a driver’s profit and a passenger’s quality of service (QoS). In this article, we propose QA-Share, a QoS-aware taxi-sharing system, by addressing two important challenges. First, QA-Share maximizes driver profit and user experience at the same time. Second, QA-Share optimizes these two metrics by dynamically adapting its schedule as new requests arrive. To address these two challenges, we formulated the optimization problem using integer linear programming and derived the optimal solution under a small system scale. Moreover, we also designed a heuristic algorithm to deal with the situation where more passenger requests for taxi service come at the same time. We evaluate our approach with a real-world dataset in a Chinese city—Zhenjiang—that contains the GPS traces recorded by more than 3,000 taxis during a period of 3 months. The results show that both QoS and profit increase by 38% compared to the current schemes. Moreover, as the first study that has conducted simulations with real traces with a population of 3 million and 3,000 taxis, we prove that taxi-sharing is a viable approach in a medium-size city.

In recent years, mobile crowdsourcing has emerged as a powerful computation paradigm to harness human power to perform spatial tasks such as collecting real-time traffic information and checking product prices in a specific supermarket. A fundamental problem of mobile crowdsourcing is: When both tasks and crowd workers appear in the platforms dynamically, how to assign an appropriate set of tasks to each worker. Most existing studies focus on efficient assignment algorithms based on bipartite graph matching. However, they overlook an important fact that crowd workers might be unreliable. Thus, their task assignment schemes cannot ensure the overall quality. In this article, we investigate the Quality-aware Online Task Assignment (QAOTA) problem in mobile crowdsourcing. We propose a probabilistic model to measure the quality of tasks and a hitchhiking model to characterize workers’ behavior patterns. We model task assignment as a quality maximization problem and derive a polynomial-time online assignment algorithm. Through rigorous analysis, we prove that the proposed algorithm approximates the offline optimal solution with a competitive ratio of 10/7. Finally, we demonstrate the efficiency and effectiveness of our solution through intensive experiments.

Self-tracking at 6 degrees of freedom in real-time is essential in lots of emerging applications such as VR/AR/MR simulation, indoor navigation, and so on. With the development of built-in sensors in smartphones, many self-tracking solutions have appeared. Many researchers try to utilize vision-based approaches combined with an Inertial Measure Unit (IMU) to realize self-tracking with smartphones. After testing these approaches, however, we find that tracking would be lost in four such common scenarios: 1) When the IMU rotates fast or for a long period of time, it will cause serious delays in orientation tracking; 2) The scenes where background features are not distinct enough; 3) When the smartphone moves fast, image features become quite different in successive frames; 4) Unstructured scenes where background features are not static. To address these issues, we propose iTracker, which utilizes Real-time Step-Length Adaption Algorithm to solve the scenario (1) and a Parallel-Multi-State Local Recovery method to deal with scenarios (2)-(4). Extensive experiments show that iTracker realizes robust and accurate self-tracking in these four scenarios with an error of 0.7% throughout the whole trajectory.

Due to limited energy supply on many Internet of Things (IoT) devices, asynchronous duty cycle radio management is widely adopted to save energy. Flooding is a critical way to quickly disseminate system parameters to adapt diverse network requirements. Capture effect enabled concurrent broadcast is appealing to accelerate network flooding in asynchronous duty cycle networks. However, when the length of flooding payload is long, due to frequently unsatisfied capture effect construction, the performance of concurrent broadcast is far from efficient. Intuitively, senders can send short packet that contains partial flooding payload to keep the efficiency of concurrent broadcast. In practice, we still face two challenges. Considering packet loss, a receiver needs an effective way to recover entire flooding payload from several received packets as soon as possible. Moreover, considering diverse channel state of different senders, how a sender chooses the optimal packet length to guarantee high channel utilization in a light-weight way is not easy. In this paper, we propose Chase++ a Fountain code based concurrent broadcast control layer to enable fast flooding in asynchronous duty cycle networks. Chase++ uses Fountain code to alleviate the negative influence of the continuous loss of a certain part of flooding payload. Moreover, Chase++ adaptively selects packet length with the local estimation of channel utilization. Specifically, Chase++ partitions long payload into several short payload blocks, which are further encoded into many encoded payload blocks by Fountain code. Then, with temporal and spatial features of the sampled RSS (received signal strength) sequence, a sender estimates the number of concurrent senders. Finally, according to the estimated number of concurrent senders, the sender determines the optimal number of encoded payload blocks in a packet and assembles the encoded payload blocks as lots of packets. Then, concurrent broadcast layer continuously transmits these packets. Receivers can recover original flooding payload after several independent encoded payload blocks are collected. We implement Chase++ in TinyOS with TelosB nodes. We further evaluate Chase++ on local testbed with 50 nodes and Indriya testbed with 95 nodes. The improvement of network flooding speed can reach 23.6% and 13.4%, respectively.

Crowdsensing applications require individuals to share local and personal sensing data with others to produce valuable knowledge and services. Meanwhile, it has raised concerns especially for location privacy. Users may wish to prevent privacy leak and publish as many non-sensitive contexts as possible. Simply suppressing sensitive contexts is vulnerable to the adversaries exploiting spatio-temporal correlations in the user's behavior. In this work, we present PLP, a crowdsensing scheme which preserves privacy while it maximizes the amount of data collection by filtering a user's context stream. PLP leverages a conditional random field to model the spatio-temporal correlations among the contexts, and proposes a speed-up algorithm to learn the weaknesses in the correlations. Even if the adversaries are strong enough to know the filtering system and the weaknesses, PLP can still provably preserve privacy, with little computational cost for online operations. PLP is evaluated and validated over two real-world smartphone context traces of 34 users. The experimental results show that PLP efficiently protects privacy without sacrificing much utility.

It has been a consensus that a certain relationship exists between personal emotions and usage pattern of the smartphone. Based on users’ emotions and personalities, more and more applications are developed to provide intelligent automation services on the smartphone, such as music recommendations or stranger introductions on social networking sites. Most existing work studies this relationship by learning large amounts of samples, which are manually labeled and collected from smartphone users. The manual labeling process, however, is very time-consuming and labor-intensive. To address this issue, we propose iSelf, a system that provides a general service of automatic detection of a user’s emotions in cold-start conditions with a smartphone. With the technology of transfer learning, iSelf achieves high accuracy given only a few labeled samples. We also embed a hybrid public/personal inference engine and validation system into iSelf, to make it maintain updates continuously. Through extensive experiments in real traces, the inferring accuracy is tested above 74% and can be improved increasingly through validation and updates. The application program interface has been open online for other developers.

Backscatter communication networks receive much attention recently due to the small size and low power of backscatter nodes. As backscatter communication is often influenced by the dynamic wireless channel quality, rate adaptation becomes necessary. Most existing approaches share a common drawback: they do not distinguish channel qualities from different nodes or sub-channels. Consequently, the transmission rate may be improperly selected, resulting in low network throughput. Through extensive experimental studies, we observe that channel diversity plays a significant role in rate selection. Therefore, there are opportunities of exploiting channel diversity for better rate adaptation, improving network throughput. In this paper, we propose a Channel-Aware Rate Adaptation framework (CARA) for backscatter communication networks. By employing a lightweight channel probing scheme, we are able to obtain fine-grained channel information that enables accurate channel estimation. We further design a novel channel selection algorithm, benefiting as many backscatter nodes as possible. On each selected channel, CARA chooses data rate with respect to the node that has the best channel condition. We implement CARA on commercial readers and the experiment results show that CARA achieves up to 4× goodput gain compared with state-of-the-art rate adaptation scheme.

Over the past few years we have observed a phenomenal growth of smartphones. Smartphones are equipped with various hardware and software resources such as Bluetooth, camera and gravity sensors. If these resources are not managed appropriately, it may cause severe problems such as battery drains and system crashes. However, the specifications of resource management are usually implicit. In this paper, we investigate the problem of mining resource management specifications from off-the-shelf apps. Our key insight is that if a set of operations to a resource are frequently performed in a specific order, it must contain the specifications of how to manage the resource. We design a tool named Automatic Resource Specification Miner (ARSM), to automatically extract resource management specifications in smartphones. In our experiments, ARSM can mine tens of rules from 100 top rated Android apps within six hours. Our work is orthogonal to existing studies on diagnosing smartphone apps. With the resource management specifications discovered, ARSM can help them pinpoint more bugs in apps.

Crowdsensing applications require individuals toshare local and personal sensing data with others to produce valuableknowledge and services. Meanwhile, it has raised concernsespecially for location privacy. Users may wish to prevent privacyleak and publish as many non-sensitive contexts as possible.Simply suppressing sensitive contexts is vulnerable to the adversariesexploiting spatio-temporal correlations in users' behavior.In this work, we present PLP, a crowdsensing scheme whichpreserves privacy while maximizes the amount of data collectionby filtering a user's context stream. PLP leverages a conditionalrandom field to model the spatio-temporal correlations amongthe contexts, and proposes a speed-up algorithm to learn theweaknesses in the correlations. Even if the adversaries are strongenough to know the filtering system and the weaknesses, PLPcan still provably preserves privacy, with little computationalcost for online operations. PLP is evaluated and validated overtwo real-world smartphone context traces of 34 users. Theexperimental results show that PLP efficiently protects privacywithout sacrificing much utility.

To meet the demand of more intelligent automation services on smartphone, more and more applications are developed based on users' emotion and personality. It has been a consensus that a relationship exists between personal emotions and usage pattern of smartphone. Most of existing work studies this relationship by learning manually labeled samples collected from smartphone users. The manual labeling process, however, is time-consuming, labor-intensive and money-consuming. To address this issue, we propose iSelf, a system which provides a general service of automatic detection for user's emotions in cold-start conditions with smartphone. Using transfer learning technology, iSelf achieves high accuracy given only a few labeled samples. We also develop a hybrid public/personal inference engine and validation system, so as to make iSelf maintain continuous update. Through extensive experiments, the inferring accuracy is tested about 75% and can be improved increasingly through validation and update.

Taxi-sharing allows occupied taxis to pick up new passengers on the fly, promising to reduce waiting time for taxi riders and increase productivity for drivers. However, if not carefully designed, taxi-sharing may cause more harm than benefit - it becomes harder to strike the balance between driver's profit and passenger's quality of service (e.g. travel time, number of strangers that share a taxi, etc.). In this paper, we propose a QoS-aware taxi-sharing system design - QA-Share - by addressing two important challenges. First, QA-Share aims to maximize driver profit and user experience at the same time. Second, QA-Share continuously optimizes these two metrics by dynamically adapting its schedule as new requests arrive, without entering an oscillation state. To address these two challenges, we have formulated the optimization problem using integer linear programming, and derived the optimal solution under a small system scale. When the number of requests and taxis becomes large, we have devised a heuristic algorithm that has a much faster execution time. We have also studied how to minimize oscillations caused by schedule re-calculations by dynamically tuning the update threshold. We have evaluated our approach with real-world dataset in a Chinese city - ZhenJiang - which contains the GPS traces recorded by over 3,000 taxis during a period of three months in 2013. Our results show that the QoS and profit is increased by 38% compared to earlier schemes.

In order to simultaneously achieve good energy efficiency and high packet delivery performance, a multihop forwarding scheme should generally involve three design elements: media access mechanism, link estimation scheme, and routing strategy. Disregarding the low-duty-cycle nature of media access often leads to overestimation of link quality. Neglecting the bursty loss characteristic of wireless links inevitably consumes much more energy than necessary and underutilizes wireless channels. The routing strategy, if not well tailored to the above two factors, results in poor packet delivery performance. In this paper, we propose L2, a practical design of data forwarding in low-duty-cycle wireless sensor networks. L2addresses link burstiness by employing multivariate Bernoulli link model. Further incorporated with synchronized rendezvous, L2enables sensor nodes to work in a lazy mode, keep their radios off most of the time, and realize highly reliable forwarding by scheduling very limited packet transmissions. We implement L2on a real sensor network testbed. The results demonstrate that L2outperforms state-of-the-art approaches in terms of energy efficiency and network yield.

How to shorten the time for channel waiting is critical to avoid network contention. Traditional MAC protocols with CSMA often assume that a transmission must be deferred if the channel is busy, so they focus more on the optimization of serial transmission performance. Recent advances in physical layer, however, allows a receiver to reengage onto a stronger incoming signal from an ongoing transmission or interference, and thus shows the potential of parallel transmissions. Indeed, even if the channel is busy, a node has opportunities to carry out a successful transmission. In this study, we propose opportunistic concurrency (OPC), a new MAC layer scheme, which enables sensor nodes to capture the opportunistic concurrency and carry out parallel transmissions instead of always waiting for a clear channel. Based on local concurrency map, which encodes the interactions among different links, OPC utilizes concurrency control algorithm to make transmission decision distributedly. Our experiments on a testbed consisting of 60 TelosB sensor motes identify the transmission opportunities in WSNs with OPC. Evaluation results show that OPC achieves a 17 percent reduction in packet latency, a 9.4 percent addition in throughput and a 10 percent reduction in power consumption compared with existing approaches.

Traditional approaches for wireless sensor network diagnosis are mainly sink-based. They actively collect global evidences from sensor nodes to the sink so as to conduct centralized analysis at the powerful back-end. On the one hand, long distance proactive information retrieval incurs huge transmission overhead; On the other hand, due to the coupling effect between diagnosis component and the application itself, sink often fails to obtain complete and precise evidences from the network, especially for the problematic or critical parts. To avoid large overhead in evidence collection process, self-diagnosis injects fault inference modules into sensor nodes and let them make local decisions. Diagnosis results from single nodes, however, are generally inaccurate due to the narrow scope of system performances. Besides, existing self-diagnosis methods usually lead to inconsistent results from different inference processes. How to balance the workload among the sensor nodes in a diagnosis task is a critical issue. In this work, we present a new in-network diagnosis approach named Local-Diagnosis (LD2), which conducts the diagnosis process in a local area. LD2 achieves diagnosis decision through distributed evidence fusion operations. Each sensor node provides its own judgements and the evidences are fused within a local area based on the Dempster-Shafer theory, resulting in the consensus diagnosis report. We implement LD2 on TinyOS 2.1 and examine the performance on a 50 nodes indoor testbed.

In large-scale wireless sensor networks, faulty link detection plays a critical role in network diagnosis and management. Most potential network bottlenecks such as network partition and routing errors can be detected by link scan. Since sequentially checking all potential links incurs high transmission and storage cost, we propose a passive scheme Link Scanner (LS) for monitoring wireless links. As we know, to maintain a sensor network running in a normal condition, many applications in flooding manner are necessary, such as time synchronization, reprogramming, protocol update, etc. During such regular flooding processes that for other purposes originally, LS passively collects hop counts of received probe messages at sensor nodes. Based on the observation that faulty links can result in mismatch between received hop counts and network topology, LS deduces all links' status with a probabilistic model. We evaluate our scheme by carrying out experiments on a testbed with 60 TelosB motes and conducting extensive simulation tests. A real outdoor system is also deployed to verify that LS can be reliably applied to surveillance networks.

The emanations of electronic and mechanical devices have raised serious privacy concerns. It proves possible for an attacker to recover the keystrokes by acoustic signal emanations. Most existing malicious applications adopt context-based approaches, which assume that the typed texts are potentially correlated. Those approaches often incur a high cost during the context learning stage, and can be limited by randomly typed contents (e.g., passwords). Also, context correlations can increase the risk of successive false recognition. We present a context-free and geometry-based approach to recover keystrokes. Using off-the-shelf smartphones to record acoustic emanations from keystrokes, this design estimates keystrokes' physical positions based on the Time Difference of Arrival (TDoA) method. We conduct extensive experiments and the results show that more than 72.2\% of keystrokes can be successfully recovered.

Being embedded in the physical world, wireless sensor networks (WSNs) present a wide range of failures, due to environment conditions, hardware limitations and software uncertainties, and so on. Once deployed, the interactivity of a WSN greatly decreases, which leads to limited visibility of network performance for managers to investigate sensor behaviors. Existing evidence-based approaches aim to explain particular network symptoms based on expert knowledge and heuristic experiences, which degrade diagnosis accuracy and perform unreliably. These diagnosis models define a limited group of network failures, emphasizing on expert knowledge too much, and thus fail to be adopted to different applications. In this work, we propose VN2, a novel tool to enhance the visibility of network performance. VN2 quantifies a node's state in terms of variation of 43 metrics, and trains a representative matrix of network exceptions with Non-negative Matrix Factorization (NMF) model. With this matrix, when a new network state coming up, VN2 automatically attributes abnormal symptoms to one or more root causes. We implement VN2 on test bed and real system traces. Experimental results show that VN2 models network exceptions involving small subsets of root causes, and the interpretation of root causes help us understand network behaviors in details.

In a large-scale wireless sensor network, thousands of sensor nodes periodically generate and forward data back to the sink. In our recent outdoor deployment, we observe that some bottleneck nodes can greatly determine other nodes' data collection ratio, and thus affect the whole network performance. To figure out the importance of a node in data collection, the manager needs to understand the interactive behaviors among the parent and child nodes. To address this issue, we present a management tool BOND (Bottleneck Node Detector). We introduce the concept of Node Dependence to characterize how much a node relies on each of its parent nodes. BOND models the routing process as a Hidden Markov Model, and uses a machine learning approach to learn the state transition probabilities in this model based on the observed traces. BOND utilizes Node Dependence to explore the hidden bottleneck nodes in the network. Moreover, we can predict how adding or removing the sensor nodes would impact the data flow, thus avoid data loss and flow congestion in redeployment. We implement our tool on real hardware and deploy it in an outdoor system. Our extensive experiments show that BOND infers the Node Dependence with an average accuracy of more than 85%.

Time synchronization plays an important role in wireless sensor networks (WSNs). Due to the unique features of WSNs such as remote deployment, low-cost hardware, and restrictive energy supply, accurate and robust time synchronization is still a challenging task. Many approaches have been proposed to improve the performance and efficiency of time synchronization. Existing schemes, however, do not pay enough attention to the impacts of varying network topology properties, including network diameter, degree distribution, and the like. They can experience unexpected performance degradation in many real systems. To address these issues, we present a novel approach, which tries to improve the performance of existing time synchronization algorithms by introducing a virtual overlay. We propose an integrated model, called topo-refiner, which encodes the impact of different network topology features to the precision and robustness of time synchronization. Also, we design a novel optimization algorithm to build a virtual overlay from the underlying communication network. Our approach is orthogonal to existing clock synchronization methods, and can improve their performance by operating these methods atop the virtual layer. Finally, in order to verify the effectiveness of our approach, we conduct extensive simulations on synthetic and real system traces, as well as testbed experiments. Results show that topo-refiner is practical, and quickly adapts to varying time synchronization accuracy requirements for applications in WSNs.

Existing approaches to diagnosing sensor networks are generally sink based, which rely on actively pulling state information from sensor nodes so as to conduct centralized analysis. First, sink-based tools incur huge communication overhead to the traffic-sensitive sensor networks. Second, due to the unreliable wireless communications, sink often obtains incomplete and suspicious information, leading to inaccurate judgments. Even worse, it is always more difficult to obtain state information from problematic or critical regions. To address the given issues, we present a novel self-diagnosis approach, which encourages each single sensor to join the fault decision process. We design a series of fault detectors through which multiple nodes can cooperate with each other in a diagnosis task. Fault detectors encode the diagnosis process to state transitions. Each sensor can participate in the diagnosis by transiting the detector's current state to a new state based on local evidences and then passing the detector to other nodes. Having sufficient evidences, the fault detector achieves the Accept state and outputs a final diagnosis report. We examine the performance of our self-diagnosis tool called TinyD2 on a 100-node indoor testbed and conduct field studies in the GreenOrbs system, which is an operational sensor network with 330 nodes outdoor.

In large-scale wireless sensor networks, it proves very difficult to dynamically monitor system degradation and detect bad links. Faulty link detection plays a critical role in network diagnosis. Indeed, a destructive node impacts its links' performances including transmitting and receiving. Similarly, other potential network bottlenecks such as network partition and routing errors can be detected by link scan. Since sequentially checking all potential links incurs high transmission and storage cost, existing approaches often focus on links currently in use, while overlook those unused yet ones, thus fail to offer more insights to guide following operations. We propose a novel scheme Link Scanner (LS) for monitoring wireless links at real time. LS issues one probe message in the network and collects hop counts of the received probe messages at sensor nodes. Based on the observation that faulty links can result in mismatch between the received hop counts and the network topology, we are able to deduce all links' status with a probabilistic model. We evaluate our scheme by carrying out experiments on a testbed with 60 TelosB motes and conducting extensive simulation tests. A real outdoor system is also deployed to verify that LS can be reliably applied to surveillance networks.

The deployment of large scale Wireless Sensor Networks generally needs network management and measurement solutions. End-to-end delay is one of the most important metrics in assessing the network performance. Many efforts have been devoted to measuring the end-to-end delay efficiently and precisely. Unfortunately, existing approaches often require sensor nodes to be tightly time synchronized which is costly in resource limited sensor network. We propose a novel scheme that can measure the end-to-end delay for each packet to the granularity of tens of microseconds without clock synchronization. Through extensive experiments on our testbed, we examine the effectiveness of our approach. The results show that our scheme achieves high performance with low overhead. We also present observations about the network states by tracking and analyzing the end-to-end delay data.

Routing dynamics are intrinsic characteristics of operational wireless sensor networks (WSNs). We present the measurement and analysis results for routing dynamics in a large-scale WSN. We seek to answer several fundamental questions: How dynamically are current routing protocols performing? What causes routing dynamics? What is the impact of routing dynamics? Answers to the above questions are critical to understanding the interactions among multiple network elements, evaluating protocol design strategies, and improving system performances. However, measurements in large-scale WSNs are challenging due to the lack of dedicated log information (be analogous to configuration files, syslog messages used in Internet). We propose an approach to identify the routing dynamics based on limited information and correlate them with system events to find out the root causes. The key findings of our study include: parent change events mainly affect local nodes, i.e. They do not cause routing instability on far-away nodes, environment dynamics and routing loops have large impact on routing, small portion of parent changes might not be necessary, while a large portion of parent changes are effective in improving network performance.

In this paper, we investigate the predictability of packet arrivals in asynchronous duty-cycling wireless sensor networks (WSNs). We conduct statistical analysis on data traces collected in both outdoor large-scale and indoor testbed WSN to show that traditional well-known traffic models, e.g., Poisson process and Self-similarity process, do not fit well for modeling packet arrivals in both networks. According to our observations, some key characteristics such as sleeping interval and sampling rate have significant impact on the traffic patterns under Low power listening (LPL) models. Hence, we raise a question: could we achieve accurate prediction on the packet arrivals in asynchronous duty-cycling WSNs? To answer this question, we design a novel data-driven predictor P2IT focusing on two prediction goals: (i) the arrival time of the next packet (ii) the number of arrival packets within a short time interval. We conduct extensive trace-driven experiments to demonstrate that our predictor achieves high acuracies on both prediction goals under various experimental settings.

Many algorithms have been introduced to deterministically authenticate Radio Frequency Identification (RFID) tags, while little work has been done to address the scalability issue in batch authentications. Deterministic approaches verify them one by one, and the communication overhead and time cost grow linearly with increasing size of tags. We design a fine-grained batch authentication scheme, INformative Counting (INC), which achieves sublinear authentication time and communication cost in batch verifications. INC also provides authentication results with accurate estimates of the number of counterfeiting tags and genuine tags, while previous batch authentication methods merely provide 0/1 results indicating the existence of counterfeits. We conduct detailed theoretical analysis and extensive experiments to examine this design and the results show that INC significantly outperforms previous work in terms of effectiveness and efficiency.

In wireless sensor networks (WSNs), diagnosis is a crucial and challenging task due to the distributed nature and stringent resources. Most previous approaches are supervised, relying on a-priori knowledge of network faults. Our experience with GreenOrbs, a long-term large-scale WSN system, reveals the need of diagnosis in an agnostic manner. Specifically, in addition to predefined faults (i.e., with known types and symptoms), silent failures that are unknown beforehand, account for a large fraction of network performance degradation. Currently, there is no effective solution for silent failures because they are often diverse and highly system-related. In this paper, we propose Agnostic Diagnosis (AD), an online lightweight failure detection approach. AD is motivated by the fact that the system metrics (e.g., radio-on time, number of packets transmitted) of sensor nodes usually exhibit certain correlation patterns. Violations of such patterns indicate potential silent failures. We implement AD on a working WSN consisting of 330 nodes. Our experimental results demonstrate the advantages of AD to discover silent failures, effectively expanding the capacity and scope of WSN diagnosis.

Traditional approaches for wireless sensor network diagnosis are mainly sink-based. They actively collect global evidences from sensor nodes to the sink so as to conduct centralized analysis at the powerful back-end. On the one hand, long distance proactive information retrieval incurs huge transmission overhead; On the other hand, due to the coupling effect between diagnosis component and the application itself, sink often fails to obtain complete and precise evidences from the network, especially for the problematic or critical parts. To avoid large overhead in evidence collection process, self-diagnosis injects fault inference modules into sensor nodes and let them make local decisions. Diagnosis results from single nodes, however, are generally inaccurate due to the narrow scope of system performances. Besides, existing self-diagnosis methods usually lead to inconsistent results from different inference processes. How to balance the workload among the sensor nodes in a diagnosis task is a critical issue. In this work, we present a new in-network diagnosis approach named Local-Diagnosis (LD2), which conducts the diagnosis process in a local area. LD2 achieves diagnosis decision through distributed evidence fusion operations. Each sensor node provides its own judgements and the evidences are fused within a local area based on the Dempster-Shafer theory, resulting in the consensus diagnosis report. We implement LD2 on TinyOS 2.1 and examine the performance on a 50 nodes indoor testbed.

Existing approaches to diagnosing sensor networks are generally sink-based, which rely on actively pulling state information from all sensor nodes so as to conduct centralized analysis. However, the sink-based diagnosis tools incur huge communication overhead to the traffic sensitive sensor networks. Also, due to the unreliable wireless communications, sink often obtains incomplete and sometimes suspicious information, leading to highly inaccurate judgments. Even worse, we observe that it is always more difficult to obtain state information from the problematic or critical regions. To address the above issues, we present the concept of self-diagnosis, which encourages each single sensor to join the fault decision process. We design a series of novel fault detectors through which multiple nodes can cooperate with each other in a diagnosis task. The fault detectors encode the diagnosis process to state transitions. Each sensor can participate in the fault diagnosis by transiting the detector's current state to a new one based on local evidences and then pass the fault detector to other nodes. Having sufficient evidences, the fault detector achieves the Accept state and outputs the final diagnosis report. We examine the performance of our self-diagnosis tool called TinyD2 on a 100 nodes testbed.