Destination choice modeling is a critical task in human mobility analysis, serving as the foundation for applications from urban planning to personalized travel services. Existing approaches often rely on individual trajectories or static surveys, limiting their adaptability and interpretability in complex urban settings. To address these issues, we propose KnowDest, a knowledge graph-based recommendation framework for destination choice modeling. Experiments on real-world data from Susono, Japan show that KnowDest improves Recall@10, Precision@10, and NDCG@10 by over 30% on average compared to baselines. Moreover, it offers structural and attribute-level explanations by explicitly identifying influential factors such as POI distributions or regional properties that influence travel decisions. Our approach presents a novel perspective on high-precision human mobility prediction, highlighting the potential of knowledge graph to support scalable, interpretable, and transferable modeling across varied urban contexts.

This paper summarizes the Optimized Road Damage Detection Challenge (ORDDC’2024), a Big Data Cup featured at the IEEE International Conference on Big Data 2024. Building on previous competitions, ORDDC’2024 aims to enhance the automatic detection and classification of road damage from images. It introduces two novel contributions: first, a standardized platform for model deployment that ensured consistent performance evaluation across all participants; second, an emphasis on inference speed as a critical evaluation criterion to meet the growing demand for real-time applications in infrastructure monitoring. The competition utilized the road damage dataset, RDD2022, released through the Crowdsensing-based Road Damage Detection Challenge CRDDC’2022, comprising road images from majorly 6 countries: India, Japan, Czech Republic, Norway, United States, and China. Attracting participants from 19 countries, the challenge yielded 76,069 submissions in Phase 1 and 353 in Phase 2. The winning model offers two solutions: one optimized for accuracy, achieving a peak F1-score of 86.18% at an inference speed of 136.41 milliseconds per image, and another optimized for speed, delivering 26.8 milliseconds per image with an F1-score of 79.27%. This paper analyzes leading solutions and challenges faced, providing insights for enhancing real-time road damage detection and improving global infrastructure maintenance strategies.

Knowledge Graphs (KGs) have recently emerged as a powerful tool for extracting directed multi-relational "knowledge" from structured facts within massive urban mobility data, supporting various urban application scenarios. However, current methods for constructing urban knowledge graphs (UrbanKGs), which rely on static data such as points of interest (POI) and road networks, fail to capture the dynamic relations that evolve over time. In this paper, we address this limitation by constructing a dynamic urban knowledge graph using heterogeneous spatio-temporal urban data to uncover deeper "knowledge". First, spatial features and their geographical relationships are extracted as entities and relations to build the UrbanKG. Next, timestamps are integrated from regional origin-destination (OD) data to form a dynamic UrbanKG. Finally, experimental results demonstrate that the dynamic models significantly outperform the static models in the knowledge graph completion (KGC) task. This confirms that incorporating temporal dynamics into UrbanKGs provides a more robust framework for capturing the complexities of urban transportation and improving urban management strategies.

Urban daily activity patterns play an important role in fields such as urban planning and traffic management, while the powerful data generation and reasoning capabilities of LLMs (Large Language Models) have sparked a surge of interest in recent years, with applications in various domains. Inspired by their natural language processing and pattern recognition functionalities, we attempted to utilize LLMs to simulate the activity patterns of people of different ages and occupations in metropolitan areas (Tokyo). By leveraging pre-processed Person Trip data, we employed 3 methods to test the ability of LLMs to generate urban daily activity data. The results were evaluated based on metrics such as rationality, diversity, and error rates. Results indicate that the fine-tuned LLaMA-3 model is capable of accurately simulating the activity distribution patterns of metropolitan populations. Among the prompt-based strategies, the few-shot approach yielded the best performance. Although designing the instruction for prompts and post-processing the data required considerable time, the few-shot prompt strategy proved to be an effective option for this task.

Large-scale commuting flow prediction is an essential task to estimate the commuting origin-destination (OD) demand within within a prefecture or the whole nation using multiple auxiliary data. Considering ranked structures of metropolitan areas and increased number of geographical units that need to be maintained, we develop a heterogeneous graph-based model to generate meaningful region embeddings at multiple spatial resolutions for predicting different types of inter-level OD flows. To demonstrate the effectiveness of the proposed method, extensive experiments were conducted using real-world aggregated mobile phone datasets collected from Shizuoka Prefecture, Japan. The results indicate that our proposed model outperforms existing models in terms of a uniform urban structure. We extend the understanding of predicted results using reasonable explanations to enhance the credibility of the model.

There is an urgent need for innovative technologies to detect road damage efficiently and cost-effectively. Traditional centralized deep learning models face challenges due to extensive data transfer and privacy concerns when sharing data among different parties. Federated Learning (FL) addresses these issues by sharing model parameters instead of raw data, enhancing collaboration without compromising privacy. While previous studies focused on similar datasets from various countries, this research showcases FL's efficacy with diverse data from Japan, China, and Norway. Our proposed solution involves developing a multi-country federated model capable of learning from diverse datasets and enhancing road damage detection accuracy across different regions, making it more robust and generalized than traditional models. Using YOLOv8l and the Flower framework with the FedAvg strategy, the FL model achieved an mAP50 of 0.467 on a multi-country test dataset, outperforming Japan's centralized model by 3.6% and Norway's and China's models by over 20%.

Human mobility generation plays a critical role in urban transportation planning. Existing human mobility generation models often fall short of understanding travelers' demographics and integrating multimodal information, including activity purposes, destination choices and transport mode preferences. Recently, mobility generation models leveraging Large Language Models (LLMs) have gained significant attention, while they are limited in directly reproducing spatial information in human mobility profiles. To address these challenges, this paper proposes the Mobility Generative Language Model (MobGLM), a novel approach for generating synthetic human mobility data to support urban planning, transport management, energy consumption and epidemic control. MobGLM addresses these limitations by capturing the complex relationships between agents' mobility patterns and individual demographics. By incorporating personal information, activity types, locations and traffic modes as encoders, MobGLM uniquely identifies and replicates features of human mobility. Our framework is evaluated using a large, real-world mobility dataset and benchmarked against state-of-the-art personal mobility generation techniques. The results demonstrate the effectiveness of MobGLM in producing accurate and reliable synthetic mobility data, highlighting its potential applications in various urban mobility contexts.

The highly accurate retrieval of large-scale road surfaces is crucial when developing nationwide digital twins. Many existing approaches have focused on detecting road surfaces from high-resolution aerial images, videos captured by dashcams, or digital surface models (DSMs), either individually or in combination with multiple data sources for stronger guarantees. The results of these implementations are highly dependent on the quality of the original data sources that are selected. Additionally, such approaches primarily employ deep learning, which is associated with challenges such as inefficient training and poor generalization of model results. Therefore, in this study, we propose a method for developing wide-ranging pavement surface models by realizing an accurate three-dimensional reconstruction module based on images with a lower resolution and introduce a multilevel Markov random field (MRF) model for mapping road networks from reconstructed sparse point cloud scenes. The experimental results demonstrated that the proposed development pipeline achieved an average of state-of-the-art performance on the International Society for Photogrammetry and Remote Sensing (ISPRS) dense matching dataset with a satisfactory root mean square error (RMSE) and obtained an overall accuracy (OA) of 90.3% and a mean intersection over union (mIoU) of 82.3% in the segmentation task. Using the Geospatial Information Authority of Japan (GSI) aerial image dataset, the robustness of the method in developing road surface models over a wide range of scenarios was further proven while accommodating less stricter source data requirements.

Remote sensing image semantic segmentation of land use, benefitted from the development of deep learning and consequently made considerable progress in terms of inferencing accuracy and speed. However, the effective training of semantic segmentation models for remote sensing imagery necessitates extensively detailed pixel-level annotations, and gathering such data is both time-intensive and laborious. Thus, this study implemented low-rank adaptation on a stable diffusion algorithm to learn the distribution of the pixel-level annotations in case of the LoveDA dataset. Consequently, the annotation-image pairs were used to train the remote sensing image generator based on stable diffusion guided by ControlNet. We proposed a stable diffusion based approach, which can generate image-annotation pairs from scratch. The generated annotation and image pairs achieved a high accuracy of 0.520 mean intersection-over-union on LoveDA dataset, which is close to the original data training result of 0.539 mIoU. Furthermore, the mixed training using generated and original data achieved 0.542 mIoU, thereby demonstrating the data augmentation function of our approach. This study provided a solution for the high-cost pixel-level annotation issue, and thus, exhibited the potential of artificial intelligence generated content.

This paper summarizes the Crowdsensing-based Road Damage Detection Challenge (CRDDC), a Big Data Cup organized as a part of the IEEE International Conference on Big Data’2022. The Big Data Cup challenges involve a released dataset and a well-defined problem with clear evaluation metrics. The challenges run on a data competition platform that maintains a real-time online evaluation system for the participants. In the presented case, the data constitute 47,420 road images collected from India, Japan, the Czech Republic, Norway, the United States, and China to propose methods for automatically detecting road damages in these countries. More than 70 teams from 19 countries registered for this competition. The submitted solutions were evaluated using five leaderboards based on performance for unseen test images from the aforementioned six countries. This paper encapsulates the top 11 solutions proposed by these teams. The best-performing model utilizes ensemble learning based on YOLO and Faster-RCNN series models to yield an F1 score of 76% for test data combined from all 6 countries. The paper concludes with a comparison of current and past challenges and provides direction for the future.

Origin–destination (OD) flow d ata a re c ritical for urban planning and traffic system design. Such data are suitable for describing movement at the macroscopic level. However, collecting them on a large scale, such as in a city, is challenging. Their form incompatibility makes using them for other tasks difficult. Therefore, we propose a deep model to learn meaningful OD information on grids within a city to address these problems. We collected multimodal characteristics of regions, such as road network densities and facility distributions, from several open-source datasets and used them as grid signals. We then constructed a spatial attention-based deep graph network to generate grid embeddings and used them to predict the OD volumes. The proposed method was evaluated against a set of baseline approaches using a real-world dataset in Japan. The analysis indicated that our model can extract more accurate latent topographical information from OD graphs and produce reasonable grid embeddings; these representations apply to other downstream tasks.

Analyzing vehicle movement trajectories is essential for understanding urban mobility and traffic flow patterns. Obtaining a reliable estimate of vehicle trajectory is challenging as it requires the vehicle to be observed and re-identified at different locations and times. Recently, a scalable citywide traffic flow estimation method has been proposed utilizing moving cameras on vehicle dashboards. This study extends the recently proposed method for traffic flow estimation by using cameras mounted on multiple moving observers to reconstruct the trajectory of detected vehicles. We develop the CARLA ReID dataset, which includes more than 50,000 images taken from 85 cameras for over 700 different vehicle models, and train a re-identification network to identify the same vehicle by multiple observers. Utilizing our proposed methodology, we conduct extensive research to estimate trajectories of vehicles in a driving simulator CARLA and evaluate the accuracy of reconstructed trajectories using Symmetrized Segment-Path Distance (SSPD) and Hausdorff Distance metrics. Our proposed method achieves a mean error of 5.13 meters evaluated using the SSPD metric for ten driving experiments in CARLA. Findings from this study will provide valuable insights for conducting traffic flow research in a simulation environment, which is otherwise challenging and costly in practice.

With the rapid trend of developing Smart Cities and Digital Twins, a better understanding of how humans move and perform a daily routine in the city area is vital. Benefiting from the recent rapidly growing location acquisition techniques, existing deep learning approaches can effectively model and predict human mobility with human mobility big data. However, it is still challenging to simulate human mobility at the population level because only the individuals with plentiful historical data can be well modeled. Moreover, the differences in complex city layouts and functions prevent us from applying the trained models to different cities. Therefore, in this study, we propose an alternative deep learning framework focusing on the destination choice to reconstruct nationwide human mobility at the population level. We design a new embedding mechanism for handling the traveler’s demographics, mobility characteristics, travel intention, and context information of locations to enrich the representations for the model prediction. Then a neural network is trained with the People Flow Dataset. We evaluated our approach based on multiple urban areas using different training data and demonstrated the advantages of our method compared with other baseline approaches.

Road rutting is a severe road distress that can cause premature failure of the road incurring early and costly maintenance costs. Research on road damage detection using image processing techniques and deep learning are being actively conducted in the past few years. However, these researches are mostly focused on the detection of cracks, potholes, and their variants. Very few research has been done on the detection of road rutting. This paper proposes a novel road rutting dataset comprising 949 images and provides both object-level and pixel-level annotations. Object detection models and semantic segmentation models were deployed to detect road rutting on the proposed dataset, and quantitative and qualitative analysis of model predictions were done to evaluate model performance and identify challenges faced in the detection of road rutting using the proposed method. Object detection model YOLOXs achieves mAP@IoU=0.5 of 61.6% and semantic segmentation model PSPNet (Resnet-50) achieves IoU of 54.69 and accuracy of 72.67, thus providing a benchmark accuracy for similar work in future. The proposed road rutting dataset and the results of our research study will help accelerate the research on the detection of road rutting using deep learning.

Vehicle class and orientation detection is fundamental to different tasks of autonomous driving such as traffic flow estimation, vehicle re-identification, etc. Recent researches using Convolutional Neural Networks (CNNs) or Vision-based Transformers for object detection tasks require a large number of images with annotations. The collection of a large amount of high-quality dataset is expensive and time-consuming. Synthetic images obtained from driving simulators allow the generation of a large synthetic dataset under varying conditions. This paper proposes YOLOv7-based neural networks for vehicle class and orientation detection that train on the synthetic dataset made available as part of the Vehicle class and Orientation Detection Challenge 2022 of IEEE BigData 2022. Our proposed approaches included the use of data resampling methods to address data imbalance in the dataset and ensemble learning to optimize the model performance. Trained models were thoroughly evaluated on real-world test datasets. Our proposed approach achieved weighted mean average precision of 0.497 allowing us to win the challenge.

Automatic monitoring of road conditions to detect different road damage types, such as cracks, potholes, etc., at low cost is important for early detection and timely maintenance of roads. This enables road maintenance bodies to perform frequent monitoring trips given the limited availability of allocated funds and manpower. In this direction, road damage detection using image processing techniques and deep learning on wide-view road images captured from smartphones or vehicle-mounted cameras has been actively researched in the past few years. This paper proposes deep learning-based neural networks and explores the feasibility of combining road damage data from different countries for road damage detection. The models were trained on the dataset made available as part of the Crowdsensing-based Road Damage Detection Challenge (CRDDC2022) of IEEE BigData 2022. The trained models were thoroughly evaluated on test datasets. Our proposed approach achieved an average F1 score of 0.628 allowing us to be in the Top 10 of the challenge. F1 scores achieved for Overall, India, Japan, Norway, and the United States were 0.697, 0.493, 0.715, 0.461, and 0.775 respectively.

A real-world traffic simulation can help better understand the need for infrastructure facilities in a region. Such simulations require digitized infrastructure information, well-represented people movement data, and efficient computing resources. Advances in computing resources and agent-based traffic simulators have made it possible to simulate real-world traffic conditions in a simulated environment. Digitized infrastructure datasets of roads, buildings, and other infrastructure facilities have enabled comprehensive visualization of infrastructure. However, the availability and accessibility of well-represented people’s mobility data are still part of some research. Additionally, due to resource limitations, traffic simulations are typically limited to geographic regions and use samples of traffic data rather than the entire population.This paper uses a novel mobility dataset, open Pseudo-PFLOW, which is a whole representation of the entire population of Japan, for a full-scale traffic modeling of Chiba prefecture. We used Multi-Agent Traffic Simulation (MATSim) tool, as an agent-based simulator. This has previously proven efficient for simulating large scenarios. The main research question focuses on improving agent trajectory data from the Pseudo-PFLOW dataset by agent-based modeling and validating the effectiveness of MATSim on network congestion and resource requirements of full-scale simulations.

Analysis of traffic fl ow pa rameters is ne cessary for Intelligent Transportation Systems (ITS) and autonomous driving research. Deep learning-based vehicle detection techniques have been widely used in reconstructing traffic fl ow parameters from video images. This research proposes a novel cross-sectional traffic fl ow es timation al gorithm to re construct tr affic volume from moving camera videos. We develop a vehicle detection dataset with more than one million annotations of vehicles with orientation and train a YOLOv4 based object detection network. We leverage the accurate vehicle detection model in tracking and estimating the distance of detected vehicles using Simple Online and Realtime Tracking (SORT) and photogrammetry techniques. The estimated distances and forward bearing of the observing vehicle are then utilized to calculate the GPS position of detected vehicles and used in the algorithm to estimate cross-sectional traffic fl ow. We ut ilize th e pr oposed al gorithm to es timate the traffic flow of 580 OpenStreetMap (OSM) road links and achieve an average accuracy of 84.30% verified a gainst 1 1 t raffic police sensor data in Susono city in Japan. The proposed large-scale dataset and cross-sectional traffic flow estimation algorithm open new avenues for ITS and autonomous driving research.

In recent years, due to various factors, including population decline, aging, and the promotion of compact cities, the social and lifestyle changes significantly impact people’s daily travel behavior. Although existing survey data and mobile phone data can reveal these phenomena, only aggregate-level results can be open to the public from the viewpoint of personal privacy. On the other hand, the conventional four-step travel demand estimation approach and human mobility models are limited to the low degree of freedom models, and hard to estimate the ever-changing traffic flow and the inability to capture the continuous behavior of people. This study develops a deep reinforcement learning-based agent model to tackle this problem and simulate the intercity people flow in the metropolitan area.

Human mobility datasets collected from various sources are indispensable for analyzing, predicting, and solving emerging urbanization and population issues. However, such datasets are only available to the public after aggregation and anonymous processing. In recent years, agent-based modeling approaches have addressed this problem by reproducing synthetic human mobility data through simulation. However, the development of such agent models typically requires a large amount of personal location histories as training data for parameter learning, leading to cost and privacy concerns. To overcome this disadvantage, we attempted to explore optimal parameters using a particle filter to alleviate the strict requirement of the data. We tested our method in a local city in Japan using aggregated real-time observation data collected from mobile phone service companies. The results show that the proposed model can achieve satisfactory accuracy using low-resolution data and can therefore be easily used by local governments for municipal applications.

This paper summarizes the Global Road Damage Detection Challenge (GRDDC), a Big Data Cup organized as a part of the IEEE International Conference on Big Data’2020. The Big Data Cup challenges involve a released dataset and a well-defined problem with clear evaluation metrics. The challenges run on a data competition platform that maintains a leaderboard for the participants. In the presented case, the data constitute 26336 road images collected from India, Japan, and the Czech Republic to propose methods for automatically detecting road damages in these countries. In total, 121 teams from several countries registered for this competition. The submitted solutions were evaluated using two datasets test1 and test2, comprising 2,631 and 2,664 images. This paper encapsulates the top 12 solutions proposed by these teams. The best performing model utilizes YOLO-based ensemble learning to yield an F1 score of 0.67 on test1 and 0.66 on test2. The paper concludes with a review of the facets that worked well for the presented challenge and those that could be improved in future challenges.

Agent-based simulations, combined with large scale mobility data, have been an effective method for understanding urban scale human dynamics. However, collecting such large scale human mobility datasets are especially difficult during rare events (e.g., natural disasters), reducing the performance of agent-based simulations. To tackle this problem, we develop an agent-based model that can simulate urban dynamics during rare events by learning from other cities using inverse reinforcement learning. More specifically, in our framework, agents imitate real human-beings' travel behavior from areas where rare events have occurred in the past (source area) and produce synthetic people movement in different cities where such rare events have never occurred (target area). Our framework contains three main stages: 1) recovering the reward function, where the people's travel patterns and preferences are learned from the source areas; 2) transferring the model of the source area to the target areas; 3) simulating the people movement based on learned model in the target area. We apply our approach in various cities for both normal and rare situations using real-world GPS data collected from more than 1 million people in Japan, and show higher simulation performance than previous models.

Unsupervised translation has become a popular task in natural language processing (NLP) due to difficulties in collecting large scale parallel datasets. In the urban computing field, place embeddings generated using human mobility patterns via recurrent neural networks are used to understand the functionality of urban areas. Translating place embeddings across cities allow us to transfer knowledge across cities, which may be used for various downstream tasks such as planning new store locations. Despite such advances, current methods fail to translate place embeddings across domains with different scales (e.g. Tokyo to Niigata), due to the straightforward adoption of neural machine translation (NMT) methods from NLP, where vocabulary sizes are similar across languages. We refer to this issue as the domain imbalance problem in unsupervised translation tasks. We address this problem by proposing an unsupervised translation method that translates embeddings by exploiting common hierarchical structures that exist across imbalanced domains. The effectiveness of our method is tested using place embeddings generated from mobile phone data in 6 Japanese cities of heterogeneous sizes. Validation using landuse data clarify that using hierarchical anchors improves the translation accuracy across imbalanced domains. Our method is agnostic to input data type, thus could be applied to unsupervised translation tasks in various fields in addition to linguistics and urban computing.

Western Japan experienced heavy, record-breaking rain from June 28 to July 8, 2018, causing approximately 600 road sections to be closed in Hiroshima and Okayama Prefectures. The government performed recovery activities pursuant to their own road-recovery plan to help damaged citizens' lives return to their original condition as soon as possible. However, it took a week after the disaster to begin to restoration of neighborhood roads. Therefore, optimal decision-making systems are necessary to improve resiliency related to real life. We adopt an off-policy algorithm of a single-agent deep Q-learning approach, because disaster situations differ based on the occurrence area, and a generalized solution for optimal decision making does not exist. To develop a decision-making system that accounts for human mobility, we utilized origin-destination pairs extracted from the location data of smartphones and, digital road maps reproduced from real topologies. Agent in our model is the entrance point of the O-D pairs passing through a damaged road section on normal days. Each agent is assigned tendency to decide actions according to the road type, and a weight factor depending on the extent of damage. In addition, a restoration rate, which refers to the score of the action, is calculated based on the delayed travel time according to the changed demand of the trip. Restoration rate is affected by the agent's action, and gives the environment information related to the answer signals. This helped us to ascertain that it takes a longer time for origin-destination pairs using a road section with heavy traffic to recover to their original state. We estimated the number of required steps of each O-D's mobility until they get their own restoration rate over than 0.85. We ensure that government administrations could utilize our results as a reference data to determine the priority of restoration work and estimate the amount of input per route.

When conducting agent economic simulation for supply chains, inter-company transaction data are essential. However, the current inter-firm transaction data are network data in which branch office information is aggregated into headquarters transaction data. This study proposes a method to estimate branch office transactions from inter-company transaction data aggregated among headquarters by using a gravity model. We also confirm the method's reliability by comparing the estimated transaction data with the inter-regional input-output tables. We analytically considered the transition for all network configurations, demonstrating that the transaction quantity depends on the amount of labor and distance. We also demonstrated that our model fits well with data from business transactions, implying that the whole network structure can be used to model money flow in the real world.

Large mobility datasets collected from various sources have allowed us to observe, analyze, predict and solve a wide range of important urban challenges. In particular, studies have generated place representations (or embeddings) from mobility patterns in a similar manner to word embeddings to better understand the functionality of different places within a city. However, studies have been limited to generating such representations of cities in an individual manner and has lacked an inter-city perspective, which has made it difficult to transfer the insights gained from the place representations across different cities. In this study, we attempt to bridge this research gap by treating cities and languages analogously. We apply methods developed for unsupervised machine language translation tasks to translate place representations across different cities. Real world mobility data collected from mobile phone users in 2 cities in Japan are used to test our place representation translation methods. Translated place representations are validated using landuse data, and results show that our methods were able to accurately translate place representations from one city to another.

Predicting the evacuation decisions of individuals before the disaster strikes is crucial for planning first response strategies. In addition to the studies on post-disaster analysis of evacuation behavior, there are various works that attempt to predict the evacuation decisions beforehand. Most of these predictive methods, however, require real time location data for calibration, which are becoming much harder to obtain due to the rising privacy concerns. Meanwhile, web search queries of anonymous users have been collected by web companies. Although such data raise less privacy concerns, they have been under-utilized for various applications. In this study, we investigate whether web search data observed prior to the disaster can be used to predict the evacuation decisions. More specifically, we utilize a session-based query encoder that learns the representations of each user's web search behavior prior to evacuation. Our proposed approach is empirically tested using web search data collected from users affected by a major flood in Japan. Results are validated using location data collected from mobile phones of the same set of users as ground truth. We show that evacuation decisions can be accurately predicted (84%) using only the users' pre-disaster web search data as input. This study proposes an alternative method for evacuation prediction that does not require highly sensitive location data, which can assist local governments to prepare effective first response strategies.

Railway disturbances occur every day and train rescheduling is conducted by human experts. Approaches for automating rescheduling have been widely studied (e.g., heuristic approach and mixed integer problem-based approach). However, extant research is still inadequate for employing the approach in practical application in terms of size, run-time, and solution accuracy. In this research, we simulate train dispatching using graph theory and propose a reinforcement learning method (i.e., Deep Q-Network (DQN)) for rescheduling. We also show experimental results of this algorithm. DQN presented positive results for over 50% of test cases, and its train rescheduling decreased approximately 20% of passengers' dissatisfaction in a certain case. It can be expected that applying a DQN approach to real-world scale cases by will improve methods to handle larger-scale networks.

Recently, the frequency and intensity of weather-related disasters are increasing and are becoming more ubiquitous, often devastating vulnerable rural areas. To prepare for speedy and effective first response, we need a flood detection method that works much faster and is able to cover a wider area compared to conventional methods that use CCTV cameras and low cost sensors, which are costly to distribute ubiquitously in all areas with possible flood threats. With the spread of mobile phones, we are able to obtain real time anonymized location information of individuals in a ubiquitous, low cost, and a continuous manner from users that have agreed to provide their location data for disaster relief purposes. Here we propose a novel method that infers flooded areas in real time by detecting anomalous behaviors of individuals using mobile phone location data. We are motivated in applying our method to rural areas that are costly to cover using cameras and sensors. To overcome the sparseness of mobile phone location signals in such rural areas, our method combines mobile phone location data with terrain information including the digital elevation model and river trajectory data. We evaluated our method using real world data from 2 severe floods in the rural parts of Japan and verified that our method is more accurate and has numerous advantages compared to conventional methods. This work presents the potential use of mobile phone data as a complementary, if not an alternative method for flood detection especially in rural areas.

This paper is the first work to replicate and simulate urban dynamics by learning individuals' decision-making processes and creating human-like agents from GPS data. We develop a novel agent model by learning from historical data via reinforcement learning techniques. We test our methodology in different scenarios at the citywide level using real world smartphone GPS data. Simulation results show that our agents can successfully learn and generate human-like travel activities. Furthermore, the performance of synthetic urban dynamics significantly outperforms existing methods.

In recent years, there has been remarkable development in unmanned aerial vehicle UAVs); certain companies are trying to use the UAV to deliver goods also. Therefore, it is predicted that many such objects will fly over the city, in the near future.

In recent decades, the frequency and intensity of natural disasters has increased significantly, and this trend is expected to continue. Therefore, understanding and predicting human behavior and mobility during a disaster will play a vital role in planning effective humanitarian relief, disaster management, and long-term societal reconstruction. However, such research is very difficult to perform owing to the uniqueness of various disasters and the unavailability of reliable and large-scale human mobility data. In this study, we collect big and heterogeneous data (e.g., GPS records of 1.6 million users 1 over 3 years, data on earthquakes that have occurred in Japan over 4 years, news report data, and transportation network data) to study human mobility following natural disasters. An empirical analysis is conducted to explore the basic laws governing human mobility following disasters, and an effective human mobility model is developed to predict and simulate population movements. The experimental results demonstrate the efficiency of our model, and they suggest that human mobility following disasters can be significantly more predictable and be more easily simulated than previously thought.

Real-time estimation of human mobility following a massive disaster will play a crucial role in disaster relief. Because human mobility in massive disasters is quite different from their usual mobility, real-time human location data is necessary for precise estimation. Due to privacy concerns, real-time data is anonymized and a popular form of anonymization is population distribution. In this paper, we aim to estimate human mobility following an unprecedented disaster using such population distribution data. To overcome technical obstacles including high dimensionality, we propose novel particle filter by devising proposal distribution. Our proposal distribution provides states considering both prediction model and acquired observation. Therefore, particles maintain high likelihood. In the experiments, our methods realized more accurate estimation than the baselines, and its estimated mobility was consistent with the survey researches. The computational cost is significantly low enough for real-time operations. The GPS data collected on the day of the Great East Japan Earthquake is used for the evaluation.

Large scale disasters cause severe social disorder and trigger mass evacuation activities. Managing the evacuation shelters efficiently is crucial for disaster management. Kumamoto prefecture, Japan, was hit by an enormous (Magnitude 7.3) earthquake on 16th of April, 2016. As a result, more than 10,000 buildings were severely damaged and over 100,000 people had to evacuate from their homes. After the earthquake, it took the decision makers several days to grasp the locations where people were evacuating, which delayed of distribution of supply and rescue. This situation was made even more complex since some people evacuated to places that were not designated as evacuation shelters. Conventional methods for grasping evacuation hotspots require on-foot field surveys that take time and are difficult to execute right after the hazard in the confusion.

Mid-level disasters that frequently occur, such as typhoons and earthquakes, heavily affect human activities in urban areas by causing severe congestion and economic loss. Predicting the irregular movement of individuals following such disasters is crucial for managing urban systems. Past survey results show that mid-level disasters do not force many individuals to evacuate away from their homes, but do cause irregular movement by significantly delaying the movement timings, resulting in severe congestion in urban transportation. We propose a novel method that predicts such irregularity of individuals' movements in several mid-level disasters using various types of features including the victims' usual movement patterns, disaster information, and geospatial information of victims' locations. Using real GPS data of 1 million people in Tokyo, we show that our method can predict mobility delay with high accuracy,

The frequency and intensity of natural disasters has significantly increased over the past decades and this trend is predicted to continue. Facing these possible and unexpected disasters, accurately predicting human emergency behavior and their mobility will become the critical issue for planning effective humanitarian relief, disaster management, and long-term societal reconstruction. In this paper, we build up a large human mobility database (GPS records of 1.6 million users over one year) and several different datasets to capture and analyze human emergency behavior and their mobility following the Great East Japan Earthquake and Fukushima nuclear accident. Based on our empirical analysis through these data, we find that human behavior and their mobility following large-scale disaster sometimes correlate with their mobility patterns during normal times, and are also highly impacted by their social relationship, intensity of disaster, damage level, government appointed shelters, news reporting, large population flow and etc. On the basis of these findings, we develop a model of human behavior that takes into account these factors for accurately predicting human emergency behavior and their mobility following large-scale disaster. The experimental results and validations demonstrate the efficiency of our behavior model, and suggest that human behavior and their movements during disasters may be significantly more predictable than previously thought.

The Great East Japan Earthquake and the Fukushima nuclear accident cause large human population movements and evacuations. Understanding and predicting these movements is critical for planning effective humanitarian relief, disaster management, and long-term societal reconstruction. In this paper, we construct a large human mobility database that stores and manages GPS records from mobile devices used by approximately 1.6 million people throughout Japan from 1 August 2010 to 31 July 2011. By mining this enormous set of Auto-GPS mobile sensor data, the short-term and long-term evacuation behaviors for individuals throughout Japan during this disaster are able to be automatically discovered. To better understand and simulate human mobility during the disasters, we develop a probabilistic model that is able to be effectively trained by the discovered evacuations via machine learning technique. Based on our training model, population mobility in various cities impacted by the disasters throughout the country is able to be automatically simulated or predicted. On the basis of the whole database, developed model, and experimental results, it is easy for us to find some new features or population mobility patterns after the recent severe earthquake, tsunami and release of radioactivity in Japan, which are likely to play a vital role in future disaster relief and management worldwide.