Abstract

nnually, millions of people die and many more sustain non-fatal injuries because of road traffic crashes. Despite multitude of countermeasures, the number of causalities and disabilities owing to traffic accidents are increasing each year causing grinding social, economic, and health problems. Due to their high volume and lack of protective-shells, more than half of road traffic deaths are imputed to vulnerable road users (VRUs): pedestrians, cyclists and motorcyclists. Mobile devices combined with fog computing can provide feasible solutions to protect VRUs by predicting collusions and warning users of an imminent traffic accident. Mobile devices’ ubiquity and high computational capabilities make the devices an important components of traffic safety solutions. Fog computing has features that suits to traffic safety applications as it is an extension of cloud computing that brings down computing, storage, and network services to the proximity of end user. Therefore, in this thesis, we have proposed an infrastructure-less traffic safety architecture that depends on fog computing and mobile devices possessed by VRUs and drivers. The main duties of mobile devices are extracting their positions and other related data and sending cooperative awareness message to a nearby fog server using wireless connection. The fog server estimates collision using a collision prediction algorithm and sends an alert message, if an about-to-occur collision is predicted. Evaluation results shows that the proposed architecture is able to render alerts in real time. Moreover, analytical and performance evaluations depict that the architecture outperforms other related road safety architectures in terms of reliability, scalability and latency. However, before deploying the architecture, challenges pertaining to weaknesses of important ingredients of the architecture should be treated prudently. Position read by mobile devices are not accurate and do not meet maximum position sampling rates traffic safety applications demand. Moreover, continuous and high rate position sampling drains mobile devices battery quickly. From fog computing’s point of view, it confronts new privacy and security challenges in addition to those assumed from cloud computing. For aforementioned challenges, we have proposed new solutions: (i) In order to improve GPS accuracy, we have proposed an efficient and effective two-stage map matching algorithm. In the first stage, GPS readings obtained from smartphones are passed through Kalman filter to smooth outlier readings. In the second stage, the smoothed positions are mapped to road segments using online time warping algorithm. (ii) position sampling frequency requirement is fulfilled by an energy efficient location prediction system that fuses GPS and inertial sensors’ data. (iii) For energy efficiency, we proposed an energy efficient fuzzy logic-based adaptive beaconing rate management that ensures safety of VRUs. (iv) finally, privacy and security issues are addressed indirectly using trust management system. The two-way subjective logic-based trust management system enables fog clients to evaluate the trust level of fog servers before awarding the service and allows the servers to check out the trustworthiness of the service demanders. Engaging omnipresent mobile device and QoS-aware fog computing paradigm in active traffic safety applications has the potential to reduce overwhelming number of traffic accidents on VRUs.; Chaque année, des millions de personnes meurent et beaucoup d'autres subissent des séquelles graves à la suite d'accidents de la route. Malgré une multitude d’initiatives, le nombre de cas mortels et d'accidents graves augmente chaque année en engendrant des problèmes préoccupants à la fois sociaux, économiques et sanitaires. En raison de leur nombre élevé et de l'absence de protection personnelle, plus de la moitié de ces décès concerne les usagers vulnérables (en anglais, vulnerable road users - VRU) regroupant les piétons, cyclistes et motocyclistes. Les appareils mobiles, combinés à la technologie de Fog Computing (ou informatique géodistribuée, ou même informatique en brouillard), représentent une solution réaliste à court terme pour les protéger en les avertissant de l’imminence d'un accident de circulation. L’omniprésence des appareils mobiles et leurs capacités de calcul élevées font de ces appareils un élément important à considérer dans les solutions de sécurité routière. Le Fog Computing offre des fonctionnalités adaptées aux applications de sécurité routière, puisqu’il s’agit d’une extension du Cloud Computing permettant de rapprocher les services informatiques, le stockage et le réseau au plus près des utilisateurs finaux. Par conséquent, dans cette thèse, nous proposons une architecture réseau sans infrastructure supplémentaire (PV-Alert) pour des fins de sécurité routière et reposant uniquement sur les appareils mobiles des VRU et des conducteurs sur la route avec l’aide du concept de Fog Computing. Les données géographiques et cinématiques de ces appareils sont collectées et envoyées périodiquement au serveur fog situé à proximité. Le serveur fog traite ces données en exécutant un algorithme de calcul de risque d’accident de circulation et renvoie des notifications en cas d'accident imminent. L’évaluation de cette architecture montre qu’elle est capable de générer des alertes en temps réel et qu’elle est plus performante que d’autres architectures en termes de fiabilité, d’évolutivité et de latence.

Abstract

Nowadays, vehicle safety is constantly increasing thanks to the improvement of vehicle passive and active safety. However, on a daily usage of the car, traffic jams remains a problem. With limited space for road infrastructure, automation of the driving task on specific situation seems to be a possible solution. The French project ABV, which stands for low speed automation, tries to demonstrate the feasibility of the concept and to prove the benefits. In this article, we describe the scientific background of the project and expected outputs. Comment: TRA (2012) Document type: Conference object

Abstract

In recent years, the growing prevalence of social networks makes it possible to utilize human users as sensors to inspect city environment and human activities. Consequently, valuable insights can be gained by applying data mining techniques to the data generated through social networks. In this work, a practical approach to combine data mining techniques with statistical analysis is proposed to implement crowd sensing in a smart city. A case study to analyze the relationship between weather conditions and traffic congestion in Beijing based on tweets posted on Sina Weibo platform is presented to demonstrate the proposed approach. Following the steps of data pre-processing and topic determination, we applied Granger Causality Test to study the causal relationships between weather conditions, traffic congestion and human outdoor activity. The mediation analysis is also implemented to verify human outdoor activity as a mediator variable significantly carrying the influence of good weather to traffic congestion. The result demonstrates that outdoor activity serves as a mediator transmitting the effect of good weather on traffic congestion. In addition, the causes of negative emotion are also studied.