Gains of Mobility for Communication and Sensing in Vehicular Sensor Networks

In this thesis, mobility information exchanged among vehicles devices is utilized to improve the communication and sensing in vehicular networks. Mobility usually causes a loss in communications, and can add an additional load in sensing. There have been research attempts to handle such challenges in vehicular networks by addressing them after realizing the mobility impact, or adaptively addressing the problem as the mobility changes. This thesis takes a different approach to enhance communication, and sensing in vehicular networks. The first objective of this thesis is to utilize mobility information in order to enhance communication in vehicular networks by reducing the excessive load on the channel, while preserving the communicated information. The second objective of this thesis is to utilize predicted mobility information in order to enhance sensing in vehicular sensor networks by efficiently providing the sensing metric, with a minimal load on the communication channel. In order to have mobility information, vehicles has to communicate that information. The first part of this thesis examines location awareness in vehicular networks via sparse recovery: that is, how vehicles would know the locations of each other in the vicinity in order to provide the optimized mobile sensing of the first part of the thesis. Locations of vehicles are exchanged periodically via beaconing to make each vehicle aware of the location of nearby vehicles for improved safety, and to provide non-safety services. The amount of data exchanged via periodical beacon broadcast can be extremely large, and the channel can become congested in dense scenarios. We proposed a novel congestion control scheme that minimizes the amount of broadcast data while preserving the location information for each vehicle using compressive sensing. This novel scheme is designed for two different modes that are suitable for two different applications. The first approach is a super-frame scheme that is designed for delay-tolerant applications, such as updating traffic maps (e.g., Google maps). The second approach is a sliding window scheme that is designed for real-time applications, such as safety packet exchange in vehicular networks. The proposed congestion control scheme was implemented on a smartphone-based testbed and shown to minimize the amount of data exchange while successfully preserving beaconing information with high accuracy in both delay-tolerant and real-time modes. Experimental tests were conducted in the highways and downtown streets of the city of Toronto. The proposed scheme is shown to reduce the number of exchanged packets while preserving the communicated information with excellent accuracy. The second part of this thesis examines the gain of predicted mobility in enhancing the coverage of targets. Herein, the sensors can be cameras, and sensing becomes the coverage of targets. Moreover, targets becomes the specific areas of the road that are of interest for coverage. Due to the limited communication channel capacity in the vehicular network, the main objective is to minimize the amount of sensed and transmitted data while preserving the coverage of all target areas. Specifically, we utilize predicted car mobility in order to provide the required coverage of target areas with less sensor activations. Activations in this context means that a sensor is selected for covering a target area, and the captured image is transmitted over the communication channel to a fusion centre. First, we formulate mathematical optimization models for the proposed mobile sensing scheme and the existing stationary sensing scheme. Then, by using probability analysis, we demonstrate that the proposed scheme outperforms the existing stationary solution in terms of sensing cost and size of the feasibility region of the optimization problem. After that, we propose two approximation algorithms that allow practical implementation of the novel coverage scheme in the centralized and distributed modes. In this part, we assume that the mobility information is known. The mobile sensing scheme is also studied when the predicted mobility information is noisy. We show that the mobile sensing scheme outperforms the stationary sensing scheme when the noise level in mobility information is small. Increasing the noise level in mobility information results in an increased sensing cost for the mobile sensing scheme. Then a breaking point exists in which the noise level in mobility information results in larger sensing cost for the mobile sensing scheme compared to that of the stationary sensing scheme. The mobile sensing scheme breaking point is found via analysis and simulation.Ph.D