Mobile crowdsensing game in vehicular networks

Vehicular crowdsensing takes advantage of the mobility of vehicles to provide location-based services in large-scale areas. In this paper, we analyze vehicular crowdsensing and formulate the interactions between a crowdsensing server and a number of vehicles equipped with sensors in the area of interest as a vehicular crowdsensing game. Each participant vehicle chooses its sensing strategy based on the sensing and transmission costs, and the expected payment by the server, while the server determines its payment policy according to the number and accuracy of the sensing reports. A reinforcement learning based crowdsensing strategy is proposed for vehicular networks, with incomplete system parameters such as the sensing costs of the other vehicles. The server and vehicles achieve their optimal payment and sensing strategies by learning via trials, respectively. Simulation results have verified the efficiency of the proposed mobile crowdsensing systems, showing that the average utilities of the vehicles and the server can be improved and converged to the optimal values in fast speed. Vehicles with less sensing costs are motivated to upload more accurate sensing data.