Vehicular crowdsensing takes advantage of the mobility of vehicles to provide location-based services in large-scale areas. In this paper, mobile crowdsensing (MCS) in vehicular networks is analyzed and the interactions between a crowdsensing server and vehicles equipped with sensors in an area of interest is formulated as a vehicular crowdsensing game. Each participating vehicle chooses its sensing strategy based on the sensing cost, radio channel state, and the expected payment. The MCS server evaluates the accuracy of each sensing report and pays the vehicle accordingly. The Nash equilibrium of the static vehicular crowdsensing game is derived for both accumulative sensing tasks and best-quality sensing tasks, showing the tradeoff between the sensing accuracy and the overall payment by the MCS server. A Q-learning-based MCS payment strategy and sensing strategy is proposed for the dynamic vehicular crowdsensing game, and a postdecision state learning technique is applied to exploit the known radio channel model to accelerate the learning speed of each vehicle. Simulations based on a Markov-chain channel model are performed to verify the efficiency of the proposed MCS system, showing that it outperforms the benchmark MCS system in terms of the average utility, the sensing accuracy, and the energy consumption of the vehicles.
Discussion(0)
No comments yet. Be the first to comment.