Collaborative Sensing Method for Intelligent Driving of New Energy Vehicles Based on Mobile Edge Computing

This paper addresses the current situation of huge amounts of data transmission and computing in the edge computing strategy of Internet of Vehicles. Firstly, an edge computing network model is constructed, which consists of two parts: vehicles and roadside units. Then, a communication model among multiple vehicles and a perception model among multiple vehicles are built. Based on these models, the motion state of the vehicle end, the operation state of the local computer, the transmission state of the wireless communication network, and the operation state of the edge server are combined into a Markov process, and the quality of the decision-making process is evaluated by reward values. This facilitates the solution of the optimal task offloading strategy through reinforcement learning theory in the subsequent text. In the process of solving the optimal parameters, a federated learning algorithm with joint auxiliary training and adaptive sparsity is used to ensure that important devices can continuously participate in multiple rounds of federated training and model solving. Finally, experiments are conducted to compare the algorithm in this paper with the previous method. The comparison shows that the offloading strategy in this paper can achieve smaller delays. By setting different vehicle scales, the improvement of the perception coverage area through the collaborative control among different vehicles is verified.