Collision Warning Algorithm Based on Efficient Federated Learning in Mobile Edge Computing Assisted Intelligent Driving

Collision avoidance tasks in intelligent driving have challenges such as extremely high latency requirements and privacy protection. First, a Gated Recurrent Unit_Support Vector Machine (GRU_SVM) collision multi-level warning algorithm based on Semi-asynchronous Federated Learning with Adaptive Adjustment of Parameters (SFLAAP) is proposed. SFLAAP can dynamically adjust two training parameters according to training and resource conditions: the number of local training times and the number of local models participating in aggregation. Then, in order to solve the efficiency problem of collaborative training of collision warning model under resource-constrained Mobile Edge Computing (MEC), according to the relationship between the above parameters and SFLAAP training delay, a model for minimizing the total training delay is established, and it is transformed into a Markov Decision Process (MDP). Finally, in the established MDP, the Asynchronous Advantage Actor-Critic (A3C) algorithm is employed to determine adaptively the optimal training parameters, thereby reducing the training completion time of the collision warning model. The simulation results show that the proposed algorithm can effectively reduce the total training delay and ensure prediction accuracy.