An Event-Triggered MPC Based on Neural Network for Collision Avoidance of Autonomous Vehicles

Collision avoidance control is essential for the driving safety of autonomous vehicles. Due to vehicle modeling uncertainties and computational burden, vehicle motion control considering road obstacles has become a challenge for autonomous driving. This paper presents a collision avoidance strategy for autonomous vehicles using the event-triggered model predictive control based on a neural network. The event-triggered MPC approach is adopted to ensure that the reference path is replanned only when the tracking error reaches the triggering level, so that path planning and motion control are conducted with different time intervals to improve the adaptability of trajectory tracking. The neural network is used to approximate the vehicle modeling uncertainties. The interference compensation is added to the vehicle nominal model to improve modeling accuracy, reducing the triggering frequency and the computational burden. The simulation results show that the proposed event-triggered MPC based on neural network can reduce the number of triggering times and, the tracking errors compared with the traditional approach.