Recurrent Neural Network-Based Model Predictive Control for Waypoint Tracking

Abstract

This paper presents an recurrent neural network-based model predictive control for an autonomous driving vehicle. Model predictive control is effective in vehicle lateral control but too computationally expensive to be applied in real-time control. To resolve this problem, we propose a recurrent neural network-based approximate model predictive control. The offline-trained neural network exhibits the ability to model the waypoint tracking system and provided the closed-loop performance. The performance of the approximate recurrent neural network-model predictive control (RNN-MPC) is validated by computational experiments of waypoints tracking control scheme.