Design of AI-powered Hybrid Control Algorithm of Robot Vehicle for Enhanced Driving Performance

Abstract

This paper presents a hybrid control of robot vehicle using behavior cloning algorithm with PID control for enhanced driving performance. The behavior cloning algorithm based on convolutional neural network (CNN) has been implemented to the autonomous driving algorithm of a robot vehicle by implementing supervised learning approach. The algorithm mimics human driving behaviors for autonomous driving by obtaining and learning manual(human) control data. In order to implement the algorithm on the real system, a low-cost camera sensor and controllers with actuators are used so that it achieves robust driving performance. With a successfully developed and validated CNN based behavior cloning model, a classic PID control algorithm has been combined as a lateral controller to propose AI-powered hybrid control architecture which improves the performance of trajectory tracking in various environments using proposed behavior cloning algorithm. The performance of the proposed algorithm has been evaluated via robot vehicle tests. It has been shown from the simulation studies and vehicle tests that the proposed algorithm enhances driving performance.