An Information Theoretic Vehicle Following System

Abstract

Vehicle following can be achieved by minimizing the relative information (Kullback-Leibler or K-L distance), between the estimated poses of leader and follower vehicles. To achieve successful vehicle following, a Bayesian formulation for the system has been derived, and two probabilistic distributions, one for each vehicle's pose, can be obtained. Based on the assumption that the two pose distributions are Gaussian functions, the K-L distance of the vehicle following system can be computed with these two computed distributions. With a series of achievable actions, such as steering and velocity commands, for the follower vehicle at each pose prediction step, and by minimizing the K-L distance, an optimized action for the follower vehicle can be obtained. The information theoretic vehicle following algorithm has been tested under a simulated environment by analyzing the performance of the follower vehicle when the leader vehicle undergoes various kinds of maneuvers. The simulated experimental results validate that the follower is able to trail the trajectories of the leader vehicle satisfactorily and at the same time maintain a safe following distance.