Vehicle backward driving control with obstacle avoidance

This paper introduces an autonomous backward driving control method using space information from stereo vision system. For forward driving, numerous autonomous driving control methods are already developed and they work better than inexperienced drivers. However, autonomous backward driving control works in very limited cases, such as for autonomous parking within short distances. In case of backward driving in a long, narrow, and bended corridor, even a human driver may have a difficulty in passing through because of the non-minimum phase nature of a backwardly driven car and a limited lateral space margin. For autonomous backward driving in such case, we design a controller using a Model Predictive Control (MPC) technique that can explicitly handle the constraints and upcoming driving conditions. The near future information can be obtained using a stereo camera. We evaluated the performance of the suggested method using simulations.