Time-to-approach-based model predictive controller for autonomous vehicle safety during unexpected wheelchairs on the path

Abstract

ABSTRACT The objectives of this study are to (i) develop a Time-to-Approach (TTA)-based Model Predictive Controller (MPC) for an Autonomous Vehicle (AV) to safely navigate around an unexpected wheelchair on its path and (ii) develop a prediction model to forecast a wheelchair's paths and potential collisions between it and an AV. Our analyses reveal that the TTA-based MPC enhances safety by reducing collision potential between an AV and a wheelchair by 44% when compared to the traditional distance-based MPC, as measured by the lateral deviation of an AV from the potential collision location. Furthermore, our approach improves passenger comfort by an average of 62% compared to the distance-based MPC.