Driver-centred Autonomous Vehicle Motion Control within A Blended Corridor

Abstract

Abstract As a potential cornerstone of the future intelligent transport system, autonomous vehicles (AVs) attract much attention of researchers across a wide range of areas from engineering to computer science. In addition, human factors issues, with respect to transfer of control and the interaction between the AVs and other road users have been studied. Current AV control algorithm development has focused on improving the safety of the vehicle, while the comfort of the drivers are normally ignored. Therefore, motion planning must not only avoid collisions between the vehicle and other road users and the road edges, but also needs to provide a sense of security and comfort for the drivers. Moreover, strict lane following can lead to overly cautious AVs relative to other road users, and thereby lead to traffic accidents. To solve these problems, we estimated the acceptable tolerance of the lateral offset based on the measured driving performance of real drivers and their reaction to a range of risk elements. Together with the vehicle dynamic constraints, the risk-based constraints are incorporated into a nonlinear Model Predictive Control (MPC) controller using a blended corridor. The result is a vehicle trajectory that produces a smooth motion within the corridor that considers the drivers’ comfort.