Lateral Motion Planning for Evasive Lane Keeping of Autonomous Driving Vehicles based on Target Prioritization

This paper describes a motion planning algorithm for evasive lane keeping in urban autonomous driving environments. Based on high-definition map (HD map), it is determined whether the surrounding vehicles and obstacles recognized by the environmental sensors exist within the driving lane of the subject vehicle. There are two methods of avoiding in lane static and dynamic obstacles: lane change and in lane avoidance behavior. Using HD map, it is possible to determine the lateral position of an obstacle, and to determine whether the subject vehicle can avoid it without changing lanes. The major in lane obstacle which determines the lateral motion decision of the subject vehicle is defined as the ‘Primary Target’. Based on the position and lateral space occupancy of the primary target, the demand for ‘Lane Change’ and ‘Biased Driving’ is determined. According to each mode demand, the motion planner determines the desired motion for lane change or lane biased driving. A model predictive controller was implemented to determine the desired steering angle and longitudinal acceleration of the ego vehicle for execution of the desired motion. Actual vehicle tests were performed to verify the performance of the proposed motion planning algorithm. The results confirmed that the proposed motion planner can effectively manage collision risk without loss of traffic flow.