Lane-Deviation Penalty for Autonomous Avoidance Maneuvers
A formulation of an offline motion-planning method for avoidance maneuvers based on a lane-deviation penalty function is proposed,which aims to decrease the risk of a collision by minimizing the time when a vehicle is outside of its own driving lane in the case ofavoidance maneuvers. The penalty function is based on a logistic function. The method is illustrated by computing optimal maneuversfor a double lane-change scenario. The results are compared with minimum-time maneuvers and squared-error norm maneuvers. Thecomparison shows that the use of the considered penalty function requires fewer constraints and that the vehicle stays less time in theopposing lane. The similarity between the obtained trajectories for different problem configurations was noticed. This property couldbe used in the future for predicting an intermediate trajectory online from a sparse data set of maneuvers.
Pavel Anistratov, Björn Olofsson and Lars Nielsen
Page responsible: webmaster
Last updated: 2019-12-17