Emergency Lane-Change Maneuvers of Autonomous Vehicles

This paper addresses the issue of collision avoidance using lane-change maneuvers. Of particular interest is to determine the minimum distance beyond which an obstacle cannot be avoided at a given initial speed. Using a planar bicycle model, we first compute the sharpest dynamically feasible maneuver by minimizing the longitudinal distance of a lane transition, assuming given initial and free final speeds. The minimum distance to an obstacle is then determined from the path traced by the optimal maneuver. Plotting the minimum distance in the phase plane establishes the clearance curve, a valuable tool for planning emergency maneuvers. For the bicycle model, the clearance curve is shown to closely correlate with the straight line produced by a point mass model. Examples demonstrate the use of the clearance curve for planning safe avoidance maneuvers.