Abstract: A system and method for reactive collision avoidance for autonomous vehicles for avoiding collisions with late-detected deterministic threats while minimizing deviations from planned trajectories and maximizing safety and comfort for vehicle occupants and others in the environment. Some implementations include receiving control commands for a planned path, determining distances to late-detected objects, determining at least one rate of change of a distance that exceeds a threshold, predicting that a collision will occur based on a collision function, and modifying at least one of the control commands to avoid the collision.

Abstract: An autonomous vehicle uses precise control that is particularly useful with difficult scenes such as ultra narrow scenes. A plan including a planned path and a speed plan for the autonomous vehicle to traverse through a portion of a vehicle transportation network is received from a trajectory planner. An optimization operation is performed to revise the plan, where the optimization operation minimizes a lateral and heading error as compared to the planned path while applying each of a lateral offset constraint and a heading angle constraint as soft constraints. The soft constraints are represented by respective slack variables, and a penalty is applied to at least one of the slack variables that is greater than a penalty applied to the lateral and heading error to prioritize collision avoidance over maintaining the planned path. At least one control system of the autonomous vehicle is operated according to the plan as revised.

Abstract: An evidence grid is established in a map frame based on the vehicle's pose. Hazard data is accumulated in the evidence grid by gathering visible grid cells, each associated with a visibility probability indicating a history of visibility determinations, and accumulating hazard presence, with each grid cell linked to a hazard associated with a hazard presence probability indicating a history of hazard presence determinations. The accumulated hazard data is then provided to a process within the autonomous vehicle for further decision-making or risk mitigation.