Task-motion planning for safe and efficient urban driving

Abstract

Autonomous vehicles plan at a task level to compute a sequence of symbolic actions to fulfill service requests, where efficiency is the main concern. The vehicle computes continuous trajectories to perform actions at the motion level, where safety is important. Task-motion planning in autonomous driving faces the problem of maximizing task-level efficiency while ensuring motion-level safety. Task-Motion Planning for Urban Driving (TMPUD) enables the task and motion planners to communicate about the safety level of driving behaviors. The motion planner incrementally advances the vehicle toward a goal with an associated incremental utility, based on at least a safety of motion trajectories. The task planner defines the goal and a sequence of the actions to advance the vehicle toward the goal, dependent an optimization of aggregate prospective utility of the task and the safety of the motion trajectories.