Interactive Corridor-Based Path Planning for Teleoperated Driving

Abstract

Teleoperated driving, where an operator remotely takes over for some time-span, enables a fallback mode for autonomous vehicles. Complex urban scenarios, however, involve a too high workload for the operator. The main reasons for this are the communication delay and the lack of three-dimensional perception. This contribution presents a supervisory control concept with distinct human-machine interactions that explicitly addresses highly complex and sensitive driving tasks. By integrating automated driving functions and human-predefined corridors, a corridor-based planning scheme based on a shared autonomy approach is derived. Within the specified corridor, a modified Constrained CHOMP algorithm determines a collision-free path online using LiDAR sensor information. Since the operator remains outside the closed control loop, the proposed approach is insensitive w.r.t. any time delays. Real driving and simulation experiments confirm the effectiveness of the proposed approach and highlight both the achieved driving safety and the real time capability.