Tele-autonomous Obstacle Avoidance

by Johann Borenstein and Yoram Koren, The University of Michigan

Tele-autonomous obstacle avoidance allows a human operator to steer vehicles or robots remotely at high speeds and in cluttered environments, even with poor visibility or without visual contact at all (see Fig. 1.3.1). Tele-autonomous operation does also allow for safe operation of remotely controlled vehicles with significant time delays or communication disturbances.


Fig. 1.3.1: Conventional tele-operated vehicles or mobile robots rely on visual contact with the operator, either directly or through video transmissions, as shown above. Guiding such a vehicle is a formidable task, often complicated by the limited view from the TV camera. Under such conditions, a human tele-operator must exercise extreme care, especially in obstacle-cluttered environments. Consequently, the actual traveling speed of the vehicle might be very slow. When dust, smoke, or steam inhibit vision-based guidance, conventional tele-operated activity is ruled out altogether.


Principle of Operation

The tele-autonomous system has a shared control architecture. In shared control systems, a complex task is divided between the human operator and the machine. In our system, the global control task of guiding the robot to a target is performed by the operator in the external loop, as shown in Fig. 1.3.2. The local control task of guiding the mobile robot around unexpected obstacles is performed automatically in the internal loop by the robot's onboard computer. The control tasks are not switched in a bang-bang control fashion. Rather, these tasks are smoothly integrated, and the level of control shifts gradually from the operator to the machine, and vice versa.


Fig. 1.3.2: We have developed a system that combines autonomous obstacle avoidance with tele-operation into what we call a tele-autonomous system. In this system, the tele-operator can guide the vehicle even without any visual contact. The vehicle follows the general direction prescribed by the operator; however, when the robot encounters an obstacle, it autonomously avoids collision with that obstacle, trying to match the operator's prescribed direction as closely as possible.


As seen in Fig. 1.3.2, the internal loop is of a sampled-data type. The internal loop provides a sampling period T that is much faster than the delay period T_d of the external loop. Depending on the application, T may be several orders of magnitude faster than T_d. This allows tight control of the vehicle, even when travelling at very high speeds.

Information about the robot's environment is derived from its onboard sensors and is shown on the operator's screen, which closes the external control loop. The visual information enables the operator to steer the robot out of difficult trap situations and to the designated target. Note that the communication bandwidth may be as narrow as only a few hundred Hz.

Applications


Click here for Project History


For more details on tele-autonomous control, see paper 15.
This file last updated on 7/4/96 by Johann Borenstein

johannb@umich.edu