Remote control of a semi-autonomous robot vehicle over a time-delayed link

Abstract

Remote control of systems over communication links having time delays of about 0.5 seconds or more is known to be a significant problem for human operators in teleoperation mode. "Move and wait" strategy is the operator's normal approach to the problem. To improve upon the inefficiency of this strategy, a number of solutions can be employed, including use of predictor displays, supervisory control and Smith Control. A predictor display aids the operator by presenting a non-delayed (ie. predicted) view of the remote system output. This allows commands to be issued before the actual delayed output is returned. Supervisory control allows the remote system to operate in semi-autonomous fashion by providing autonomous capability that can be directed by the operator using high-level commands. Smith Control provides stable, closed-loop control of systems with delay by effectively moving the delay out of the loop. This thesis presents a robot vehicle control system that includes all of these techniques in the overall design. The robot vehicle being controlled is intended for underground mining applications. This is a difficult environment for control systems. Movement of a vehicle in such an unconstrained environment coupled with the problem of sensor readings suggests an excellent application for fuzzy control. Remote control of a vehicle basically involves speed and steering. A multivariable, fuzzy control system has been developed for this task. A simplified version of Smith Control is provided to compensate not only for the time delay but also for the human operator's dynamics. This simplified method requires only the loop time delay. Semi-autonomous operation includes tunnel-tracking and turning into intersecting tunnels. Obstacle avoidance, based on fuzzy "obstacle factors" forms an essential part of the system. A neural network-based predictor is included for the operator's assistance. Simulation results prove that simplified Smith Control compensates for both the time delay and the human operator's dynamics. The length of the time delay is essentially irrelevant. The predictor allows the operator to control the vehicle over the time delay by driving the non-delayed predictor vehicle via a console display.