Driver Modeling for Teleoperation with Time Delay

Abstract The results from a user study on teleoperated steering tasks are used to create a driver model under different latency conditions. The 31-subject user study explored the effects of latency on teleoperated steering tasks using a simulated mobile robot receiving input commands from a teleoperator via a computer gamepad. Using fundamental concepts from automotive steering models, and examining the users' input commands to the simulated robot under different latency conditions, a model of a human teleoperator for steering tasks was developed, tuned and validated. The model is a PD controller with feedback based on the projected lateral displacement of the robot. The tuning of the model gains for different latency scenarios reflects the real-world control strategies that users must employ when adapting to system latency. Simulation results show that the control gains can be interpolated to predict teleoperator performance under latency scenarios that were not tested with users. An analysis of the closed-loop stability of the system confirms our empirical observation that the ratio of the controller gains Kd/Kp increases as the latency increases.