Extended Predictive Model-Mediated Teleoperation of Mobile Robots through Multilateral Control Michael Panzirsch1,2, Harsimran Singh1, Martin Stelzer1, Martin J. Schuster1, Christian Ott1, Manuel Ferre2

Abstract

Despite the substantial progression of autonomous driving systems, their application is often limited e.g. due to safety margins which can be caused by uncertainties in the environment reconstruction. Then, via teleoperation as a fallback solution, a human-in-the-loop can be introduced as the main decision maker. However, high delay in the communication channel distorts the performance of direct force Feedback teleoperation for example in space or disaster scenarios. On the other hand, model-mediated teleoperation can provide instantaneous and even predictive force feedback to the user, but the performance is limited due to state mismatches, incomplete models, model errors and the modeling challenges of complex wheel-ground contacts. Therefore, in this paper we ntroduce the concept of extended model-mediated eleoperation with a car like interface for mobile robots by fusing local fictitious and remote force feedback, which can be measured, computed or fictitious. We provide a method to guarantee stability of the extended model-mediated teleoperation (involving time delay, multilateral coupling, fictitious force feedback and permanent updates of the local model) based on the passivity theorem. The benefits of the approach are highlighted by human-in-the-loop experiments with a wheeled mobile robot.