Improving the Prediction Accuracy of Predictive Displays for Teleoperated Autonomous Vehicles

Abstract

Autonomous vehicle teleoperation is used to remote vehicle control, manage car-sharing fleets, or in case of autonomous driving failure. Yet, the communication delay is one of the major challenges that jeopardize system stability and transparency. Vehicle trajectory prediction as the Predictive Display (PD) is the state-of-the-art technique to mitigate this problem. However, its effectiveness is highly dependent on the remote vehicle inputs and actual communication latency. To solve this issue, we propose a new model that considers the operator inputs in the control loop. To validate the feasibility of the proposed approach, an evaluation was conducted. The novel predictive model was implemented in the BMW R&D virtual simulator. Experimental results show that the proposed model predicts the vehicle states with 7.3% less euclidean deviation.