Model Predictive Control based Haptic Shared Steering System: A Driver Modelling Approach for Symbiotic Driving

Abstract

Advanced Driver Assistance Systems (ADAS) aim to increase safety and reduce mental workload. However, the gap in the understanding of the closed-loop driver-vehicle interaction often leads to reduced user acceptance. In this research, an optimal torque control law is calculated online in the Model Predictive Control (MPC) framework to guarantee continuous guidance during the steering task. The novelty lies in the integration of an extensive driver-in-the-loop model within the MPC-based haptic controller to enhance collaboration. The driver model is composed of a preview cognitive strategy based on a Linear-Quadratic-Gaussian, sensory organs, and neuromuscular dynamics, including muscle co-activation and reflex action. Moreover, an adaptive cost-function algorithm enables dynamic allocation of the control authority. Experimental data was gathered from 19 participants in a fixed-base driving simulator at Toyota Motor Europe, evaluating an MPC controller with two different cost functions against a commercial Lane Keeping Assist (LKA) system as an industry benchmark. The results demonstrate that the proposed controller fosters symbiotic driving and reduces driver-vehicle conflicts with respect to a state-of-the-art commercial system, both subjectively and objectively, while still improving path-tracking performance. Summarising, this study tackles the need to blend human and ADAS control, demonstrating the validity of the proposed strategy.