A Trust-based Human-Vehicle Co-driving System with Model Free Adaptive Dynamic Programming Controller

Trust is one of the crucial factors influencing the performance and safety of human-vehicle co-driving system. The evolution mechanism of human-vehicle trust is studied in this work, and a trust-based steering control model (SCM) is designed to allow the autonomous driving system to adjust its control behavior based on real-time trust level, on the purpose of improve the efficiency and trust. The contributions made in this paper are as follows: 1) a novel quantitative model of human-vehicle dynamic trust is established for the first time in shared steering control by considering the deviation of human-vehicle driving expectations; 2) a trust-based steering controller using model free adaptive dynamic programming (MFADP) is designed which can solve the optimal control policy according to the value of trust without the dependencies on parameters of dynamic model of the controlled system. The rationality of proposed trust model and trust-based steering control method are validated by high-fidelity Carsim-Simulink simulations.