A Robust Non-Cooperative Control Approach for Human-Machine Shared Driving

In vehicles employing human-machine shared control, the human driver and the advanced assistance driver systems (ADAS) often generate conflicting expected trajectories, which must be fused to generate one trajectory to optimize driving comfort, to maximize energy efficiency, and to mitigate risks on the road. The application of existing game-theoretical approaches to the human-machine shared driving problems is not straight-forward. One of the main obstacles lies in the unique unpredicted characteristTics of the human driver in a human-machine system. For example, the drivers' reaction lag and possible driving errors can hinder precise adherence to his/her expected trajectory, significantly impacting the equilibrium of a conventional game-theoretic approach. This paper proposes an receding horizon robust non-cooperative approach to address this problem. We show that, the robust non-cooperative formulation can guarantee the robustness of human-machine equilibrium, in the presence of driver's reaction delay and possible driving errors. Simulations based on human-machine lane-changing task is provided to validate the efficacy of the proposed non-cooperative approach.