Simple Tuning and Low-Computational-Cost Controller for Enhancing Energy Efficiency of Autonomous-Driving Electric Vehicles

抄録

Previous studies have proposed various optimization algorithms, such as dynamic programming (DP) and model predictive control (MPC), to reduce the energy consumption of autonomous-driving vehicles. The difficulties in the industrial applications of these methods are their computational costs and tuning parameters. In this paper, we propose a linear quadratic regulator (LQR), a low-computational-cost algorithm. The proposed controller calculates the input within a sampling period of 10kHz. By the approximated linear-parameter-varying (LPV) modeling of a vehicle and a motor, we considered the energy loss in the cost function of the LQR. Thus, the proposed method had only one tuning parameter. The effect of changing this parameter, the solver of the LQR for the LPV model, and the influence of the approximation of the models were analyzed. We compared the proposed LQR and DP using computer simulations, a simulation bench, and field experiments. Based on these comparisons, the validity of the proposed method for enhancing the energy efficiency for industrial applications without additional computational hardware was demonstrated.