鈴木 隆

<jats:p>Torque vectoring (TV) is a commonly used method for four in-wheel motor electric vehicles (4-IWM EVs). Several existing studies based on model predictive control (MPC) focus on improving system stability and energy efficiency by minimizing or maximizing a performance function, defined as the time integral of the weighted sum of two cost functions. However, this approach must address the challenge of balancing these two objectives. Furthermore, the MPC framework lacks sufficient robustness against model uncertainties and external disturbances. This study proposes a two-layer TV controller for a 4-IWM EV, designed to enhance both robustness and energy efficiency, as specified in the Autonomous Driving Control Benchmark Challenge of IEEE CDC 2023. The first layer includes a direct yaw moment control module and a longitudinal force control module based on first-order sliding mode control, integrated with nonlinear disturbance observers (NDOBs) to estimate disturbance amplitudes from rough roads and reduce chattering. The second layer employs MPC to allocate torque among the wheels to minimize total energy consumption. Simulations, performed using a full 4-IWM EV simulator developed in Modelica, focused on the ISO double lane change on a rough road, as required by the benchmark challenge. The results demonstrate that the proposed control system significantly improves the vehicle’s robustness and energy efficiency in this scenario.</jats:p>