DMIC: Decision-Making Integrated Predictive Control for Intelligent Vehicles

The active crash avoidance system of intelligent vehicles faces challenges in accurately triggering and achieving effective multiobjective coordination control in complex driving environments characterized by multiple vehicles and variable road conditions, thereby posing a threat to driving safety for human. To address those issues, this article introduces a decision-making, path planning, and tracking integrated predictive control method (DMIC). First, DMIC incorporates vehicle actuation system characteristics, road conditions, and environmental information to design a dynamic characteristic-based risk indicator, which is applied to design different control modes. DMIC then applies continuous activation functions to activate optimized states and control inputs under various control modes, forming an integrated event-triggered continuous decision-making objective function. Afterward, DMIC employs the receding horizon optimization to calculate the front-wheel steering angle and wheel torques based on the integrated predictive model and time-varying constraints. Verification on a driver-in-the-loop (DiL) platform demonstrates that DMIC can accurately and smoothly switch among different optimization states, ensuring crash avoidance with arbitrary approaching vehicles under varying road conditions, thereby maintaining smooth vehicle responses, enhancing the driving stability and ride comfort, while meeting real-time and robustness requirements.