Collision avoidance of vehicles equipped with integrated differential and nonlinear model predictive control

Abstract

Every year, rear-front collisions, a significant form of accident, result in substantial economic and physical harm to individuals inside the vehicles involved. To address this issue, car manufacturers have put forth a solution known as collision avoidance, which employs subsystems like Autonomous Emergency Braking (AEB), Autonomous Emergency Steering (AES), or a combination of both. This research aims to introduce and enhance a comprehensive longitudinal-lateral integrated control system for collision avoidance, incorporating AEB and AES subsystems. The system presented in this study effectively manages the complete dynamic model of a passenger car and analyzes the surroundings of the host vehicle, while five different types of sensors, namely radar, ultrasonic, object camera, lane marker, and vehicle-to-vehicle acceleration transceiver are employed. The system also continuously monitors the behavior of the target obstacle following the activation of the AEB system, enabling the decision-making process to be updated based on newly observed motion behavior. A combined longitudinal-lateral Nonlinear Model Predictive Control (NMPC) approach is employed to govern the vehicle, working in conjunction with an additional derivative controller. The NMPC system takes into account inputs such as the steering wheel angle, throttle percentage, and brake pressure to ensure effective control. To evaluate the system’s performance under different circumstances, four scenarios involving three moving obstacles are executed.