Emergency Driving Support Algorithm for Collision Avoidance

Abstract

In recent years, automakers have been trying to help drivers to avoid or mitigate collision with active safety system. Practical applications have become possible due to recent advances in exterior sensors and electronically controllable actuators. These advances have opened up many possibilities for active safety systems like lane keeping assistance system (LKAS), advanced emergency braking system (AEBS) and blind spot detection (BSD). The further enhancement of these technologies will lead to an automated driving system that requires a collision avoidance function using by automatic braking and even automatic throttle and steering. This dissertation focuses on automated collision avoidance using automated steering control and control of motor-driven power steering (MDPS) torque overlay and differential braking for emergency driving support (EDS). A robust Model Predictive Control (MPC) method is used in order to guarantee safety constraints despite of disturbances and model uncertainty. A minimum Robust Positively Invariant (RPI) set of vehicle state error is calculated and the robust MPC calculates the appropriate collision avoidance steering. The performance of the proposed algorithm has been investigated via computer simulations. The simulation studies show that the controlled vehicle can achieve safe collision avoidance maneuver using small lateral acceleration in a long distance preceding vehicle avoidance scenario, and it can achieve safe collision avoidance maneuver using high lateral acceleration in a sudden appeared vehicle avoidance scenario. Electrically controllable actuators, MDPS and differential brake system are used as actuators instead of automated steering and a radar and camera are used as a sensor system for the EDS algorithm. Using environment and vehicle information obtained from the sensor system, a risk of collision and drivers intention are determined. A trapezoidal acceleration profile (TAP) is generated incorporating the drivers intention and based on the TAP, the MDPS overlay torque is determined in order to assist the drivers speed of response. The differential braking is determined to maximize the minimum vehicle-to-vehicle distance to avoid collision. From the non-linear optimal control problem, the rule-based control algorithm is designed for real-time application. The performance of the proposed algorithm has been investigated via computer simulations and real-time human-in-the-loop simulations. The simulation studies show that the controlled vehicle can secure additional vehicle-to-vehicle distance in severe lane change maneuvering for collision avoidance. The success rate of collision avoidance has been investigated for 8 test drivers using the human-in-the-loop simulations. It has been shown that the most of the test drivers can benefit from the proposed support system.