Fuzzy Logic Based Collision Avoidance System For Autonomous Vehicles

Abstract

To decrease traffic accidents in the real world, scientists developed Advanced Driver-Assistance Systems (ADAS) to avoid accidental human error and make the driving experience better. Many types of research in the (ADAS) programs have been reported. The vehicle path tracking with the Collision-Avoidance System (CAS) and the lane assists, departure warning with the rear-end collision avoidance are good examples. However, they are not fully automated yet. A fuzzy logic controller framework is proposed in this thesis for improving CAS in self-driving vehicle navigation. The framework of the thesis accommodates two navigation controllers for tracking a heading-vehicle and a reference path. It uses the accurate GPS/INS/OD integrated sensor, which gives its data to the navigation controllers. The controllers process this data and give the desired tire angle to keep the follower-vehicle on the right track till reaching the reference path or a fixed distance to the heading-vehicle. The framework also accommodates a fuzzy collision avoidance controller (CAC) processing the 2D laser information in case of a collision or pedestrian detection, leading to safety maneuvering or sudden stopping. The previous controller's decisions are propagated to the vehicle's actuators by the last fuzzy controller called the weight controller (W), which mixes the decision among the controllers or gives the upper hand to the appropriate controller whether to maneuvering, stop, accelerate, or deacceleration. Experimental work was performed on the GAZEBO simulation tool, which has a robust platform with vehicle dynamics. On the other hand, the fuzzy controllers were designed and implemented on MATLAB Simulink. The framework used the robotics operating system (ROS) to integrate between GAZEBO and the MATLAB by (ROS) messages synchronously to control the vehicle actuators/ sensors.