Machine Learning-Based Forward Collision Avoidance System: A Case Study for the Kayoola EVS

A forward collision avoidance system is an advanced driver assistance system that alerts the driver or maneuvers for safe motion in case of the occurrence of an imminent collision. In this research, an efficient reinforcement learning algorithm that actuates the car to move forward, steer left, right, and stop was designed for autonomous vehicles. Currently, forward collision avoidance systems are based on input commands from the sensors like Lidar and Camera to the system and the output is based on the commands initialized. With this model, the vehicle gathers data using an RGB Camera and collision sensor while moving on the road in a simulated environment. Scenarios are developed which include cars moving around corners, straight road, and in a more urban layout with other obstacles like cars within the environment. Reward flags are given for no collision and penalty for collision with obstacles within the environment. Model testing was done in Carla’s simulator and analysis of the model was done on a Tensor board and recorded simulation as the vehicle moves within the environment. An optimized deep Q-learning algorithm that relies on deep reinforcement learning was developed under constrained conditions in a Carla simulation environment with an overall accuracy of 64% and a metric loss of 20%. The algorithm relies on dynamic programming which can buffer data during the training process.