Advanced Artificial Neural Network for Steering and Braking Control of Autonomous Electric Vehicle

Abstract

Sensors are necessary for an autonomous electric vehicle (AEV) system to identify its environment and take appropriate action, such avoiding obstacles and crashes. Despite their limitations about color, light, and non-metallic items, cameras, radar, and lidar are widely employed to detect objects surrounding a vehicle. Ultrasonic sensors are weather and light-resistant. Thus, the goal of this work was to create object detectors by combining multiple long-range ultrasonic sensors into a multi-sensor circuit. The Arduino processor incorporates an artificial neural network that uses the advanced artificial neural network as a novel approach to control the sensors. There are two steps to this method: offline training and implementation test. The most ideal neural network weights are found offline using the adaptive back propagation algorithm, and the best fixed weight is then embedded into the neural network software on Arduino for implementation test. Because the system can sense more detail about the vehicle's surroundings and accurately avoid obstacles, the definition of actions by taking the object's distance into consideration is better. As a result, the training can produce an output that is closed to the target with 0.001 errors.