A Strategy of Potential Fields and Neural Networks in the Control of an Autonomous Vehicle Within Dangerous Environments

Abstract

This article focuses on the development of an autonomous navigation system by generating real-time 3D maps of different urban environments with different properties within simulation software. This system used the Pioneer 3-DX vehicle, a LiDAR sensor, GPS, and a gyroscope. For the elaboration of the trajectory, the mathematical tool of artificial potential fields was used, which will generate an attractive field to a dynamic goal identified by the robot and repulsive to the obstacles present in the environment, recognized with great precision thanks to the use of a neural network. The topology neural network 8–16–32 was developed using forward propagation, reverse propagation, and gradient descent algorithms. By combining the tools of potential fields and neural networks, a path was traced through which the robotic system will be able to move freely under an off-center point kinematic control algorithm. Finally, a 3D map of the environment was obtained to provide information on the morphology and most outstanding characteristics of the deployment environment to users who use the system.