Approach of Kinematic Control for a Nonholonomic Wheeled Robot using Artificial Neural Networks and Genetic Algorithms
Date:
The quest for an improvement in the quality of life of human beings directs the idea of integrating different areas of knowledge to solve problems that it is impossible to solve using traditional methods. The presented idea applies to the development of autonomous robotic systems; such as cars, wheelchairs, boats, and airplanes. In those cases, given a high level of autonomy can bring countless benefits to the quality of life of human beings. In this work, a reactive navigation hybrid controller for a nonholonomic mobile robot is presented. The controller was designed using the algorithm "Neuroevolution of Augmented Topologies" (NEAT) and trained using a developed simulator, which integrates different areas of knowledge such as control, system modeling and discrete-time simulation. The used methodology allowed to reach a level of autonomy for the vehicle, obtaining a stable controller with good performance in the analyzed scenario.