Genetic Algorithm based Obstacle Avoidance for 4-Wheeled Robot

This paper proposes an algorithm that uses for decision-making and control in Autonomous Mobile Robots (AMR) and performs in a virtual environment for simulation. In the decision-making stage, the algorithm uses the Velocity Obstacle (VO) and Artificial Potential Field (APF) methods to select the optimal velocity for avoiding collision and reaching the goal point in real-time. It additionally computes fitness that helps eliminates rapid changes in velocity. In control, the PID controller moves the robot along the velocity determined in the decision-making stage. The PID gain is updated by the gradient descent method to find the optimal gains. Proper brake torque is applied for the robot moves smoothly according to velocity previously determined. This parameter, $\alpha$ affects chance of avoidance obstacles, however it increase the travel time in experiment. The travel time averagely increase 9.54% at every $\alpha$ 0.1. When comparing the smallest and largest periods, the larger period resulted in a 29.9% decrease in control error. The results also showed that using the proposed PID controller resulted in a 13.6% decrease in control error compared to using a fixed PID controller.