THE KINEMATIC MODELING OF A FOUR MECANUM WHEEL ROBOT FOR ENVIRONMENT MAPPING

This research presents the kinematic modeling of a four mecanum wheel robot for environment mapping and navigation. The study developed the kinematic model of the robot and then used Bayesian and kalman filter to improve the Simultaneous Localization and Mapping (SLAM) of 2D LIDAR sensor used for scanning. Q learning algorithm was used as an agent in reinforcement learning to control the action of the robot with respect to the SLAM data input. The performance of the learning algorithm when trained with 5000 episodes of the environment, achieved loss function approximately zero which implied that the episodes was correctly learnt. When deployed on the robot and then tested in an environment with strategically places obstacles, the result showed that the robot was able to learn as it navigated from each episodes until it was able to correctly avoid obstacles in the path and carry out its desired function.