MODELING A DEEP LEARNING AND FUZZY LOGIC-BASED BEHAVIORAL APPROACH FOR AUTONOMOUS NAVIGATION OF A ROBOT IN A GLOBAL POSITIONING SYSTEM (GPS) DENIED ENVIRONMENT

This paper presents the modeling of a deep learning and fuzzy logic behavioral approach for the autonomous navigation of a robot in a Global Positioning System (GPS) denied environment. The study was aimed at addressing the optimization problems experienced by mobile robots due to the dynamics of an environment as a result of global positioning system unavailability. This problem was addressed by collecting data from the workspace environment and then training a deep neural network model to generate a cognitive algorithm that was used for intelligent Simultaneous Localization and Mapping (SLAM) using the fuzzy logic approach. The algorithm was integrated into a deferential wheel drive robot using Simulink and was tested. The result showed an accuracy of 99.80% and a loss function of 0.20%, which implied good training performance and SLAM intelligence. The deep fuzzy algorithm when integrated into the robot and tested in a dynamic environment that has no GPS was able to intelligently maneuver obstacles in the workspace.