IMPROVING THE PERFORMANCE OF OBSTACLE DETECTION AND AVOIDANCE AUTONOMOUS MOBILE ROBOT USING TRANSFER LEARNING TECHNIQUE

Subject Area: ROBOTICS

This paper is targeted at improving the performance of an obstacle avoidance autonomous mobile robot using transfer learning technique. This research was embarked on to address the problem of obstacle collusion in mobile robots. To address this problem, transfer learning was adopted was Alex.Net algorithm and used to develop an improved obstacle recognition algorithm using convolutional neural network. The method was guided by Dynamic Systems Development Model (DSDM) methodology, while the system design was done using objected oriented system analysis approach. The model was implemented with Simulink and evaluated. The result after cross validation showed high obstacle detection and recognition accuracy of 98.7%. The accuracy achieved by the algorithm was further justified using comparative algorithm with other state of the art obstacle detection and recognition algorithms. The result showed that the adopted algorithm have a percentage improvement of 1.89%.