UGV Navigation and Tracking of a Moving Target Using Artificial Neural Network Model

There is much interest and research in applying Machine Learning (ML) methods for Unmanned Ground Vehicle (UGV) applications. This work is concerned with implementing a ML system based on Artificial Neural Networks (ANN) for a small UGV that navigates its environment, avoids obstacles, tracks a moving target, and optimizes its path. The work includes designing and implementing a wheeled robot prototype equipped with a camera and Time-of-Flight (ToF) sensors, Arduino and Raspberry PI controllers. The research methodology was divided into three main stages. First, data was collected from the manual operation of the UGV through different environments. Then, the data was used to build and tune a Feed-Forward ANN model. Finally, the ANN model was embedded in the UGV and tested in new environments. Results show that the UGV could maneuver its environment and reach its moving target in real-time.