Abstract
Robots are showing great impact and in recent trends making an appearance in areas such as education and entertainment. Robotic soccer is becoming more prevalent in universities as projects for students as well as annual international competitions furthering research in robotics and Artificial Intelligence (AI). Reconfigurable robotics is used in application domains such as cleaning, multi-terrain locomotion and logistical support but the reconfigurability yet to be introduced in robotic soccer. This will offer potential in providing increased flexibility and adaptability in the game of soccer. This thesis focuses on using Reinforcement Learning (RL) to train an agent to kick a toward a goal using reconfiguration. A robotic soccer competition has been conducted and the variation of the ball trajectory with shapes, approach direction and reconfiguration have been analysed. With the use of Kinovea software, the videos that were captured from an overhead camera underwent motion analysis. RL was used with the Proximal Policy Optimisation (PPO) algorithm to train and optimise goal scoring. The environment was developed and trained in Unity. Training included the agent to learn to approach the ball in an optimal position to hit the ball into a goal using reconfiguration. Two use cases of penalty kicks and free kicks were used to validate accuracy of the proposed model which resulted in goal conversion of 81% and 67% respectively. Moreover, the results confirm that this method allows a reconfigurable robot to adapt to dynamic conditions in the soccer field and perform the best move out of the myriad possibilities that can occur in this complex yet competitive game.