Innovative Production Machines and Systems Conference

The slide above shows some paths taken using RL and non-AI control from seven random starting positions in environment E3. It demonstrates that when using non-AI control the robot always attempts to head directly towards the recharging station even if this means attempting to move through obstacles. This results in a larger average number of moves, and in some situations an inability to move round an obstacle. This is because it cannot learn a new strategy. In contrast the paths for RL control show that RL can learn to find a path which navigates the robot around obstacles in order to reach the recharging station rather than attempting to move it through obstacles.