Caterpillar Robot Locomotion Based on Reinforcement Learning Using Subjective Reward

This note presents an application of reinforcement learning to caterpillar robot locomotion. An excellent advantage of reinforcement learning is that an action can be acquired using only a simple reward. In our previous work, the reward was a forward distance measured using a sensor. This reward was completely an “Objective Reward.” On the other hand, this study uses the reward given by the human's subjective judgment, which is defined as “Subjective Reward.” The main purpose of this study is to compare its performance between the “Objective Reward” obtained from the sensor and the “Subjective Reward” given by a human teacher. The results show that the “Subjective Reward” can give better results than that of the “Objective Reward”, because the “Subjective Reward” has more information than the “Objective Reward”. On the other hand, this note discusses the good teacher who gives an excellent “Subjective Reward”