Multi-Agent Machine Learning

5.6 Fuzzy Controller Structure

We use two inputs (fuzzy variables) to the fuzzy controller and generate one output from the fuzzy controller. The inputs for the pursuer are the angle difference and its rate of change . The inputs for the evader are the angle difference and the distance . We add the distance as an input to the fuzzy controller for the evader. The reason is that the evader has higher maneuverability than the pursuer and the distance between the evader and the pursuer is critical for the evader to decide if it needs to make a sharp turn.

For simplicity and to avoid the curse of dimensionality, we use two inputs and three fuzzy sets for each input to construct the controller. The pursuer's fuzzy sets are negative (N), zero (Z), and positive (P) for the angle difference and its derivative . The evader's fuzzy sets are negative (N), zero (Z), and positive (P) for the angle, ...