Neuro-fuzzy control of satellite attitude by reinforcement learning

Neuro-fuzzy controller with reinforcement learning is applied in the attitude control of satellites. The detailed design method is presented and the algorithm of reinforcement learning is deduced. Parameters of the controller are adjusted only by reinforcement signal, but not by the learning sample. Simulation results show that the method can effectively copy with the uncertainty of satellite and thus posses good robustness. Under the proposed method, higher precision of attitude control of satellite can be achieved.