| Abstract |
This paper considers a novel parallel control algorithm of Best Weight CMAC (Cerebella Model Articulation Controller)
and PD (Proportional Derivative) for the actuator electric load simulator of Unmanned Aerial Vehicles (UAVs). Although the conventional parallel controller can effectively reduce loading error and restrain the surplus torque of load simulator, it could destabilize the system due to overlearning of CMAC when tracking continuously variable loading signals like sinusoidal instructions. The overlearning issue of CMAC is investigated in detail and its divergent trend is predicted. A
novel weight updating rule of CMAC, Best Weight Method, is proposed and its convergence is theoretically analyzed.
This novel parallel controller can not only retain the normal learning process of conventional parallel controller, but also
avoid the overlearning of CMAC. The experimental results have demonstrated that the proposed controller has good
robustness, can effectively reduce the loading error, eliminate the surplus torque and assure the stability of system as well. |
