| Summary: | 碩士 === 長庚大學 === 機械工程學系 === 100 === The purpose of this thesis is to develop a robust and practical controller for an experimental overhead crane system to smoothly transport payloads.
In order to exploit the readiness of industrial servo motors, the motor drivers were assumed to be set in velocity mode. Besides, Microsoft Kinect sensor was used for the estimation of payload positions to derive the swing angle. The acquiring of swing angle using Kinect introduces significant time delay, approximately 0.23 seconds, which should be counteracted by the controller. A adaptive neural network was developed for the estimation of real-time swing angle, compensating the delay.
A sliding mode controller and a PD controller were developed for the crane. Both use the adaptive neural network for swing angle feedback. To achieve anti-swing performance, there are more design parameters for the sliding mode controller. The PD controller, although relatively simple, requires reference trajectories to follow.
Experimental results using a prototype system show that adaptive neural networks are superior in angle estimation than the neural networks. And both controllers are able to achieve the requirement of smooth manipulation. In comparison, the sliding mode controller was more robust than the PD controller in the face of disturbances.
|
|---|
