The Research On Control Of Complicated Robot Manipulators With Adaptive Neural Network

With the rapid development of science and technology,robotics has been widely applied to various areas,such as manufacture,aerospace,military,healthy and so on,and has been developing rapidly.The Robot main executive body is manipulators,which are a mechanical device to simulate human arms.Undoubtedly,there exist nonlinearity,strong coupling and uncertain disturbances in the manipulator dynamic system.At the same time,complicated operating condition,for example,the uncertainty of the load,the uncertainty of the system parameter,and unmolded dynamic,leads to the difficulty of accurate mathematical model to be set up.In the other hand,there will be a flexibility between the motor shaft and the actuating shaft in the transmission.Thus,taking flexibility of joint into consideration has been a focus and difficulty of the current research.Otherwise,for saving cost or reducing the error of measurement leading to some unmeasured state variable in the industrial application,it is becoming an important part of designing controllers for the design of the observer.Firstly,taking the adaptive sliding mold controller with the first-order filter observer addresses the uncertainty of system parameter and the partial unmeasured states in this paper.Then,an adaptive BP neural-network observer was presented with completely unmolded dynamics in the practical work.Contemporary,using the observed states,the back-stepping controller was applied to track the desired trajectory for the manipulator.Secondly,singular perturbation based on a joint flexibility compensator,which is created by increasing bandwidth of the system filter in fact,was adopted to achieve trajectory tracking control of the flexible joint manipulator,especially weak joint flexibility.The RBF neural-network observer was presented to reconstruct unmeasured state variables and design a PD controller.Furthermore,combining self-learning ability and associative capability in neural-network with the ability to be easy to understand reasoning process in the fuzzy system for the more advanced intelligent controller,the fuzzy-neural observer was used to estimate state variables and approximate the unknown nonlinearfunction.And then,the back-stepping controller was adopted to achieve trajectory tracking control of the manipulator.And the result of simulation demonstrated that the controller was reasonable and feasible as well.Finally,an experimental platform was building to communicate with the DSP target board and the servo system of the manipulator.Simultaneously,through running control algorithm and transmitting control signal to the servo system,the torque from the servo motor achieves trajectory tracking control of the manipulator.Meanwhile,without interrupting work of DSP,the CCS-Link,that is a module of the software in Matlab,was applicable to transfer data among DSP,CCS and Matlab in real time.At last,to be more important,the result of experiment verify that the control strategy was feasible and valid.