Neural Network Synchronization Control For Multi-manipulator Systems In Task Space

In the case of increasingly complex work tasks and increasingly higher work accuracy requirements,a single manipulator can no longer meet the requirements of some work tasks.Synchronization and collaborative control of multi-manipulator systems has attracted more and more experts and scholars.And mechanical arm usually exist parameter perturbation and vulnerable to the interference of external environment factors,such as the nonlinear friction of mechanical arm joint,the end of the actuator operating object diversity,etc..It is a typical multi-input,multi-output,strong coupled complicated nonlinear system,and often difficult to obtain all its structural and dynamic parameters.In this case,the traditional control method is often difficult to meet the control accuracy requirements.In engineering applications,most of the operation tasks are planned in the task space,so designing synchronization controller directly in the task space is more conducive to the implementation of control tasks.In this paper,firstly,the dynamics model of the multi-manipulator system task space is established based on the dynamics model of the manipulator joint space.The communication topology of the multi-manipulator systems is established based on the graph theory,and the synchronization error of each sub-manipulator system is defined.On this basis,a multi-manipulator task space adaptive synchronous controller based on RBF neural network is designed.The weight of RBF neural network is updated online by using adaptive law.The effectiveness of the designed controller is verified by Lyapunov stability analysis and simulation.Aiming at the multi-manipulator system with uncertainty,firstly,the system model uncertainties is analyzed,and the corresponding sliding mode synchronization controller is designed,then the model uncertainty is approximated and compensated by RBF neural network,and the task space sliding mode synchronization controller based on RBF neural network is designed.Aiming at the multi-manipulator systems with unknown model,combined with the sliding mode control method,a model-free adaptive synchronization controller is designed by using multiple independent RBF neural networks approaching to each sub manipulator system and.Through the continuous on-line iteration process of the weights of the neural network,the real-time approximation of the dynamic model of the multi-manipulator which are changing with the working tasks can be realized.Using this method can break away from the limitation of mathematical model and enlarge the application scope of the controller.Finally,the weight iteration process is optimized,which a single parameter is used to replace the weight of the neural network and the RBF neural network synchronization controller based on this method is designed to better meet the requirements of real-time control.