Intelligent Robot Control Algorithms Applied To Human-Robot Interaction

Robot technology has developed rapidly,and it has been widely used in industrial production,scientific research and military armament.Compared with traditional manual work,robots have the advantages of high precision,high repeatability,low errors,etc.More importantly,robots can operate in a variety of harsh environments and can work continuously,which is impossible for ordinary production workers.However,due to factors such as the nonlinearity of the model and the uncertainty of the inertia parameters,the robot system is difficult to be modeled in practical applications.Therefore,the intelligence of robots is often not high under traditional control algorithms.In addition,in the fields of industrial processing,assembly,etc.,multi-robots cooperation can accomplish more complex,more dexterous and wider operating space tasks than a single robot,which is more research significances and application values.However,the multi-robots cooperative control system currently faces the problems of high coordination control difficulty,long motion programming cycles and low intelligences sharing.Therefore,this thesis develops adaptive neural learning controllers and design a human-robot demonstration system based on mixed reality.The main research of this thesis is to make robot control and interaction smarter,more efficient and more friendly.In this thesis,an adaptive finite-time convergence neural learning control algorithm is proposed for the manipulator system.The algorithm can also guarantee the neural learning and transient tracking performance in the presence of model uncertainty.The traditional adaptive neural controller only pays attention to the system control performance and rarely considers the neural learning performance.The neural learning algorithm based on finite-time convergence can make the neural weights converge to a small neighbourhood of the optimal values in a finite time under persistent excitation conditions.In addition,the system state variables errors conversion mechanism introduced in the controller design can convert the original constrained system into an unrestricted system,thereby improving the system transient tracking performance.This thesis also probes into the adaptive learning and generalization of neural networks sharing knowledge between different independent robot in isomorphic multi-robots cooperative systems.By establishing the communication topology between neural weights learning algorithms,the dynamic model sharing of robots is carried out.In addition,this thesis designs a robotic teleoperation demonstration system based on mixed reality technology,in order to enable people to interact with the robot in the remote location more safely,thus achieving more efficient and friendly robot teleoperation.This virtual demonstration system can expand the skills of the humans to the robots,thus broadening the skills range of the intelligent robots.