Compliance Control Method Of Collaborative Robot In Multi Interactions For Unknown Surface Operations

With the development of aviation,aerospace,and automobile industries,the requirements for surface processing performance and efficiency are getting higher.In order to adapt to the high efficiency,high precision and high safety processing of different parts,the multiple interactions in human,robot,and environment nee d to be fully considered.Most of the previous studies only considered the single interaction problem between human and robot or robot and environment,and assumed that the surface features were known or could be obtained through some sensors such as vision.For the unknown surface processing and multiple interactions among human,robot,and environment,the compliance control methods of collaborative robots are develoepd in this dissertation which include trajectory representation and generalization on unknown surfaces,collision detection and external force estimation,reference pose generation and adaptive compliance control.The experimental system is set up to carry out some experiments.In order to solve the problems of trajectory representation and ge neralization of the manipulator on the unknown surface,the modified Dynamic Movement Primitive(DMP)method to solve the limitations of original DMP is proposed.As one of the trajectory representation methods,DMP has been widely used due to excellent performance in trajectory generalization,obstacle avoidance and robustness.However,the original DMP method cannot finish trajectory generalization when the goal point coincides with the initial point,the goal point is very close to the initial point,the new goal point and the old goal point are distributed on both sides of the initial point,or the trajectory is on the curved surface.To solve these problems,the modified DMP method is proposed which uses the adjusted cosine similarity to optimize and evaluate the similarity of the curves and the force coupling term to achieve force control.Finally,simulations and experiments are conducted to verify the effectiveness of the method.In order to solve the collision detection and external force estimation problems in human-robot-environment interactions,the extended momentum observer is developed.Most of the existed methods can solve the collision detection of the single arm.In this dissertation,the momentum observer method is extended to the solve the collision detection and force estimation problems of the gerneral object.In this way,the safety of the human-robot-environment system is fully guaranteed.The estimating external force is further used for force control during the cooperative operation of dual-arm robot.Finally,simulations and experiments are conducted to verify the effectiveness of the method.In order to solve the problem that the reference pose of the arm is difficult to obtain in unknown surface operations,the reference pose acquisition method is proposed.Estimating the surface normal vector and curvature based on force sensors has been extensively studied.However,in these studies,force analysis is often conducted for a special grinding tool,such as a sphere,a cylinder,etc..In addition,the friction model is simplified to get a concise expression.In this dissertation,the active surface learning method is proposed.Through learning on the known plane,the relationship between the force sensor information and the surface norma l vector is obtained.After that,this learned policy is applied to the unknown surface.The accuracy of position and direction control is further improved through adaptive pose compensation and iterative learning.Finally,simulations and experiments are conducted to verify the effectiveness of the method.It is difficult to achieve high-precision trajectory tracking and force control due to the complex dynamic coupling characteristics of the human-robot-environment multiple interactions,the uncertain dynamics,and the uncertain environmental impedance model.In order to solve these problems,a unified neural adaptive control method to achieve high stability control for the whole process containing multiple interactions is proposed.First,the multiple interactions are modeled as the equivalent second-order impedance model.Then,the human force and the environmental force are achieved by using a force sensor and a momentum observer.The selection matrix S is used to decouple the multiple interactions.Finally,the neural adaptive control method compensating position errors caused by the model uncertainty is addressed.Finally,simulations and experiments are conducted to verify the effectiveness of the method.Finally,the experimental system is built.Based on this experimental platform,a series of experiments are carried out including trajectory representation and generalization experiments based on improved DMP,collision detection and external force estimation experiments,position and direction control experiments on unknown surfaces,adaptive control experiments for multiple interactions.The results verify the algorithms proposed in this dissertation.The research in the dissertation provides a feasible solution for the safety and compliance control of the collaborative arm on unknown surfaces,multiple interactions.It has important theoretical and practical significance for the realization of intelligent manufacturing.