Attractive Region In Environment Based Method For Robotic Optimal Grasping And Compliant Assembly

Robotic manipulation is an important issue in the research field of robotics and can be widely used in industries,service industries,national defense,and medical fields.Therefore,it can greatly expand the scope and applications of robots.However,compared with the traditional structured environment,errors,such as sensor errors,object model errors,and robot control errors,are wildly existing in an unstructured environment and can hardly be eliminated.They are huge challenges for autonomous and stable robotic manipulation.Besides,when robots perform manipulation tasks,they would often interact with the environment,so compliance of robotic manipulation should be ensured.In all,considering the errors in the unstructured environment and the requirements for manipulation compliance,designing effective robotic manipulation strategies and compliant control methods are of great significance to improve the stability,robustness,and compliance of the robotic manipulation.The attractive region in environment method is firstly proposed in the study of robotic peg-in-hole assembly tasks.By using environmental constraints,the uncertainty of the peg’s position can be effectively eliminated to achieve successful assembly tasks.Later on,the attractive region in environment method is extended and applied to robotic grasping and localization tasks.Designing robotic manipulation strategies with the help of the attractive region in environment method can eliminate part of uncertainty of the system states,but it is necessary to design more robust strategies given the problem of various errors in an unstructured environment.On this basis,considering the compliance to the environment during the robotic manipulation,it is necessary to design a compliant strategy and controller.Given this,in order to solve the problem of various errors in the unstructured environment and the requirement for the compliance of robotic manipulation,this thesis adopts the method based on the attractive region in environment and impedance control and systematically studies the robotic optimal grasping and compliant assembly strategies and control methods.The main contributions of this thesis are listed in the following:(1)In the robotic high-precision assembly task,the uncertainties of the peg’s states and the excessive contact forces between the peg and hole in the robotic assembly are likely to cause jamming or damage to the objects.For these problems,a force-guided robotic compliant assembly strategy is proposed based on the attractive region in environment method.By modeling and analyzing the configuration space of the peg in hole,the attractive region in environment is constructed and the feasible range of the peg’s initial position and assembly action strategy is planned.On this basis,a strategy is designed by using the direction of the contact force to guide the horizontal movement of the peg and adjusting the peg’s orientation "from coarse to fine" to realize the robotic high-precision and compliant peg-in-hole manipulation.This strategy can effectively eliminate the position errors of the peg,and only the direction of the contact force is used to adjust the position of the peg to ensure that the contact force between the peg and hole during the robotic assembly process within a safe range.Therefore,it can improve the robustness and compliance of the robotic assembly.(2)In the robotic autonomous grasping task,the various uncertainties in the unstructured environment pose huge challenges to the robotic stable grasping.Considering this problem,a strategy based on the attractive region in the environment and deep learning is proposed for robotic grasping with a four-pin gripper.This thesis models and analyzes the configuration space of the four-pin gripper when it grasps three-dimensional objects.By constructing the attractive region in the environment,form-closure grasping points are planned.Considering the uncertainties in the unstructured environment,a calculation method of grasping points robustness is proposed.Moreover,a four-pin robotic grasping dataset is generated independently.Using this dataset,a grasping quality evaluation network is trained for the robot to quickly select the robust grasp points.In the unstructured real scene,this strategy can realize robotic fast and stable grasping of a variety of objects of different shapes.(3)In the robotic stable grasping and in-hand manipulation tasks,given the complex calculation of the multi-fingered robotic grasping configuration and the problem that excessive grasping force or environmental force will damage the object,this thesis proposes a robot torque control based on optimal grasping and impedance control for robotic in-hand manipulation.Considering the multi-fingered robot stable grasping with constraints,the minimum grasping force optimization function is proposed.And the grasping quality index is designed to optimize the grasping configurations,so as to plan the optimal grasping configuration for the multi-fingered robot.On this basis,a torque controller for the multi-fingered robot based on impedance control is designed to track the desired trajectory of the object and keep it compliant with the environmental force.This method can realize the multi-fingered robotic stable and optimal grasping,and at the same time ensure the compliance of the robotic in-hand manipulation to prevent damage to the object.The research results of this thesis can provide a theoretical basis for improving the stability,robustness,and compliance of robotic manipulation.At the same time,it is of great value for the wider applications of robots in practice.