| With the progress of industrial production,most emerging manufacturing tasks are either too complex to be automated and manually operated,or even impossible to be completely completed by robots or humans.This urgently requires robots to work together with humans.During the interaction between robots and humans or the environment,phenomena such as system instability,accuracy loss,saturation,and mechanical nonlinearity can occur,seriously affecting system performance.For collaborative robot systems,in addition to ensuring high-precision closed-loop motion control of the robot,it is more important to be able to adapt to different work tasks.Therefore,modular robots with different modules and the ability to quickly assemble the most suitable for completing a given task are increasingly widely used in the field of human-robot collaboration.Modular robots are organically connected modules with simple functions and certain perceptual abilities.Modular robots are usually composed of a set of modules with different sizes and performance characteristics,through which workers can quickly assemble the most suitable robot for completing a given task to adapt to different working environments and task requirements.Therefore,modular robots undoubtedly have high adaptability to humanrobot physical interaction tasks.On the one hand,the integration of multiple sensors in modular robots further enhances the robot’s environmental perception ability,enabling it to complete tasks in highly unstructured and random environments,avoiding excessive interaction forces that can cause harm to humans or damage to the robot’s body.On the other hand,modular robots can meet the conditions required by humans in physical human-robot interaction tasks.In the process of human-robot physical interaction,modular robots engage in a large amount of force interaction between human upper limbs and the end of the robot.If humans cannot take the correct measures in real-time for the current environmental task,it can damage some parts of the robot,damage the performance of the robot,and even pose a threat to the human body during robot interaction.Therefore,ensuring the safe and stable completion of corresponding task objectives in the human-robot physical interaction system,and maintaining the interaction force between modular robots and human upper limbs within a safe range,making modular robots demonstrate flexibility,is of greater research significance.In addition,one of the key issues in human-robot physical interaction tasks is to enable robots to understand the motion intentions of their human partners,so that they can "actively" collaborate with humans.In this regard,it is not applicable to make the robot track the prescribed trajectory.When the robot changes its motion based on the force exerted by humans,once humans want to change their motion intention,it naturally becomes a load on the robot.To solve this problem,it is necessary to estimate the motion intentions of human partners and integrate them into robot control.This article focuses on the above issues and studies the impedance control research method of modular robot systems for human-robot physical interaction.The specific content includes:(1)Decentralized Robust Control of Human Modular Robot Based on RBF Neural NetworkThe dynamic equation of the robot with joint torque sensor is derived,which explicitly represents the nonlinear multivariable structure.Assuming that each modularization is integrated by a rotary joint and a joint torque sensor,this paper establishes a dynamic model of the modular robot by installing a torque sensor on the joint for torque feedback,which enables the conventional control algorithm to be applied to the modular robot and ensures that the robot can obtain high control accuracy.It lays a solid foundation for the following human intention identification,adaptive impedance control method and model uncertainty compensation.(2)A Control Strategy for Physical Human-Modular robot Interaction Based on RBF Neural NetworkUsing the establishment of a modular robot system dynamic model based on joint force feedback,the impedance model of the upper limb is established and simplified.The human motion intention is further analyzed as an unknown function of interaction force,upper limb end position,and velocity.At the same time,a radial basis function neural network is used to dynamically estimate the motion intention of cooperative person,and accurate human motion intention is calculated.In order to eliminate the disturbance of model uncertainty when the dynamic model is established,a decentralized robust controller is designed to dynamically compensate the model uncertainty.To enable modular robots have active flexibility and effectively limit interaction forces,impedance controllers are designed to maintain the desired impedance relationship between the modular robot end effector and interaction forces.Finally,the Lyapunov theorem is used to prove the tracking error ultimately uniformly bounded of the dual closed-loop control system.(3)A Physical Human-Modular Robot Interaction Control Strategy Based on Impedance Parameter Self LearningDesign a human-robot interaction control strategy based on impedance parameter selflearning,which can not only achieve motion constraints on the robot,ensure the safety performance of human-robot physical interaction,but also improve the flexibility of humanrobot physical interaction.Based on the proposed method for estimating human motion intention,a self-learning method for human impedance parameters is considered.Combining joint torque feedback information,joint position,and other information,an adaptive impedance control strategy combined with neural networks and motion constraints is proposed for modular robots that collaborate with humans to perform tasks.The uniform ultimate boundedness of trajectory tracking errors is proved through Lyapunov theorem.Finally,an experimental comparison was conducted between the proposed control strategy and existing impedance control strategies to verify the effectiveness of the proposed control strategy. |
PDF Full Text Request