Research On Collision Detection And Compliant Control Based On Joint Torque Sensor

The collaborative manipulator can directly interact with people to complete complex and varied tasks,and improve work efficiency,therefore,it has become a hotspot in the field of robotics research gradually.In the process of human-robot interaction,the manipulator is required to be able to perceive the interaction force with human beings.At the same time,in order to ensure human safety,the manipulator must be able to detect the collision with the environment.Therefore,the external force sensing ability of the manipulator is particularly important.At present,joint torque sensors are often used in most researches to perceive the force information.To estimate the external force of the robot arm by using the torque sensor,it is necessary not only to deal with the load torque brought by the inertia of the robot arm,but also to solve the problem of torque interference and signal noise of the torque sensor itself.In order to solve the problem of torque ripple and temperature drift of joint torque sensor,a method of hardware compensation for torque ripple and temperature drift compensation is proposed.Aiming at the torque ripple caused by the deformation of the flexible wheel in the transmission process of the harmonic reducer,the method of eliminating the torque ripple is studied.In view of the influence of the joint temperature increase on the performance of the torque sensor during the continuous operation of the manipulator,the temperature compensation module is designed to solve the temperature drift problem of the sensor.The dynamic calibration of torque sensor is accomplished by using dynamic calibration platform.The estimation of the external moment is completed by using the disturbance observer based on generalized momentum.The disturbance observer can track the external moment of the manipulator quickly and accurately.Aiming at the observer residual problem caused by inaccurate parameters in the dynamic model of the manipulator,a time-varying threshold algorithm is proposed to envelope the observer residual.When the observed value exceeds the threshold limit,the collision is judged.A method of signal noise processing in collision detection is proposed to reduce the noise of observer and retain the original information of collision force.A position-based compliance control algorithm is proposed for the problem that the compliance control algorithm can only interact with the end force sensor and is difficult to apply in commercial robots.The proposed control algorithm including force-free control,impedance control and impedance control with force tracking.These three compliance control algorithms make use of the external force estimates of the observer and can be used for compliance interaction at any position of the linkage.To test the performance of torque sensor and verify the effectiveness of the proposed algorithm,a single-joint experiment platform was designed,and the dynamic parameters of the single-joint were estimated by optimized excitation trajectory.Using the torque information,the effectiveness of the disturbance observer based on generalized momentum is verified,and the collision detection experiment of the manipulator and the position-based compliance control algorithm experiment are further completed.