Research On Flexible Joint Robot And Its Kinematic Calibration And Vibration Suppression Research

Light-weight robots, because of their lightweight, human-like configuration, multi-sensory and so on, achieve widely employment and play very important roles in a lot ofspheres recently. At the aspect of end pose accuracy improvement, close-loop control isalways chosen. However, those end pose errors made by the manufacturing and assem-bling errors can not be eliminated by such control method. These errors can be reducedthrough robot kinematic parameter calibration based on end pose measurements. Besides,elastic elements like harmonic gears and torque sensors are always included inside light-weight robot joints to get higher load weight ratio and detect joint torque. Joint ?exibilityis imported since these elastic elements. Lightweight robots with ?exible joints are al-ways named ?exible joint robot. Joint ?exibility will cause vibration, and seriously affectimplements of robot. More serious problem is that the residual vibration can damage themeaning of the end pose accuracy improvement made by kinematic calibration. As a re-sult, this vibration must be suppressed. This paper will give a deep research on ?exiblejoint robot and its kinematic calibration method and vibration suppression.At first, a 6 degrees of freedom ?exible joint robot system will be developed, in-cluding mechanism system, sensor system and electrical system. Specifically, a generalanalysis to end-effector dexterity, requirement of load ability and requirements of inversekinematic decoupling will be implemented to decide the degrees of freedom and con-figuration. According to different location each joint would be placed, a kind of partialmodular concept is used to design robot joints. They will use the inside cabling struc-ture, and include features like lightweight, large ratio, high integration, multisensory andmechatronics. Besides, for the facility of repair, spring needle electrical connection struc-ture will be proposed.The kinematic error model which re?ected relationship between kinematic param-eter errors and end pose errors will be constructed to carry out kinematic calibration.According to the kinematic feature that shoulder joint axle and elbow joint axle paral-lel each other, a modified D-H parameter method will be applied to represent pose errorbetween two consecutive joints under this situation. To research residual vibration sup-pression problem, robot dynamic model will also be constructed. Because of the existence of joint ?exibility, joint angle along the motor side do not equal to the angle along the linkside. That is the main reason of vibration. In the dynamic equation, motor side joint an-gle is used as input to solve link side angle. However, for the reason of large magnitudeof system stiffness matrix, robot dynamic differential equations will become stiff differ-ential equations. Traditional numerical method used to solve this kind of equation willalways make the solution divergent. In order to obtain convergent results, the fourth orderimplicit Runge-Kutta method will be applied.Kinematic calibration method with minimum pose number selection will be re-searched base on kinematic error model. Including the minimum measurement posesnumber selection and optimization of pose set. Based on changes of observability indexand pose accuracy according to poses number increase, the minimum pose number wasselected. Robot pose accuracy will not improve a lot after the number of poses used in cal-ibration exceeds the minimum number. To avoid selection of unobservable poses, a kindof"change-add-change"method will be used to pose set optimization. Kinematic calibra-tion method with minimum pose number selection can balance efficiency and accuracy ofcalibration. That is to say, using least time to obtained better calibration results.To suppress robot residual vibration, an of?ine joint trajectory optimization methodwill be researched. The resource of residual vibration—residual elastic energy is chosenas objective function. A novel constriction factor particle swarm optimization methodwith punish function is proposed to optimize joint trajectory which corresponding to theminimum residual elastic energy, and finally to suppress residual vibration. In order toget smooth joint trajectory, decrease velocity mutation at start and end point, a modi-fied piecewise cubic spline function is designed as joint trajectory function. During jointtrajectory selection, if joint control points were not constrained, unsafe joint trajectorymay be generated. This trajectory may lead to dangerous situation like collision accident.Therefore, control increments which used to generate joint trajectory must be constrained.Unfortunately, particle swarm optimization method belongs to optimizing method with-out variable constrains. In this paper, punish function will be included to particle swarmoptimization, based on punish factor, constrains on trajectory control points will be in-cluded to a new objective function. So the joint trajectory optimization with constrainconditions is achieved.To suppress vibration during robot moving, a novel vibration prediction rule is pro- posed based on joint torque sensor. The three sections DKA ( Decelerating-Keeping-Accelerating ) real-time trajectory modification method is proposed to change joint pathat the time vibration trend is detected to decrease this trend and suppress robot vibra-tion which would happen in the future. The prediction rule include to conditions—jointerror torque exceeds torque threshold and its value is detected increasing within a timethreshold. Joint error torque is acquired by joint real torque measured by torque sensorand joint theoretic torque calculated from dynamic model. When joint error torque ful-fills those two conditions, the prediction rule would determine the generation of vibrationtrend and figure out future vibration. Initial joint expected trajectory is changed by thethree sections real-time trajectory modification method which composed three decelerat-ing, keeping and accelerating sections to decrease the vibration trend. Finally, vibrationsuppression during robot motion is achieved.