Interaction Torque Estimation Algorithm for Wrist Gimbal Rehabilitation Robot Mohammadhossein Saadatzi, David C. Long and Ozkan Celik Biomechatronics Research Laboratory, Department of Mechanical Engineering, Colorado School of Mines ļ Wrist Gimbal is a two degree-of-freedom (DOF) wrist and forearm rehabilitation exoskeleton. ļ Active DOF are wrist flexion/extension and forearm pronation/supination. Implementation of force/torque estimation algorithms requires a dynamic model of the robot, which is realized using Denavit Hartenberg convention, and Newton-Euler dynamics formulation. SolidWorks model and assigned coordinate frames for wrist gimbal. šš = š“š“ š½š½ š½š½Ģ + šŖšŖ š½š½ š½š½Ģ + š®š®(š½š½) š¾š¾šš šššš + š¾š¾š£š£ ššššĢ − Disturbance Observer + ššĢ šš ššĢĢ =-LššĢ + šæšæ{šš ļæ½ šš ššĢ + šš šš š¶š¶Ģ šš, ššĢ ššĢ + šŗšŗļæ½ šš } Estimated interaction torques: ššĢ šššššš 0.4 Dynamic model+Simplified friction model Motor torque 0.2 0 -0.2 Friction Model Ģ šššššššš Ģ = šš(šš) šššššššš Ģ -0.4 2 0 4 6 8 10 12 time (sec) 14 16 18 20 14 16 18 20 Second joint 0.05 ļ For torque estimation, first the total disturbance torque at the robot handle šššš is estimated using a nonlinear disturbance observer (NDO). ļ The estimated torque ššĢ šš includes the friction at the joints and the torque applied by the patient/user. ļ By subtracting the friction of the joints šššššššš from the estimated disturbance, the torque applied by the user's hand to the robot handle can be calculated. Friction Modeling First joint frictional torque Original friction model Simplified friction model Experimental data Estimated motor torques, using dynamic and friction models, closely track the recorded motor torques. 0 -0.05 -0.1 -0.15 -0.2 2 0 4 6 8 12 10 time (sec) Validation of the NDO algorithm via the same experiments: First joint 0.2 0.1 0 -0.1 -0.2 ļ Friction (šššššššš Ģ ) at joints is modeled as a combination of stiction and viscous effects. 0.1 First joint Torque (N.m) − + Torque (N.m) + Joint kinematics: šš, ššĢ + Validation of the derived robot dynamic model and the friction model is completed using a closed-loop sinusoidal trajectory tracking experiment scenario under PD control. Torque (N.m) Device Description Desired feedback torque šššššššš Torque Controller Results and Discussion 0 2 4 6 8 10 12 time (sec) 14 16 18 20 14 16 18 20 Second joint Estimated torque Simplified friction model 0.1 Torque (N.m) Wrist Gimbal mechanism, and the implemented target matching game for it. Total external disturbances: šššš = šššššššš + šššššššš Torque (N.m) ļ Repetitive movement exercises help re-establish part of the lost motor function in stroke and spinal cord injury patients. ļ Rehabilitation robots present opportunities for lowering therapy costs, increasing patient motivation in therapy and enabling intensive exercise protocols. ļ This work focuses on development of force/torque estimation algorithms that will allow accurate tracking of patient interaction torques in rehabilitation robots, without relying on force sensors. Disturbance Observer Design Estimated torque, using NDO algorithm, follows the friction torque acting on the robot fairly closely. 0.05 0 -0.05 0 0 -1.5 -1 -0.5 0 0.5 1 1.5 2 dθ1/dt (rad/sec) Second joint frictional torque 0.1 0.05 0 -0.05 -0.1 -2 -1.5 -1 -0.5 0 0.5 dθ2/dt (rad/sec) 1 1.5 2 2 4 6 8 10 12 time (sec) Conclusion -0.1 -2 Torque (N.m) Introduction and Motivation ļ Robot dynamic model and friction model were developed. ļ Correct operation of the NDO, as a force estimator, was verified. ļ It was found that the pursued NDO structure has significant potential for enabling accurate interaction force/torque estimations for force feedback robots. 5th Annual Regional Meeting of the Rocky Mountain ASB April 17-18, 2015 | Estes Park, Colorado