Parallel robot for knee rehabilitation: Reduced order dynamic linear model, mechanical assembly and control system architecture

Abstract

In this work we present the development of a dynamic linear model of a 3UPS+1RPU parallel robot for knee rehabilitation, which allows the reduction of the error with respect to simulation carried out based on its non-linear model. Furthermore, the design and implementation of a control algorithm in a real robot is detailed, for which a dynamic linear model has been developed based on inertial parameters including a friction model in Coulomb and viscosity parameters. Subsequently, the linear model has reduced applying the numerical method of decomposition into singular values, resulting in a model expressed as function of base parameters. This method uses a base parameter identification path obtained by finite Fourier series. This path is optimized through minimization algorithms restricted by distance, velocity and acceleration of the linear actuators of the robot, as well as the working space of its spherical joints. Then, the compatibility level of the reduced dynamic model is quantified by estimating mean square error determined between the generalized forces of the independent joints obtained from the model and compared with those resulting from simulations performed in Adams/View software for a trajectory obtained by finite Fourier series. Afterwards, mechanical components involved in the implementation of the prototype are selected and the control system of its actuators is designed. Finally, tests are performed in a laboratory through photogrammetry equipment, in order to validate joints mobility in the robot and study its performance, for this task defined trajectories based on criteria of a physiotherapist are used.