[home] [Personal Program] [Help]

---

Session: Musculoskeletal System
Session starts: Thursday 24 January, 10:40
Presentation starts: 11:10
Room: Lecture room 558

---

Stijn Van Eesbeek (TU Delft)
Erwin de Vlugt (TU Delft)
Frans van der Helm (TU Delft)
Michel Verhaegen (TU Delft)

Abstract:
By making use of haptic manipulators and system identification techniques parameters describing the neural and mechanical contributions to joint impedance can be obtained [1] These parameters are useful in rehabilitation practice for diagnostics and treatment selection [2]. Linear system theory has been used extensively for this purpose, but is no longer valid during functional tasks as parameters describing the neuromuscular system are known to vary with state, i.e. torque and angle, and task. Time-variant identification techniques have been developed which can potentially be used to identify human joint impedance through its complete range of motion out of a single observation [3]. This makes the method suitable for evaluation of joint impedance during functional tasks and appropriate for clinical application where short duration of the experiments is highly desired. Time-variant system descriptions are obtained with a LPV-subspace identification routine, where prescribed scheduling functions are used to give the model freedom to vary in time. Earlier studies under isometric conditions have shown a strong dependency of joint stiffness on exerted torque, using a single scheduling function based on torque. To have a valid system description of the joint undergoing large rotations the nonlinear elasticity of passive tissues has to be incorporated. The nonlinear passive stiffness can be incorporated in a LPV model by expanding the number of scheduling functions, in this study based on polynomials of the joint angle. Simulation studies have been used to show the validity of the proposed method. Subsequently, the method was applied on a batch of subjects. Torque perturbations were applied to the subject’s ankle using a haptic manipulator while the subject was moving through a large range of motion. For all subjects, the nonlinear passive elasticity could be retrieved out of a single observation using the proposed method. This opens many possibilities for identifying neural and non-neural components of joint impedance during tasks with large joint rotations.