A Data-driven Gain Adaptation Mechanism for Flexible Usability in the UGent Knee Rig

Abstract

This paper presents a data-driven gain adaptation mechanism in order to make the usability of the UGent Knee Rig (UGKR) more flexible. The design of the UGKR is unique as it is characterized by a movable ankle joint contrary to traditional dynamic knee rigs which are used to investigate knee kinematics and surgical techniques. This feature allows the UGKR to perform both bicycle motions and squat motions while applying a quadriceps force and hamstring forces. The control of the UGKR is a model-based control strategy which requires time-consuming system identification for each new knee specimen. A data-driven gain adaptation mechanism is developed reducing the necessary time for identification creating a flexible usability. Two adaptation mechanisms are implemented and tested on both a mechanical hinge and a cadaver knee: a rectangular sequence and a polar sequence. The results from squat and bicycle tests indicate that rectangular sequence provides good performance for the rigid mechanical hinge, however, it fails for the elastic cadaver knee during full extension. It is shown that the polar sequence results in good performance for the cadaver knee compared to the rectangular mechanism.

Revisión de los sistemas de control aplicados a los simuladores de rodilla

“…En los modelos basados en la plataforma de Oxford ha sido utilizado LabVIEW para programar el código del controlador y la interfaz de usuario (5-8, 20, 22, 24-28). En (3)(4)(5)(6)(7)(8)15,17,18,(20)(21)(22)(23)(25)(26)(27)(28)(29)(30)(31) se han implementado controladores Proporcional-Integral-Derivativos (PID) o alguna de sus variantes, (P, PI o PD). De todos los simuladores existentes en la literatura, los más reportados son los de la Universidad de Kansas y la Universidad de Gante.…”

“…Los autores también mencionan que, con cada espécimen probado en su simulador, se requieren ganancias específicas, mientras que las estrategias de diseño probadas utilizan parámetros de sintonización que no cambian (31). Para solucionar este problema de ganancias, se propone un mecanismo de "adaptación" en el que se determinan las ganancias en un corto pre-experimento donde el tobillo sigue cierta trayectoria (30).…”