Challenges in using compliant ligaments for position estimation within robotic joints

Abstract

The mechanical advantages of bio-inspired condylar robotic knee joints for use in prosthetics or rehabilitation has been argued extensively in literature. A common limitation of these designs is the difficulty of estimating joint angle and therefore accurately controlling the joint. Furthermore, the potential role of ligament-like structures in robotic knees is not very well established. In this work, we investigate the role of compliant stretch sensing ligaments and their integration into a condylar robotic knee. Simulations and experiments are executed out in order to establish whether measurement of stretch in these structures can be used to produce a new feedback controller for joint position. We report results from a computer model, as well as the design and construction of a robotic knee that show, for a chosen condyle shape, ligament stretch is a function of muscle force and joint velocity as well as joint angle. We have developed a genetic algorithm optimised controller incorporating ligament feedback that demonstrates improved performance for a desired joint angle in response to step inputs. The controller showed marginal improvement in response to a cyclic command signal and further investigation is required in order to use these measurements in robust control, nevertheless we believe these results demonstrate the that ligament-like structures have the potential to improve the performance of robotic knees for prosthetics and rehabilitation devices.