A Complete robotic knee as a tool for a better understanding of joint dynamics

Abstract

The human knee is critical to human locomotion, performing both efficient walking and high load tasks. Failure of structural elements in this joint can cause pain or loss of function and often necessitates surgical interventions that do not always completely relieve symptoms. To address this, significant research effort has gone into improving understanding of the joint mechanics. Of specific interest is the coupling of the patellofemoral and tibiofemoral joint with ligaments and its impact on joint mechanics and movement. Challenges in producing dynamic loads and repeatable test conditions with cadaver, human and computer studies means that this coupling is not fully understood. This thesis describes a method to overcome these challenges using robotic modelling, an area that is relatively unexplored.

An anthropomorphic robotic joint that is a sagittal plane physical mechanical model of the human knee is built. The joint has rolling-sliding joint surfaces to represent condyles, elastic links to represent cruciate ligaments, a floating bearing to represent the patella and an antagonistic pair of actuators that drive the joint via cables to represent muscles and tendons. The system is tested in planar leg squat to examine dynamic interaction of joint elements. It is able to perform squats of 10°-110° flexion angle at up to 0.25 Hz, similar to rates achieved by humans. In the final model stiffnesses and geometries are derived from anthropomorphic data, however, the thesis describes design methodologies that can be applied to develop bicondylar joints with other design requirements including simplifications of the patella.

The results show a reduction in ACL tightness, or complete removal of the ACL spring force, has no effect on the quadriceps' mechanical advantage or the force required to perform squat despite changes to tibiofemoral contact point and ligament forces. This is because, when ACL spring force is reduced, the patellofemoral and tibiofemoral joints move together, changing the ratio of tensions in the quadriceps to patella tendons. Despite this, ACL loss compromises the stability of the knee and activities that require small flexion angles <15°, such as a healthy walking gait cannot be adequately controlled. The robotic work is supplemented by results from a quasi-static computer model. This shows that the ACL length and stiffness is the key factor controlling range of motion whereas the PCL provides control of friction losses. This has applications both for improving ligament surgeries, where osteoarthritis, thought to be linked to contact mechanics, causes joint pain, and in prosthetic knees, where joint properties are often tuned for individual amputees.

Additional investigation shows that increasing the amount of co-contraction in the antagonistic pair of actuators decreases payload capacity but allows the stiffness of the joint to be increased without changing the joint angle. This has applications for legged robots where compliance is critical for both facilitating efficient gait cycles and for limiting the damage from collisions. Also explored is the use of ligament stretch information for measuring joint state. Results shows that under physiological loads ligament length provides no benefit for estimating joint angle but may have applications for detecting user intent for application in prosthetic knee joints.

The work described in this thesis lends critical insights into knee joint mechanics, robotic knee investigation techniques and novel humanoid robotic joint design. These insights will have applications for both informing future joint surgeries and for addressing a need for improved robotic joints for legged robotic system such as prosthetic knees, walking robots and exoskeletons.