Institute of Engineering and Computational Mechanics
Development of energy-saving transfemoral prostheses
The use of models of the musculoskeletal system as a tool to support the development of lower limb assistive devices is becoming an established approach. Although some models developed provide a good insight in many aspects of the design of lower limb assistive devices like prosthesis, they present important limitations. Specially the metabolic energy expended during the gait, which is an important criteria to evaluate the performance of, for example, lower limb prosthesis, must be better accounted for.
A better estimation of the variations in the metabolic cost and in the kinematics of the body during the mechanically disturbed gait depends on the proper modeling of the human musculoskeletal system. Recent research in the field of human musculoskeletal modeling achieved good results, but at the expense of a high degree of complexity and computational effort. These drawbacks make the use of musculoskeletal modeling to design lower limb assistive devices challenging. On the one hand, the models should be complex enough to predict variations in the metabolic cost and kinematics of the gait satisfactorily. On the other hand, the development of efficient strategies and the use of simplified models are required in order to keep the problem computationally tractable and ensure the straightforwardness of the analysis.
In this context, the aim of this project is to develop straightforward approaches to model and analyze mechanically disturbed gaits using musculoskeletal models that can be used in the future to design lower limb prosthetic devices. The project focused up to date on the following three topics:
Realization of experiments on mechanically disturbed walking and analysis of the required torques at the joints obtained by conventional inverse dynamics using 3-D skeletal models of the subjects.
Development of an approach to solve the muscle force distribution problem in biomechanics that considers the muscle activation and contraction dynamics, based on a large-scale optimization that allows for the use of time-integral cost functions, e.g. the total metabolic cost. This approach permits the estimation of metabolic cost for the case of known or measured kinematics by employing phenomenological muscle energy expenditure models recently proposed in the literature.
Development of an approach to predict walking kinematics, kinetics, muscle force, and neural excitation patters, based on the parameterization of the kinematics and of the muscle forces. The approach requires the solution of a large-scale optimization problem that minimizes the metabolic cost per unit of distance traveled. We believe this approach is computationally more efficient than the extremely expensive dynamic optimization approach, because it does not require any forward integration of the differential equations describing the dynamics of the musculoskeletal system.