For effective control and interaction of active prosthetic and orthotic (P/O) devices with the human, understanding of human control of gait is needed. Feedback, provided by sensors and reflexes in the body, can compensate for unexpected environmental conditions or sensory noise. This was shown by a model of human gait purely based on reflexove feedback, presented by Geyer [1] and Song [4], referred to as neuromuscular model control (NMC). The NMC framework requires only basic supraspinal input of foot clearance height and foot placement location. Additionally, NMC uses local low-level muscle reflex signals (stretch, stretch rate, and force) to generate muscle activation. Eleven simulated Hilltype [3] muscles per leg generate force, resulting in a net torque around the joints. In this work, a first step is done towards adjusting the neuromuscular model for the use in P/O devices. Predefined trajectory-based impedance controllers are currently the golden standard of control of gait in P/O devices [5]. These controllers require switching to different trajectory sets when in different environments, like rough or sloped terrain. In simulation, NMC has shown to be able to generate stable human-like gait for different environments without adjusting parameters nor switching to different trajectories [1]. Safety and stability are an important part of controller design for P/O devices. A robust gait controller should at least not decrease, but preferably increase the stability against slipping, tripping or perturbations on the patient wearing the P/O device. A safe and robust device will minimize injury, allow walking in more challenging environments and increase the patient’s confidence in his/her own mobility. However, this part is often neglected in research. The goal of this study is to enhance NMC to be more robust and more subject specific. The robustness against perturbations of the NMC is investigated as a first step. The next step is to make NMC more subject specific by re-optimizing the model parameters using human kinematic and torque data. By driving the model using subject specific input data from previous research [6], instead of optimizing by evaluating the model’s forward dynamics, the model
[1]
Hartmut Geyer,et al.
A Muscle-Reflex Model That Encodes Principles of Legged Mechanics Produces Human Walking Dynamics and Muscle Activities
,
2010,
IEEE Transactions on Neural Systems and Rehabilitation Engineering.
[2]
H. van der Kooij,et al.
Center of mass velocity-based predictions in balance recovery following pelvis perturbations during human walking
,
2016,
Journal of Experimental Biology.
[3]
Seungmoon Song,et al.
A neural circuitry that emphasizes spinal feedback generates diverse behaviours of human locomotion
,
2015,
The Journal of physiology.
[4]
Robert Riener,et al.
Control strategies for active lower extremity prosthetics and orthotics: a review
,
2015,
Journal of NeuroEngineering and Rehabilitation.
[5]
Nikolaus Hansen,et al.
The CMA Evolution Strategy: A Comparing Review
,
2006,
Towards a New Evolutionary Computation.
[6]
O. Schmitt.
The heat of shortening and the dynamic constants of muscle
,
2017
.