Modeling human walk by PCPG for lower limb neuroprosthesis control

In this paper, we propose an original and biologically-inspired leg prosthesis control system. We demonstrate that human walk periodic patterns can be modeled by a Programmable Central Pattern Generator (PCPG) algorithm. Assuming that high-level commands reflecting the user's intention (such as accelerate, decelerate or stop) - and optionally, with their associated confidence level - are available, we show that the PCPG can generate an output signal directly exploitable to control the prosthesis actuators at different desired walking speeds in a smooth way. Thanks to an adequate tuning of the PCPG parameters relying on realistic human walk kinematics, such a prosthesis would undoubtedly increase the comfort of the patient. In this study, we modeled the kinematics of foot angle of elevation of seven subjects walking on a treadmill at 10 different speeds. The method we used to modify at best the PCPG parameters is presented. We found that a low-level order polynomial interpolation of the PCPG parameters as a function of speed provides good similarity indices between real walk and generated patterns at different speeds. This proves the relevancy of our approach and paves the way for numerous applications of human walk rehabilitation. Additionally, results suggest that walk would be advantageously modeled by two PCPGs.