Stochastic Optimization of Impedance Parameters for a Powered Prosthesis Using a 3D Simulation Environment

Developing controllers for powered prostheses is a daunting task that requires involvement from clinicians, patients and robotics experts. Difficulties arise for tuning prosthetic devices that perform in multiple conditions and provide more functionality to the user. For this reason, we propose the implementation of a simulation framework based on the open-source 3D simulation environment Gazebo, to reduce the burden of experimentation and aid in the early stages of development. In this study, we present a minimalist plugin for the simulator that allows the interfacing of a virtual model with the native prosthesis controller and renders the finding of impedance parameters as a pattern search problem. To demonstrate the functionality of this approach, we used the framework to obtain the parameters that offer the most similar joint trajectory to the respective biological counterpart during swing phase for ground level walking. The optimization results are compared against the response of a physical 2DOF knee-ankle prosthesis operating under the optimized parameters, showing congruence to our model-based results. We found that a simulation-based solution is capable of finding parameters that create an emerging behavior that approximates to the kinematic trajectory goals within a tolerance (mean absolute error ∼10%). This provides an appropriate method for development and evaluation of impedance-based controllers before deployment to the physical device.