Start Date
5-2020 12:00 AM
Description
Past work has shown Model Predictive Control (MPC) to be an effective strategy for controlling continuum joint soft robots using rudimentary models, however the in-accuracies of these models often mean that an integration scheme must be combined with MPC. Presented in this work is a novel dynamic model formulation for continuum joint soft robots which is more accurate than the authors’ previous models yet remains fast enough for MPC. This model is based on the Piecewise Constant Curvature (PCC) assumption and a relatively new configuration representation and allows for computationally efficient simulation. Due to the difficulty in determining model parameters (damping and spring effects) as well as effects common in continuum joint soft robots (hysteresis, complex pressure dynamics, etc.), we submit that most model-based controllers of continuum joint soft robots would benefit from some form of model adaptation. In this work a novel form of adaptive MPC is presented based on Model Reference Adaptive Control (MRAC). We call our novel adaptive MPC approach MRPAC. We show that like MRAC, MRPAC is able to compensate for ”known unknowns” such as unknown inertias and spring and damper coefficients. Our experiments also show that like MPC, MRPAC is also robust to ”unknown unknowns” such as unmodeled external forces and any forces not represented in the form of the adaptive model. Experiments in simulation and hardware show that MRPAC outperforms MPC and MRAC in every case except for that in which MPC uses a perfect model in simulation.
Included in
Model Reference Predictive Adaptive Control for Large Scale Soft Robots
Past work has shown Model Predictive Control (MPC) to be an effective strategy for controlling continuum joint soft robots using rudimentary models, however the in-accuracies of these models often mean that an integration scheme must be combined with MPC. Presented in this work is a novel dynamic model formulation for continuum joint soft robots which is more accurate than the authors’ previous models yet remains fast enough for MPC. This model is based on the Piecewise Constant Curvature (PCC) assumption and a relatively new configuration representation and allows for computationally efficient simulation. Due to the difficulty in determining model parameters (damping and spring effects) as well as effects common in continuum joint soft robots (hysteresis, complex pressure dynamics, etc.), we submit that most model-based controllers of continuum joint soft robots would benefit from some form of model adaptation. In this work a novel form of adaptive MPC is presented based on Model Reference Adaptive Control (MRAC). We call our novel adaptive MPC approach MRPAC. We show that like MRAC, MRPAC is able to compensate for ”known unknowns” such as unknown inertias and spring and damper coefficients. Our experiments also show that like MPC, MRPAC is also robust to ”unknown unknowns” such as unmodeled external forces and any forces not represented in the form of the adaptive model. Experiments in simulation and hardware show that MRPAC outperforms MPC and MRAC in every case except for that in which MPC uses a perfect model in simulation.
Comments
Due to COVID-19, the Symposium was not able to be held this year. However, papers and posters were still submitted.