Dynamic Modeling Of Pneumatic Robots
Dynamic Modeling of flexible or "soft" robots is an active research area. The type of dynamic modeling of soft robots depends a lot on the type of soft joints one is trying to model. The robot arm, Kaa, shown to the right is composed of antagonistic inflatable bladders for joints and pseudo-rigid inflatable links. Because the joints on this robot arm act similarly to revolute joints, this arm can be modeled effectively with traditional rigid body dynamics where the pressures in the antagonistic bladders can be mapped to a torque applied on the joint.
In contrast to Kaa, the orange robot arm, Sir Hiss, shown to the left has flexible segments that are approximately the same size as the rigid links. The joints are modeled as constant curvature (CC) segments. This assumption is accurate in the absence of high external loads or large inertial effects. The CC assumption begins to break down as high external loads are included. Several different formulations exist that each have different pros and cons. One formulation involves the Euler-Lagrange method, but this method becomes intractable for more than one joint. Another formulation developed by the lab uses traditional recursive Newton-Euler methods to solve for the dynamics with slight variations.
The above models fail to model things such as hysteresis and plastic creep. To remedy some of these problems, we are also actively developing machine learning approaches. Machine learning models can be combined with other dynamics derivations to account for the nonlinear effects that we are neglecting in our models.
Johnson CC, Quackenbush T, Sorensen T, Wingate D and Killpack MD (2021) Using First Principles for Deep Learning and Model-Based Control of Soft Robots. Front. Robot. AI 8:654398. doi: 10.3389/frobt.2021.654398
Hyatt P, Johnson CC and Killpack MD (2020) Model Reference Predictive Adaptive Control for Large-Scale Soft Robots. Front. Robot. AI 7:558027. doi: 10.3389/frobt.2020.558027
M. T. Gillespie, C. M. Best, E. C. Townsend, D. Wingate and M. D. Killpack, "Learning nonlinear dynamic models of soft robots for model predictive control with neural networks," 2018 IEEE International Conference on Soft Robotics (RoboSoft), 2018, pp. 39-45, doi: 10.1109/ROBOSOFT.2018.8404894.
Hyatt P, Johnson CC and Killpack MD (2020) Model Reference Predictive Adaptive Control for Large-Scale Soft Robots. Front. Robot. AI 7:558027. doi: 10.3389/frobt.2020.558027
M. T. Gillespie, C. M. Best, E. C. Townsend, D. Wingate and M. D. Killpack, "Learning nonlinear dynamic models of soft robots for model predictive control with neural networks," 2018 IEEE International Conference on Soft Robotics (RoboSoft), 2018, pp. 39-45, doi: 10.1109/ROBOSOFT.2018.8404894.