Dynamic Shape Control of Soft Trunk Robots
Dynamic Shape Control of Soft Trunk Robots
This research introduces a method for controlling the dynamic shape of soft trunk robots using adiabatic spectral submanifolds (aSSMs) within a model predictive control (MPC) framework. The authors trained an aSSM model on data from a trimmed-helicoid soft robot and applied it to hardware, achieving high-accuracy closed-loop end-effector tracking and full-body shape regulation. Their approach outperformed linear baselines by a factor of 10 in dynamic shape tasks, successfully completing section-by-section shaping and balancing a tennis ball for over 25 seconds in human-in-the-loop demos, whereas linear methods failed. A model-free reinforcement learning policy trained for 12 hours on hardware could not complete the same balancing task. These results validate aSSM-based MPC as a practical solution for full-body shape control on real soft robotic hardware.
The study presents a model-based control approach using adiabatic spectral submanifolds (aSSMs) for full-body shape regulation on soft trunk robots, demonstrating superior performance over linear baselines in both tracking and balancing tasks on hardware.
Source: arXiv Robotics — research abstracts · Read original article ↗
Loading article text…
What the source reports
Publisher-reported claims, with original evidence. These results have not been independently verified by RoboSignal.
Reported numbers
training hours
12 hours
View original evidence
es fall short. A model-free reinforcement learning policy trained for 12 hours on hardware fails on the same balancing task. These results establish aSSM-based MPC as a practical route to full-body shape control on real soft hardware.
Open source S5balancing duration
>25 seconds
Reported trials
View original evidence
oid soft trunk robot from data and use this in a model predictive control (MPC) scheme to demonstrate a range of shape control tasks on hardware. On closed-loop dynamic shape regulation, our method outperforms the available linear baselines by a factor of 10. In human-in-the-loop interactive demos, our method completes section-by-section shaping and balances a tennis ball for over 25 seconds, while the linear baselin
Open source S4performance factor
10
View original evidence
oid soft trunk robot from data and use this in a model predictive control (MPC) scheme to demonstrate a range of shape control tasks on hardware. On closed-loop dynamic shape regulation, our method outperforms the available linear baselines by a factor of 10. In human-in-the-loop interactive demos, our method completes section-by-section shaping and balances a tennis ball for over 25 seconds, while the linear baselin
Open source S4
- paper: not_reported
Source excerpts and review record
Automatically extracted; no manual editorial approval recorded.
oid soft trunk robot from data and use this in a model predictive control (MPC) scheme to demonstrate a range of shape control tasks on hardware. On closed-loop dynamic shape regulation, our method outperforms the available linear baselines by a factor of 10. In human-in-the-loop interactive demos, our method completes section-by-section shaping and balances a tennis ball for over 25 seconds, while the linear baselin
Open source S4
es fall short. A model-free reinforcement learning policy trained for 12 hours on hardware fails on the same balancing task. These results establish aSSM-based MPC as a practical route to full-body shape control on real soft hardware.
Open source S5
Source：arXiv Robotics — research abstracts · arxiv.org