Recursive Self-Improvement of Exoskeleton Assistance Policies through Human-in-the-Loop Feedback
Recursive Self-Improvement of Exoskeleton Assistance Policies through Human-in-the-Loop Feedback
This research proposes a framework for the autonomous evolution of exoskeleton assistance policies using human feedback. An intelligent agent generates executable control code, evaluates it with human participants incorporating subjective feedback, movement performance, and muscle activation data, and autonomously revises the policy based on identified deficiencies. The system establishes a closed loop of generation, feedback, evaluation, and refinement. Evaluated using GPT, DeepSeek, GLM, and Kimi, all models achieved effective assistance after a single refinement iteration. Across three tasks, biceps and triceps muscle activation levels were reduced by 29.4% and 35.2%, respectively, compared to measurements without an exoskeleton. However, the study observed limited adaptability to unexpected situations, indicating a need for further improvements in robustness and safety.
The study demonstrates a closed-loop framework where an intelligent agent autonomously refines exoskeleton assistance policies using human feedback, achieving reduced muscle activation across three tasks with four different large language models after one refinement iteration, though the authors note limitations in adaptability to unexpected situations.
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.
- Control
- Reported autonomousOpen source S2
Reported numbers
number of tasks tested
3
View original evidence
istance performance and refinement outcomes are compared across models. Despite differences in initial performance, effective assistance is achieved with all four models after a single autonomous refinement iteration under the tested conditions. Across three tasks, biceps and triceps muscle activation levels are reduced by 29.4% and 35.2%, respectively, compared with those measured without an exoskeleton. However, li
Open source S4number of tasks across which assistance is achieved
3
View original evidence
istance performance and refinement outcomes are compared across models. Despite differences in initial performance, effective assistance is achieved with all four models after a single autonomous refinement iteration under the tested conditions. Across three tasks, biceps and triceps muscle activation levels are reduced by 29.4% and 35.2%, respectively, compared with those measured without an exoskeleton. However, li
Open source S4
- paper: not_reported
Source excerpts and review record
Automatically extracted; no manual editorial approval recorded.
arXiv:2611.01581v1 Announce Type: new Abstract: To reduce reliance on manual design and extensive parameter tuning, a framework for the autonomous evolution of exoskeleton assistance policies through human feedback is proposed. Assistance policies are autonomously generated as executable control code by an intelligent agent and evaluated in experiments with human participants. Subjective user feedback, movement perfo
Open source S2
istance performance and refinement outcomes are compared across models. Despite differences in initial performance, effective assistance is achieved with all four models after a single autonomous refinement iteration under the tested conditions. Across three tasks, biceps and triceps muscle activation levels are reduced by 29.4% and 35.2%, respectively, compared with those measured without an exoskeleton. However, li
Open source S4
rmance, and muscle activation measurements are incorporated into policy assessment. Based on these assessments, policy deficiencies are autonomously identified and the control code is revised by the agent for subsequent trials. A closed loop of policy generation, human feedback, evaluation, and autonomous refinement is thereby established. The framework is evaluated using GPT, DeepSeek, GLM, and Kimi, and initial ass
Open source S3
Source：arXiv Robotics — research abstracts · arxiv.org