Learning to Suppress Tremors:
A Deep Reinforcement Learning-Enabled Soft Exoskeleton for Parkinson's Patients

Tamás Endrei^1,2*, Sándor Földi^1,2, Ádám Makk³, and György Cserey^1,2*
1Faculty of Information Technology and Bionics, Pázmány Péter Catholic University, Budapest, Hungary
²Jedlik Innovation Ltd., Budapest, Hungary
³András Pető Faculty, Semmelweis University, Budapest, Hungary

Official implementation for the code used in:
Learning to Suppress Tremors: A Deep Reinforcement Learning-Enabled Soft Exoskeleton for Parkinson's Patients

---

Usage:

Agent:

The agent folder contains the modified version of the replay buffer to store multiple experiences collected throughout multiple reference movements. It also contains the TD7 agent with the detailed parameters present in the paper.

Environment:

The environment contains the folder of the reference movements which hold the 8 distinct recordings of 4 different dynamic movements.

It also contains the pybullet physical simulation environment used in the reinforcement learning environment. With the reinforcement learning environment and the upper body / exoskeleton urdf file.

Simulation:

The simulation folder contains the code used for the training and evaluation of the exoskeleton control's performance.

Trained Agents:

The trained agents' folder contains the actor, critic and encoder neural networks used for the evaluation of each tremor type.

Utilities:

The utilities' folder contains various scripts used in the physical simulations joint angle and end effector position calculations. As well as various scripts used for the evaluation and plotting of the results.

Evaluation logs:

The evaluation logs hold the logs used to determine the tremor suppressing effect of the exoskeleton.

Software

Results were originally collected with:

• Gym: 0.26.2
• Numpy: 1.23.5
• Pybullet: 3.2.5
• Python: 3.10
• Pytorch: 1.12.1
• SciPy: 1.11.4

📚 Citation

If you found this repository useful, please consider citing:

@ARTICLE{endrei2025learning,
  AUTHOR={Endrei, Tamás and Földi, Sándor and Makk, Ádám and Cserey, György},
  TITLE={Learning to suppress tremors: a deep reinforcement learning-enabled soft exoskeleton for Parkinson’s patients},
  JOURNAL={Frontiers in Robotics and AI},
  VOLUME={Volume 12 - 2025},
  YEAR={2025},
  URL={https://www.frontiersin.org/journals/robotics-and-ai/articles/10.3389/frobt.2025.1537470},
  DOI={10.3389/frobt.2025.1537470},
  ISSN={2296-9144},
}