Views
No views yet
Joint-Space Empowerment for Dexterous Coordination in Tendon-Driven Hands
James Heald, Vittorio Caggiano, Vikash Kumar, Maneesh Sahani
ICML 2026 (Spotlight)
Paper | Code | Project Page
1joint-space-empowerment
2├── Adroit
3│ ├── BaodingBalls
4│ │ ├── CustomAdroitBaodingP1-v1.yaml # Environment config
5│ │ ├── seed_0
6│ │ │ ├── rl_model_10000000_steps.zip # Trained model
7│ │ │ └── rl_model_vecnormalize_10000000_steps.pkl # Observation normalization statistics
8│ │ ├── seed_1 … seed_4
9│ ├── DieReorient
10│ │ └── ...
11│ ├── KeyTurn
12│ │ └── ...
13│ └── PenTwirl
14│ └── ...
15├── MyoHand
16│ ├── BaodingBalls
17│ │ └── ...
18│ ├── DieReorient
19│ │ └── ...
20│ ├── KeyTurn
21│ │ └── ...
22│ ├── PenTwirl
23│ │ └── ...
24│ ├── Reorient100
25│ │ ├── Pretraining_Play
26│ │ │ ├── seed_0
27│ │ │ │ ├── rl_model_replay_buffer_1000000_steps.pkl # Replay buffer
28│ │ │ │ └── rl_model_vecnormalize_1000000_steps.pkl
29│ │ │ ├── seed_1 … seed_4
30│ │ └── Training
31│ │ └── ...
32│ └── Reorient8-sparse
33│ └── ...
34└── README.mdPrerequisites: clone and install the JoSE repository.
play.py script in the JoSE repository downloads a pretrained model and executes it in the corresponding environment, allowing agent behaviour to be visualized. The script is called from the repo root as follows:uv run python play.py --hand <hand> --task <task> --seed <seed>| Argument | Choices |
|---|---|
--hand | Adroit, MyoHand |
--task | BaodingBalls, DieReorient, KeyTurn, PenTwirl, Reorient8-sparse, Reorient100/Training |
--seed | 0, 1, 2, 3, 4, 5 |