🐝 BeeWalker v6 — Training Results
Training results for the BeeWalker bipedal robot, using PPO and LSTM-based reinforcement learning in MuJoCo.
GitHub: ThomasVuNguyen/makeshit
Training Runs
| Run | Type | Steps | Videos | Checkpoints |
|---|---|---|---|---|
| lstm_20260209_154855 | LSTM | 17.9M | 4 | 0 |
| lstm_20260209_004543 | LSTM | 29.6M | 296 | 59 |
| lstm_20260208_191752 | LSTM | 9.9M | 99 | 19 |
| lstm_20260208_190415 | LSTM | 200K | 2 | 0 |
| lstm_20260208_171335 | LSTM | 1.2M | 12 | 2 |
| lstm_20260207_013141 | LSTM | 24.9M | 2489 | 49 |
Robot
- Simulation: MuJoCo (BipedWalker custom env)
- Servos: 6× MG996R (hip, knee, ankle × 2)
- Controller: RP2040 (target deployment)
- Algorithm: RecurrentPPO (LSTM hidden_size=32)
How to Use
# Upload results to HuggingFace
python tools/upload_hf.py --latest
# Upload all runs
python tools/upload_hf.py
# Upload specific run
python tools/upload_hf.py --run lstm_20260207_013141