Lerobot_MLP-SFP
Collection
End-to-End trained model and dataset for push-T task based on the Lerobot framework. Optimized with MLP-SFP.
•
8 items
•
Updated
•
1
FMLP-Policy: Flow Matching MLP for Robotic Control
This project explores RRAM-compatible neural network architectures for robotic manipulation policies, replacing UNet with pure MLP (Linear + ReLU only) for deployment on analog RRAM accelerators.
Overview
Diffusion Policy achieves SOTA on robotic manipulation but requires 50-100 denoising steps — impractical for RRAM deployment (each step needs ADC/DAC conversion). We explore:
- Streaming Flow Policy (SFP) reduces to 1-4 integration steps
- MLP velocity networks replaces UNet with RRAM-friendly architecture
- Quantization + noise tolerance validates INT8 deployment with device variation
Models
| Model | Architecture | Description |
|---|---|---|
| pusht_diffusion_v3 | ResNet18 + UNet | DP baseline, 136 episodes |
| pusht_diffusion_v4 | ResNet18 + UNet | DP baseline, 226 episodes |
| pusht_diffusion_v5 | ResNet18 + UNet | DP baseline, 255 episodes (best) |
| pusht_sfp_v9 | ResNet18 + UNet | SFP working baseline |
| pusht_sfp_v14 | ResNet18 + UNet | SFP with h50/k2/σ1 params |
| pusht_sfp_v15 | ResNet18 + MLP | SFP with cond_residual MLP (RRAM-compatible) |
Dataset
| Dataset | Episodes | Description |
|---|---|---|
| pusht_real_merged | 255 | Real robot Push-T task, SO-101 arm, 320x240 |
Key Results
Sim (2D Push-T):
- MLP achieves 0.86-0.88 FP32 vs UNet 0.74
- INT8 quantization: Bottleneck128+Skip achieves 0.86
- Noise tolerance: <6% accuracy drop at 10% multiplicative noise
Real Robot:
- DP v5: >90% success rate
- SFP v9: >70% success rate
- SFP v14/v15: > 80% success rate
Code
Hardware
- Robot: SO-101 (LeRobot compatible)
- Camera: HB Camera, top-down, 320x240 @ 30fps
- Training: 2x RTX 4090
Citation
Coming soon.
Acknowledgments
- LeRobot framework
- Streaming Flow Policy paper
- HKU EEE
Part of FYP project at The University of Hong Kong, supervised by Prof. Han Wang.