rocPAI-Forge

When a Robot Arm Invents Its Own Grip: An RL Practice with OpenArm

Mon, 06 Jul 2026 00:00:00 +0000

总览回放 / Overview replay

📖 This is the concise version (~3 min). For the full engineering details (design decisions, algorithm / reward, diagnostics, reproduce commands), read the deep-dive →

UniLab & Joint Release

UniLab is a heterogeneous robot-RL training infrastructure: CPU-parallel physics simulation (MuJoCo / Motrix) and GPU policy learning are coupled through a unified runtime and shared memory — instead of pinning physics, rollout collection, and learning on a single GPU-resident simulation path. Tasks, rewards, and backend selection are expressed as Hydra owner YAMLs; training goes through a unified uv run train / uv run eval CLI covering PPO, SAC, TD3, APPO, and more.

Overview

Wed, 01 Jul 2026 00:00:00 +0000

rocPAI-Forge is dedicated to forging Physical AI on AMD ROCm. We organize our work as a closed simulation ↔ real-world loop, with every stage running on the ROCm acceleration stack.

Solution Architecture

flowchart TB
subgraph REAL["Real World"]
RR["Real robots / arms / mobile platforms"]
SEN["Sensor data capture"]
end
subgraph SIM["Simulation"]
ENG["Sim engines
MuJoCo / Genesis, etc."]
TWIN["Digital twins / scenes"]
end
subgraph LEARN["Learning"]
RL["Reinforcement Learning
locomotion / manipulation"]
WM["World Models"]
VLA["VLA models
Vision-Language-Action"]
end
ASSET["3D assets & scene reconstruction
meshes / environments / twins"]
SEN -- "Real2Sim: reconstruct dynamics/scenes" --> ASSET
ASSET --> TWIN
TWIN --> ENG
ENG --> RL
ENG --> WM
ENG --> VLA
RL --> POLICY["Policies / models"]
WM --> POLICY
VLA --> POLICY
ENG -. "Sim2Sim: cross-simulator validation/transfer" .-> ENG
POLICY -- "Sim2Real: deploy" --> INF["Real-robot inference
low-latency ROCm deploy"]
INF --> RR
RR --> SEN
ROCM["AMD ROCm acceleration (train / sim / inference)"]
ROCM --- SIM
ROCM --- LEARN
ROCM --- INF

Focus Areas

Area	What We Explore
Sim2Real	Closing the gap between simulation and real robots — domain randomization, calibration, and deployment
Sim2Sim	Cross-simulator validation and transfer (e.g. MuJoCo ↔ Genesis, etc.) for robust policies
Real2Sim	Reconstructing scenes and dynamics from real data to improve simulation fidelity
3D Assets & Scene Reconstruction	Meshes, environments, and digital twins for robotics and RL
Reinforcement Learning	Locomotion, manipulation, and task-specific RL on ROCm-accelerated stacks
World Models	Predictive models of environment dynamics for planning and control
VLA Models	Vision–Language–Action models for generalist robot policies
Real Robot Inference	Low-latency deployment on manipulators and mobile platforms with ROCm

Principles

Open by default — code, configs, and learnings shared with the community
ROCm-first — optimize and validate on AMD hardware and software stack
End-to-end — from data and sim to train, eval, and real-world inference
Evidence over hype — reproducible benchmarks, clear contracts, and honest trade-offs

Roadmap

Wed, 01 Jul 2026 00:00:00 +0000

🚧 Under construction — stay tuned.

The macro roadmap will be published here once finalized.

Feeding the VLA: Generating Expert Grasp Trajectories for OpenArm on AMD ROCm

Tue, 30 Jun 2026 00:00:00 +0000

📖 This is the concise version (~3 min). For the full engineering details (design decisions, algorithm / reward, diagnostics, reproduce commands), read the deep-dive →

Overview

VLA (Vision-Language-Action) models are data-hungry, and real-robot collection is slow and expensive. So we flip it: on an AMD Instinct MI300X + ROCm box, we turn OpenArm’s pick-and-place into an in-sim expert trajectory data engine with openarm_mp_labs — given an object and a grasp pose, auto-solve a smooth, physically feasible, sub-mm-accurate demonstration trajectory that’s reproducible at scale.