MiniCPM-Robot Enables Fully Local Robot Inference: Offline Tracking at 33Hz
HeyToha · x · 2026-07-25
The author points out that most current robot demos are fragile, relying heavily on stable Wi-Fi and cloud compute, which causes them to fail under weak networks or long tasks. The newly released MiniCPM-Robot series addresses this by combining contextual memory with fully local tracking.
The release includes three core components:
- MiniCPM-RobotTrack: Achieves fully offline target tracking on the Unitree Go2 Edu robot dog with 180ms end-to-end latency (5+ Hz). It continues making decisions even when physically disconnected from the internet.
- MiniCPM-RobotManip: A 1.5B VLA (Vision-Language-Action) model that utilizes historical visual context during multi-step tasks (e.g., remembering button presses, moved objects, or resuming after occlusion) without exploding inference costs.
- PhyAI Open-Source Framework: An inference infrastructure for Physical AI that uses CUDA Graph optimization and custom Triton kernels to boost inference rates from 10 Hz to 33 Hz or even 36 Hz.
Additionally, the model uses visual-token compression to cut visual processing computation by about 50% while remaining competitive with Qwen-VLA and π0.5, effectively reducing hardware costs and power draw.
Related event: OpenBMB Open-Sources MiniCPM-Robot Embodied AI Series(10 posts)→
More from Embodied
- Mondo’s Beni pairs cute hardware with millions of simulated RL trials — AnandSwa · 2026-07-25
- Almond Robotics Sells Out First Batch, Starts Production on V2 — chris_j_paxton · 2026-07-25
- Robotics research is mostly limited by robot hours or GPU hours, says one researcher — chris_j_paxton · 2026-07-25
- A cotton-picking robot with 108 arms is the latest robotics spectacle — chris_j_paxton · 2026-07-25
- From Bricked to MCP: Developer Turns Dead IoT Cameras into AI Servers — ahstanin · 2026-07-25
- EngineAI says its T800 humanoid is rolling out from a Henan factory — CyberRobooo · 2026-07-25