AI

OpenRUA: Robot-Use Agents Are Zero-Shot Visuomotor Policies

Read full article at source ↗

Based on a feed excerpt; full article content has not been acquired. Source evidence retrieved 06/10/2026, 09:18:25 UTC.

Briefings

OpenRUA is a zero-abstraction harness that enables off-the-shelf coding agents to operate as visuomotor policies on robots using only terminal access to ROS 2. Powered by Claude Opus 5, the system achieves 99.0% success on CaP-Bench and 87.0% on LIBERO-PRO without bespoke primitives or training. Key findings include spontaneous perception processing in 96.80% of episodes and manipulation client building in 95.87% of episodes, demonstrating emergent behaviors from a minimalist design.

Key points

• OpenRUA provides a zero-abstraction harness allowing coding agents to access robots via terminal ROS 2 interfaces without custom orchestration layers.
• The system achieves 99.0% success rates on CaP-Bench and 87.0% on LIBERO-PRO using Claude Opus 5 to generate zero-shot visuomotor policies.

Context

This research addresses whether complex custom harnesses are necessary for agents to interact with physical robots. OpenRUA simplifies this by offering only terminal access to ROS 2, enabling agents to organize their own workflows and perceive/manipulate through file I/O and coding.

Limitations

Evidence is limited to two benchmark datasets (CaP-Bench and LIBERO-PRO) with no additional robot platforms or real-world deployment data. Results are based on 99.0% and 87.0% success rates from a single preprint snapshot, with no long-term stability or safety validation provided.

Generated by Bittrees-hosted qwen35-2b sha256:aaf42c8b7c3c on 19/10/2026, 15:48:57 UTC. Based on publisher-provided feed evidence; may not cover the full original article. Check the source for details.

View archived briefing on IPFS
bafybeifh2k2s7noru4eil7lkhnbw5zhsajokcb44mzq7fr75di5hgswjxa