Reproduce “AEROS: A Single-Agent Operating Architecture with Embodied Capability Modules” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Xue Qin, Simin Luan, John See, Cong Yang, Zhijun Li
AEROS is a robotic operating architecture that models a robot as a single persistent intelligent agent whose capabilities are extended by installable Embodied Capability Modules (ECMs). This architecture achieved 100% task success rates in simulation under high failure conditions, outperforming baseline approaches by enabling dynamic re-planning and robust recovery mechanisms while enforcing safety through a policy-separated runtime.
Key insights
- Modeling a robot as a single, persistent intelligent agent with unified decision authority provides coherence and simplifies reasoning about complex system behavior.
- Decoupling capability logic from execution policies allows for dynamic extension of robot functionality through installable modules while maintaining robust system-level safety.
- Combining dynamic re-planning and explicit recovery actions within a closed-loop architecture significantly enhances task success and resilience against stochastic failures.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.