AEROS: A Single-Agent Operating Architecture with Embodied Capability Modules

2026-04-08

AI Summary

This work addresses the lack of a unified abstraction for integrating intelligence and capabilities in existing robotic systems, whose architectures often suffer from either excessive coupling or loosely coordinated modules. The paper proposes AEROS, a novel architecture that models a robot as a single persistent agent, encapsulating skills, models, and tools within hot-pluggable Embodied Capability Modules (ECMs). A strategy-decoupled runtime mechanism is introduced to jointly manage execution and safety. This design constitutes the first modular robot operating system centered around a single-agent paradigm. Evaluated on a Franka Panda manipulator, AEROS achieves 100% task success—significantly outperforming baselines (67–93%)—with zero false acceptance of invalid actions at the policy layer. Moreover, it maintains 100% success after ECM hot-swapping without re-tuning, demonstrating exceptional composability, system-wide safety consistency, and runtime extensibility.

Abstract

Robotic systems lack a principled abstraction for organizing intelligence, capabilities, and execution in a unified manner. Existing approaches either couple skills within monolithic architectures or decompose functionality into loosely coordinated modules or multiple agents, often without a coherent model of identity and control authority. We argue that a robot should be modeled as a single persistent intelligent subject whose capabilities are extended through installable packages. We formalize this view as AEROS (Agent Execution Runtime Operating System), in which each robot corresponds to one persistent agent and capabilities are provided through Embodied Capability Modules (ECMs). Each ECM encapsulates executable skills, models, and tools, while execution constraints and safety guarantees are enforced by a policy-separated runtime. This separation enables modular extensibility, composable capability execution, and consistent system-level safety. We evaluate a reference implementation in PyBullet simulation with a Franka Panda 7-DOF manipulator across eight experiments covering re-planning, failure recovery, policy enforcement, baseline comparison, cross-task generality, ECM hot-swapping, ablation, and failure boundary analysis. Over 100 randomized trials per condition, AEROS achieves 100% task success across three tasks versus baselines (BehaviorTree.CPP-style and ProgPrompt-style at 92--93%, flat pipeline at 67--73%), the policy layer blocks all invalid actions with zero false acceptances, runtime benefits generalize across tasks without task-specific tuning, and ECMs load at runtime with 100% post-swap success.

Problem

• robotic systems
• intelligence abstraction
• capability organization
• execution architecture
• control authority

Innovation

• AEROS
• Embodied Capability Modules
• modular robotics
• runtime safety enforcement
• single-agent architecture