Harnessing Embodied Agents: Runtime Governance for Policy-Constrained Execution
This research paper proposes a runtime governance framework for policy-constrained execution in embodied agent systems, separating agent cognition from execution oversight. It formalizes a control boundary among a persistent embodied agent, modular capability packages, and a runtime governance layer that handles capability admission, policy checking, execution monitoring, rollback, and human override. Evaluated across 1,000 randomized simulation trials, the framework effectively intercepts unauthorized actions and detects runtime violations while providing structured recovery.
Summary and content fields are drawn from the document text and are pending editorial review.
Who issued it
Who it governs
Researchers and developers building autonomous embodied AI systems who need structured runtime safety and governance mechanisms.
What the document contains
The document addresses the architectural and systemic challenges of governing embodied agents at runtime, formalizing a control boundary and execution pipeline that integrates admission control, policy checking, monitoring, rollback, and human override.
The text does not provide hardware-level deployment validation on physical robots, nor does it detail a universal natural-language policy specification language.
- Embodied agent execution requires runtime governance that separates agent cognition from execution oversight.
- A dedicated runtime governance layer can mediate capability admission, policy checks, and execution monitoring without altering the underlying agent model.
- Externalizing governance enables environment-sensitive adaptation across simulation, real robots, and human-shared spaces.
- Continuous runtime monitoring and structured recovery significantly reduce unsafe continuations compared to pre-execution-only filters.
· 1 Introduction · 2 Related Work · 3 Method · 4 Experiments · 5 Discussion · 6 Conclusion
How it scores
Each score is backed by a quote verified against the document text. Editorial review has not yet confirmed them.
Fairness
0/5Non-discrimination and equitable treatment
screening score · 2026-10-23
Human Dignity
0/5Respect for the inherent worth of persons
screening score · 2026-10-23
Human Oversight
4/5Humans remain in control of consequential decisions
Governance is externalized into a dedicated runtime layer performing policy checking, capability admission, execution monitoring, rollback, and human override.
Abstract
screening score · 2026-10-23
Privacy
0/5Protection of personal data and private life
screening score · 2026-10-23
Transparency
4/5Openness about how AI systems work and are used
The governance layer operates on an explicit, inspectable policy knowledge base comprising admission rules, constraint predicates, environment profiles, and recovery strategies
Section 1
screening score · 2026-10-23
Impact Assessment
4/5Required assessment of an AI system's risks/impacts before or during use
every capability invocation passes through admission control, policy checking, and execution monitoring, with recovery and human override at runtime
Section 1
screening score · 2026-10-23