AI RESEARCH

Embodied Foundation Models at the Edge: A Survey of Deployment Constraints and Mitigation Strategies

arXiv CS.AI

ArXi:2603.16952v2 Announce Type: replace-cross Deploying foundation models in embodied edge systems is fundamentally a systems problem, not just a problem of model compression. Real-time control must operate within strict size, weight, and power constraints, where memory traffic, compute latency, timing variability, and safety margins interact directly. The Deployment Gauntlet organizes these constraints into eight coupled barriers that determine whether embodied foundation models can run reliably in practice.