This paper introduces PHILIA, a multi-robot agent architecture built around a robot gateway abstraction. It separates low-frequency, high-semantic agent reasoning from high-frequency robot execution, while exposing perception, navigation, speech, robot policies, monitoring, and cancellation through a shared capability interface. PHILIA preserves OpenClaw’s user-interface and tool ecosystem and was validated on Astribot S1 robots. Demonstrations include household organization, backpack packing, and lifting a garbage bag. The authors emphasize long-horizon memory, scene grounding, heterogeneous embodiment support, and human confirmation or adjustment during execution.
No heat snapshots are available in the last 24 hours.