The World as a Graph Database for Distributed Robotic Intelligence and Autonomous Optimization

Robotic systems working in shared physical environments repeatedly acquire information about objects, resources, and changing conditions. Exchanging that information may reduce redundant work, but it also creates costs and risks associated with communication, stale evidence, and conflicting actions. This Perspective proposes a federated world graph connecting specialized onboard capabilities with task-relevant shared knowledge through local gateways that may also provide charging. The graph represents partial observations and their provenance rather than a complete digital copy of reality. Robots retrieve bounded, authorized views while retaining local control and explicitly constrained behavior during disconnection. Building on shared robotic knowledge, dynamic scene graphs, cloud robotics, and task allocation, the paper argues for evaluating evidence freshness, resource ownership, and charging access together. A hypothetical tomato-monitoring scenario illustrates the distinction between measurements, inferences, assignments, and verified actions. The paper then proposes controlled comparisons against independent robots, message sharing, and centralized graph services, with total energy per verified successful task as a primary outcome. No implementation, experimental results, or demonstrated energy savings are claimed. The resulting agenda makes the broad vision of distributed intelligence and autonomous optimization testable through bounded deployments, explicit knowledge contracts, and system-level resource accounting.