We present GeoFlow, a method that automatically generates agentic workflows for geospatial tasks. Unlike prior work that focuses on reasoning decomposition and leaves API selection implicit, our method provides each agent with detailed tool-calling objectives to guide geospatial API invocation at runtime. GeoFlow increases agentic success by 6.8% and reduces token usage by up to fourfold across major LLM families compared to state-of-the-art approaches.

Current research on robot navigation is focused on clear recognition of the map and optimal path planning. The human cleaning path is, however, not optimal regarding time but optimal to the cleaning purpose. We have analyzed in this paper the cleaning behaviors in home environments and understood the user's path planning behaviors through usage tests of various vacuuming robots. We discovered that the actual user cleans with methods unique to specific areas of the house rather than following an optimal cleaning path. We not only suggest a path planning method for the vacuuming robot by using a layered map, but also a cleaning area designating method reflecting each area's characteristics. Based on these, we have designed a vacuuming robot's actions.