Modern cities generate vast streams of multimodal data, including citizen engagement inputs (e.g., 311 service requests, social media posts) and geospatial datasets (e.g., GIS maps, satellite imagery). However, existing urban planning models often treat these data sources in isolation, failing to capture the interdependencies necessary for holistic, participatory decision-making. This research introduces a multi-agent AI framework to integrate and reason over multimodal urban datasets, leveraging GeoAI and large language models (LLMs) to enhance participatory urban governance.

The system consists of specialized Text and Spatial Agents, which extract, process, and fuse citizen concerns with geospatial insights. A Coordinator Agent synthesizes these inputs, generating an urban knowledge graph that informs decision-makers. Additionally, a Participatory Planning Agent acts as an interactive assistant, summarizing urban issues, generating AI-driven recommendations, and facilitating transparent policymaking.