SpatialLLM: From Multi-modality Data to Urban Spatial Intelligence

2025-05-19

AI Summary

Existing approaches for complex spatial intelligence tasks—such as urban planning, ecological assessment, and traffic management—rely heavily on specialized GIS tools or domain experts, hindering zero-shot, low-barrier intelligent analysis. Method: This paper proposes a prompt-only, zero-shot spatial understanding framework that requires no model fine-tuning, task-specific architecture, or manual annotation. It automatically parses heterogeneous geospatial data into structured scene descriptions and designs LLM prompting strategies explicitly tailored for spatial reasoning, enabling pretrained large language models to perform cross-domain spatial understanding and decision support. Contribution/Results: To our knowledge, this is the first end-to-end urban spatial intelligence framework operating without GIS expertise. It achieves state-of-the-art zero-shot performance across multiple tasks, empirically demonstrating the critical roles of contextual modeling, multi-source knowledge integration, and long-range spatial reasoning in urban intelligence.

Technology Category

• Knowledge Representation and Reasoning: Geometric, Spatial, and Temporal Reasoning
• Computer Vision: Remote Sensing / Geospatial AI
• Planning, Routing, and Scheduling: Optimization of Spatio-temporal Systems

Application Category

• Semantics and Knowledge: Data modeling to support human-machine intelligence, including LLMs agents, intelligent system behavior, explanations, and user-friendly interactions
• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs
• Search and Retrieval-Augmented AI: Web query analysis, representation and understanding

Abstract

We propose SpatialLLM, a novel approach advancing spatial intelligence tasks in complex urban scenes. Unlike previous methods requiring geographic analysis tools or domain expertise, SpatialLLM is a unified language model directly addressing various spatial intelligence tasks without any training, fine-tuning, or expert intervention. The core of SpatialLLM lies in constructing detailed and structured scene descriptions from raw spatial data to prompt pre-trained LLMs for scene-based analysis. Extensive experiments show that, with our designs, pretrained LLMs can accurately perceive spatial distribution information and enable zero-shot execution of advanced spatial intelligence tasks, including urban planning, ecological analysis, traffic management, etc. We argue that multi-field knowledge, context length, and reasoning ability are key factors influencing LLM performances in urban analysis. We hope that SpatialLLM will provide a novel viable perspective for urban intelligent analysis and management. The code and dataset are available at https://github.com/WHU-USI3DV/SpatialLLM.

Problem

• Advancing spatial intelligence in urban scenes without training
• Enabling zero-shot execution of urban planning tasks
• Improving LLM performance in multi-field urban analysis

Innovation

• Unified language model for urban spatial tasks
• Zero-shot execution without training or expertise
• Detailed scene descriptions prompt pre-trained LLMs