Geo-Agent: a framework for intelligent geographic information systems with natural language interaction

doi:10.13474/j.cnki.11-2246.2025.1019

Abstract

Abstract: Traditional geographic information systems (GIS)often encounter multiple challenges in the human-computer interaction process, such as cumbersome operation procedures and limited intelligence.With the rapid development of general artificial intelligence technology, new engines centered on generative AI are driving the geographic information industry to accelerate its evolution from digitalization to intelligence.Typical practices include innovative research such as Autonomous GIS, MapGPT, and LLM-Find.Existing studies have confirmed the huge potential of large language models (LLMs)in tasks such as GIS knowledge Q&A and map-making.However, current research still has the following limitations: on the one hand, the models lack the ability to autonomously understand geographic information data and perform complex spatial task analysis; on the other hand, they highly rely on the task parsing and code generation capabilities of the large models themselves.In addition, the API calling mode may lead to the risk of privacy and sensitive geographic data leakage.To address these challenges, this paper innovatively proposes a geographic information intelligent agent, Geo-Agent, based on an open-source architecture.This framework proposes a multi-level instruction parsing strategy based on spatial thinking chains and a data retrieval strategy oriented to graph structures, effectively solving the problems of geographic semantic understanding deviation and spatial logic disconnection.Experimental verification shows that Geo-Agent can understand, manage, and deeply analyze geographic information data, and can complete complex spatial analysis tasks through natural language interaction, providing an innovative path for realizing fully autonomous and intelligent next-generation geographic information systems.

Key words: intelligent agent, large language models, gartography, Geo-Agent

CLC Number:

P208

LIANG Hailei, WANG Yong, DU Kaixuan, ZHOU Weixiang. Geo-Agent: a framework for intelligent geographic information systems with natural language interaction[J]. Bulletin of Surveying and Mapping, 2025, 0(10): 114-118,126.

References

[1] [TOMLINSON R F. Geographical information systems,spatial data analysis and decision making in government[D]. London:University of London,1974. [2] 罗斌,刘文豪,吴进,等. 从地理信息系统到地理智能体[J]. 地球信息科学学报,2025,27(1): 83-99. [3] VASWANI A,SHAZEER N,PARMAR N,et al. Attention is all you need[C] //Proceedings of the 31st International Conference on Neural Information Processing Systems. [S. l. ]:ACM Press,2017. [4] ZHAO W X,ZHOU K,LI J,et al. A survey of large language models[EB/OL]. [2025-02-21]. http://10. 48550/arXiv. 2303. 18223v15. [5] BUBECK S,CHANDRASEKARAN V,ELDAN R,et al. Sparks of artificial general intelligence: early experiments with GPT-4[EB/OL]. [2025-02-21]. https://doi. org/10. 48550/arXiv. 2303. 12712. [6] MOONEY P,CUI Wencong,GUAN Boyuan,et al. Towards understanding the geospatial skills of ChatGPT: taking a geographic information systems (GIS)exam[C] //Proceedings of the 6th ACM SIGSPATIAL International Workshop on AI for Geographic Knowledge Discovery. Hamburg:ACM Press,2023: 85-94. [7] 艾廷华. 深度学习赋能地图制图的若干思考[J]. 测绘学报,2021,50(9): 1170-1182. [8]YANG H,YUE S,HE Y. Auto-GPT for online decision making: benchmarks and additional opinions[EB/OL]. [2025-02-21]. https://doi.org/10.48550/arXiv.2306.02224. [9]LI Zhenlong,NING Huan. Autonomous GIS: the next-generation AI-powered GIS[J]. International Journal of Digital Earth,2023,16(2): 4668-4686. [10] FENG Y,DING L,XIAO G. Geoqamap-geographic question answering with maps leveraging LLM and open knowledge base (short paper)[C]//Proceedings of the 12th International Conference on Geographic Information Science (GIScience 2023). Leeds:[s. n. ],2023. [11] ZHANG Yifan,HE Zhengting,LI Jingxuan,et al. MapGPT: an autonomous framework for mapping by integrating large language model and cartographic tools[J]. Cartography and Geographic Information Science,2024,51(6): 717-743. [12] NING H,LI Z,AKINBOYEWA T,et al. An autonomous GIS agent framework for geospatial data retrieval[EB/OL]. [2025-02-21]. https://arxiv.org/pdf/2407.21024. [13] BROWN T B,MANN B. Language models are few-shot learners[EB/OL]. [2025-02-21]. https://doi.org/10.48550/arXiv.2005.14165. [14] KOJIMA T,GU S S,REID M,et al. Large language models are zero-shot reasoners[J]. Advances in Neural Information Processing Systems,2022,35:22199-22213. [15] YAO S,ZHAO J,YU D,et al. React: synergizing reasoning and acting in language models[EB/OL]. [2025-02-21]. https://doi.org/10.48550/arXiv.2210.03629. [16] HUANG L,YU W,MA W,et al. A survey on hallucination in large language models: principles,taxonomy,challenges,and open questions[J]. ACM Transactions on Information Systems,2025,43(2): 1-55.