An Autonomous GIS Agent Framework for Geospatial Data Retrieval

Abstract: Powered by the emerging LLMs, autonomous geographic information systems (GIS) agents have the potential to accomplish spatial analyses and cartographic tasks. However, a research gap exists to support fully autonomous GIS agents: how to enable agents to discover and download the necessary data for geospatial analyses. This study proposes an autonomous GIS agent framework capable of retrieving required geospatial data by generating, executing, and debugging programs. The framework utilizes the LLM as the decision-maker, selects the appropriate data source (s) from a pre-defined source list, and fetches the data from the chosen source. Each data source has a handbook that records the metadata and technical details for data retrieval. The proposed framework is designed in a plug-and-play style to ensure flexibility and extensibility. Human users or autonomous data scrawlers can add new data sources by adding new handbooks. We developed a prototype agent based on the framework, released as a QGIS plugin (GeoData Retrieve Agent) and a Python program. Experiment results demonstrate its capability of retrieving data from various sources including OpenStreetMap, administrative boundaries and demographic data from the US Census Bureau, satellite basemaps from ESRI World Imagery, global digital elevation model (DEM) from OpenTopography.org, weather data from a commercial provider, the COVID-19 cases from the NYTimes GitHub. Our study is among the first attempts to develop an autonomous geospatial data retrieval agent.

• The paper introduces an autonomous GIS agent framework that leverages LLMs to automate and optimize geospatial data retrieval with a success rate of 80-90% from over 70 test requests.
• The framework employs iterative program generation, debugging, and a plug-and-play handbook system to seamlessly integrate diverse data sources like OpenStreetMap, US Census, and ESRI World Imagery.
• The study highlights future improvements, including multimodal data integration and Retrieval Augmented Generation techniques to enhance data understanding and selection in autonomous GIS operations.

Overview of An Autonomous GIS Agent Framework for Geospatial Data Retrieval

This paper presents a significant advancement in the field of Geographic Information Systems (GIS) through the development of an autonomous GIS agent framework, leveraging LLMs to facilitate geospatial data retrieval. The authors address the prominent challenge that impedes the complete automation of GIS agents: the ability to autonomously discover and fetch essential geospatial data for analyses and cartographic tasks.

Framework and Methodology

The proposed framework represents a comprehensive approach to automating geospatial data retrieval by generating, executing, and debugging programs. It utilizes LLMs to function as decision-makers, selecting the appropriate data sources from a predefined list indexed within a system of flexible handbooks. These handbooks contain metadata and technical details necessary for efficient data retrieval from various sources. The framework is designed with a plug-and-play architecture, enabling easy addition of new data sources by updating handbook metadata.

A prototype of this framework has been implemented as a QGIS plugin named GeoData Retrieve Agent and an interactive Python program. Through iterative testing and program generation, the framework has demonstrated the ability to successfully retrieve geospatial data from multiple sources, including OpenStreetMap, US Census Bureau datasets, ESRI World Imagery, OpenTopography, OpenWeather, and COVID-19 data from the NYTimes GitHub repository.

Results and Implications

The results from over 70 data retrieval requests showcased a success rate between 80% and 90%, highlighting the agent's capability to autonomously fetch diverse types of geospatial data. For instance, data from OpenStreetMap involves complex querying practices that the agent can automate through integrated queries and templates, effectively managing different data structures such as nodes, ways, and relations.

The framework's methodology aligns with the broader objectives of autonomous GIS, aiming to reduce human intervention and democratize access to GIS technology. By enabling autonomous agents to obtain data independently, researchers and practitioners can direct their focus towards complex analytical tasks and decision-making processes.

Limitations and Future Directions

Despite the promising potential of this framework, certain limitations persist, particularly concerning handbook constraints related to multi-modal data retrieval and the length restrictions imposed by LLMs. The authors suggest that further research is necessary to support more comprehensive and multimodal handbooks, which could include vector files or raster maps for remote sensing data. Additionally, integrating Retrieval Augmented Generation (RAG) approaches may enhance the framework's capability to handle extensive datasets and metadata.

Moreover, the paper signifies the need for data understanding and assessment modules within geospatial data retrieval agents. These modules would enable agents to evaluate and select between similar datasets, mimicking the decision-making acumen of human GIS analysts.

Conclusion

The autonomous GIS agent framework proposed in this paper embodies a forward-looking approach to automating the complex task of geospatial data retrieval. By leveraging LLMs and a structured framework of handbooks, the researchers illustrate an innovative method to facilitate a more autonomous workflow in GIS environments. Future developments should target overcoming current limitations and expanding the framework’s adaptability and functionality. Such advancements hold substantial promise for the future of autonomous GIS and its broader applications in geospatial analysis and research.