Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Agentic AI in Remote Sensing: Foundations, Taxonomy, and Emerging Systems

Niloufar Alipour Talemi, Julia Boone, Fatemeh Afghah

TL;DR

This survey addresses the need for autonomous, agentic AI in Earth Observation to manage long-horizon geospatial workflows. It presents a taxonomy that separates single-agent copilots from multi-agent orchestrators and outlines foundational components (Foundations, Agents, Systems, Evaluation) connecting perception, reasoning, tool use, and memory to EO tasks. By reviewing RS foundation models, multimodal reasoning, and a broad ecosystem of datasets, benchmarks, and safety considerations, it highlights opportunities and gaps in grounding, memory, and orchestration. The work offers a strategic roadmap toward Earth-native, memory-enabled, and safety-conscious autonomous RS systems, with open platforms and standardized evaluation as critical enablers for real-world deployment and reproducibility.

Abstract

The paradigm of Earth Observation analysis is shifting from static deep learning models to autonomous agentic AI. Although recent vision foundation models and multimodal large language models advance representation learning, they often lack the sequential planning and active tool orchestration required for complex geospatial workflows. This survey presents the first comprehensive review of agentic AI in remote sensing. We introduce a unified taxonomy distinguishing between single-agent copilots and multi-agent systems while analyzing architectural foundations such as planning mechanisms, retrieval-augmented generation, and memory structures. Furthermore, we review emerging benchmarks that move the evaluation from pixel-level accuracy to trajectory-aware reasoning correctness. By critically examining limitations in grounding, safety, and orchestration, this work outlines a strategic roadmap for the development of robust, autonomous geospatial intelligence.

Agentic AI in Remote Sensing: Foundations, Taxonomy, and Emerging Systems

TL;DR

This survey addresses the need for autonomous, agentic AI in Earth Observation to manage long-horizon geospatial workflows. It presents a taxonomy that separates single-agent copilots from multi-agent orchestrators and outlines foundational components (Foundations, Agents, Systems, Evaluation) connecting perception, reasoning, tool use, and memory to EO tasks. By reviewing RS foundation models, multimodal reasoning, and a broad ecosystem of datasets, benchmarks, and safety considerations, it highlights opportunities and gaps in grounding, memory, and orchestration. The work offers a strategic roadmap toward Earth-native, memory-enabled, and safety-conscious autonomous RS systems, with open platforms and standardized evaluation as critical enablers for real-world deployment and reproducibility.

Abstract

The paradigm of Earth Observation analysis is shifting from static deep learning models to autonomous agentic AI. Although recent vision foundation models and multimodal large language models advance representation learning, they often lack the sequential planning and active tool orchestration required for complex geospatial workflows. This survey presents the first comprehensive review of agentic AI in remote sensing. We introduce a unified taxonomy distinguishing between single-agent copilots and multi-agent systems while analyzing architectural foundations such as planning mechanisms, retrieval-augmented generation, and memory structures. Furthermore, we review emerging benchmarks that move the evaluation from pixel-level accuracy to trajectory-aware reasoning correctness. By critically examining limitations in grounding, safety, and orchestration, this work outlines a strategic roadmap for the development of robust, autonomous geospatial intelligence.
Paper Structure (22 sections, 2 figures, 3 tables)

This paper contains 22 sections, 2 figures, 3 tables.

Table of Contents

  1. Introduction
  2. Background
  3. Sensors, Data, and Applications
  4. Vision and Multimodal Foundation Models
  5. Remote Sensing Foundation Models
  6. Vision Foundations in Remote Sensing
  7. Multi-Modal Foundations in Remote Sensing
  8. Taxonomy of AI Agents in Remote Sensing
  9. Agent Types in Remote Sensing
  10. Agentic AI Applications in Remote Sensing
  11. Systems, Technologies, and Platforms
  12. Knowledge Representation and RAG
  13. Tool Integration and API Orchestration
  14. Benchmarks and Evaluations
  15. Datasets and Benchmarks
  16. ...and 7 more sections

Figures (2)

  • Figure 1: Overview of the Agentic AI ecosystem in remote sensing. The proposed framework consists of four key components: 1) Foundations: Data acquisition and foundation models; 2) Agents: A classification of systems into single-agent copilots vs. multi-agent orchestrators; 3) Systems: The technological stack (RAG, Tools, Memory) empowering the agents; and 4) Evaluation: Benchmarks for assessing planning and reasoning capabilities. The figure also maps these components to specific Earth observation applications.
  • Figure 2: Benchmarks and datasets for agentic remote sensing AI. GeoRSMLLM zhang2025georsmllm includes referring-expression tasks, change detection, scene classification, and geo-localization; OmniGeo yuan2025omnigeo covers health geography, RS scene classification, urban perception, and geospatial semantics; ThinkGeo shabbir2025thinkgeo pairs RS patches with multi-tool reasoning; and RingMo-Agent hu2025ringmo supports multi-modal RS tasks such as relation reasoning.