Towards Spatio-Temporal Intelligence Systems in Urban Space: A Twin-Agent Perspective
Authors: Zhen Dong, Haiping Wang, Wei Tu, Yuan Liu, Bing Wang, Xuan Sun, Jinzhou Cao, Ronald Clark, Zhiheng Fu, Mohammed Bennamoun, Danfeng Hong, Yuanzheng Cui, Gemine Vivone, Pedram Ghamisi, Bisheng Yang, Qingquan Li, Deren Li
Organizations: Wuhan University · The Hong Kong University of Science and Technology · Shenzhen University · The Hong Kong Polytechnic University · Nankai University · Shenzhen Technology University · University of Oxford · The University of Western Australia · Southeast University · Chinese Academy of Sciences · National Research Council of Italy · University of Iceland
Cities must coordinate domains like infrastructure, mobility, environmental management, emergency response, and public services as conditions evolve. An urban Spatio-Temporal Intelligence (STI) system helps the city monitor citywide states, analyze problems, plan and execute interventions, and adapt through feedback, making urban governance more timely, effective, and transparent. Recent urban twinning and agentic AI techniques provide foundations for such systems. From a twin-agent perspective, this survey takes a system-level view of how these disparate techniques can be coherently integrated into an urban STI system. Specifically, we review how the twin reconstructs and represents a persistent, information-enriched urban environment, and how the agent understands the twin via interfaces, reasons over it, acts reliably, and continuously evolves with the city. Built upon this synthesis, we summarize a two-loop blueprint, where an inner agent loop for grounding, reasoning, and verification, and an outer system loop linking data, twin, agentic action, feedback, and update. We then use this to structure an STI research agenda and show how a shared prototype supports diverse urban domains.
Figures & tables
Figure 1. Scope of this review. With the twin-agent perspective, we review the technologies for constructing an urban twin, developing grounded urban agent, and integrating both components into a future STI system for urban domains.
Reviews
Twin
Agent
System loop
Concept
Techniques
Concept
Techniques
Concept
Techniques
Twin Reviews ( Lehtola et al., 2022 ; Xu et al., 2024b ; Jeddoub et al., 2025 ; Sajadieh and Noh, 2025 ; Cao et al., 2025a ; Kong et al., 2025c )
✓
✓
–
–
–
–
Agent Reviews ( Dorobantu and Badea, 2026 ; Han et al., 2025 ; Hashemi and Züfle, 2026 ; Liang et al., 2025d ; Zhang et al., 2024b ; Zhang et al., 2026a ; Zhang et al., 2025e ; Shao et al., 2025 )
–
–
✓
✓
–
–
System Reviews ( Xu et al., 2026 ; Ye et al., 2026b )
✓
–
✓
–
✓
–
This review
✓
✓
✓
✓
✓
✓
Table 1. Scope and technical depth across related reviews. “Concept” indicates an explicit framing, whereas “Techniques” indicates comprehensive coverage of its key subtopics, technical approaches, and implementation details. “ ✓ ” marks comprehensive coverage at the indicated level; “–” marks the absence of such coverage.
Figure 2. Detailed technical organization of this review. Urban twinning reconstructs, completes, represents, and enriches city states; urban agents access these states through textual and feature interfaces, reason over them, and operate through governed agentic architectures; the future vision couples both subsystems through a two-loop blueprint and adapts it across urban domains. Leaf labels summarize closely related technical topics.
Figure 3. The four-stage urban twinning framework. Recovery and completion determine the physical state contained in the twin, while physical representation and information embedding organize and enrich that state for subsequent operation.
Figure 4. Recovering observable urban states. Analytical and feed-forward paradigms process images, 3D data, or multimodal observations to recover static or dynamic scenes.
Figure 5. Inferring unobserved scene states. Generation synthesizes new 3D scenes or 4D evolution from weak conditions, whereas completion extends partial scene observations while preserving consistency with the measured spatial and temporal evidence.
Figure 6. Three forms of physical representation. Explicit representations enumerate directly addressable primitives, implicit representations encode continuous scene functions, and hybrid representations attach learned features or attributes to explicit spatial anchors.
Figure 7. Strategies for embedding information in an urban twin. Inside-out methods lift or learn semantic features from reconstruction signals, whereas outside-in methods ground external fields, attributes, and relations on twin carriers.
Figure 8. The four-layer organization of urban agent. General language models are grounded in city evidence, strengthened with spatio-temporal reasoning, embedded in agentic architectures, and constrained for reliable operations.
Figure 9. Urban agent systems organized by task. Analysis and planning transform urban evidence into analytical results and plans; interaction and control apply decisions to environments; mobility prediction and urban simulation model individual futures and collective dynamics.
Figure 10. The proposed system blueprint. The inner loop grounds, reasons, and acts, then replans before action. The outer loop updates the twin, executes an approved decision, observes the outcome, and starts the next full cycle.
Urban environments are shaped by design choices with long-term implications for health, safety, and quality of life, yet evaluating proposed interventions remains costly, time-consuming, and often impractical. Existing geospatial vision methods largely focus on monitoring urban indicators from aerial and street-view imagery, rather than proposing interventions and estimating their effects on such indicators. Moving beyond recognition, we introduce the problem of discovering interventions that improve target indicators for a given aerial or street-view image. We argue that a black-box indicator model, combined with a generative editing model, can serve as an implicit digital twin for testing intervention hypotheses. We present VIDA-Geo , a multi-agent system that explores this intervention space by coordinating segmentation, diffusion-based inpainting, and indicator scoring models to produce interventions that are both perceptually realistic and aligned with real-world policies. We evaluate our system on 8 indicators across aerial and street-view imagery, measuring changes in factors such as perceived safety and greenery. Our approach outperforms existing baselines in many cases, achieving up to 2X higher perceptual quality and policy alignment scores. Finally, our model provides users with multiple candidate interventions, supporting an expert city-planner-in-the-loop workflow.
Hosam Elgendy, Utkarsh Mall
Mohamed bin Zayed University of Artificial Intelligence, Abu Dhabi, UAE
Modern cities rely on an increasing number of digital services to operate, but residents' daily needs are still difficult to meet. Services are fragmented and have little interoperability, placing a heavy operational burden on users. Existing digital platforms, urban foundation models, and intelligent assistants each address only isolated aspects of an urban task. But they struggle to reliably convert complex natural-language requests into executable cross-system workflows. We propose Urban-Agent, a tool-augmented agent framework for cross-system urban tasks. It couples the cognitive and reasoning capabilities of a large language model with a tool-set supporting code execution, API calls, and Model Context Protocol. Through one adaptive closed loop, it clarifies missing information before acting, grounds tool use in live observations, and aligns the final response with observed evidence and task constraints. To address the evaluation gap, we introduce Urban-Eval, a benchmark specifically designed for cross-system urban request. Unlike prior benchmarks that assess either general tool use or urban knowledge and reasoning, Urban-Eval evaluates both task results and execution quality, including required tool coverage, dependency validity, and evidence traceability. Experimental results indicate that Urban-Agent reaches a 71% task success rate, 10 points above the strongest baseline. This lead holds across GPT-5-mini, Gemini-2.5-flash, DeepSeek-V4-flash, and Qwen3-235B-A22B.
City systems are shaped by interactions among infrastructure, human activity, and environmental conditions. Existing earth observation and urban benchmarks are limited in unified modeling and in capturing their joint dynamics. We introduce AgenticCity-Bench, a longitudinal multimodal benchmark for multi-layer unified state modeling and agentic city task reasoning. The benchmark covers 16 cities in 12 countries across six continents from 2000 to 2025, and includes 18 modalities spanning urban function, activity, and environmental state. It organizes heterogeneous data sources around a common spatial reference and calendar index while preserving spatial and temporal coverage, provenance and tile-level availability. On this foundation, AgenticCity-Bench supports urban cross-modal translation together with diverse agentic city tasks involving forecasting, planning and resource analysis. This study provides a unified benchmark for studying the transition from multimodal urban and earth observation to joint understanding and agentic city decision-making. The dataset is available at https://anonymous.4open.science/r/AgenticCity-Bench.
Xiaofei Yu, Kailai Sun, Yanyu Mao +7
Renmin University of China · Singapore-MIT Alliance for Research and Technology, MIT · Massachusetts Institute of Technology +4