Multi-Modal AI Framework for Intelligent Urban Infrastructure Management Using Deep Reinforcement Learning
Journal of Multidisciplinary Knowledge · 2025
DOI: 10.36676/jmk.v5.i2.101 ↗Kurz gefasst
Fachartikel in der Zeitschrift Journal of Multidisciplinary Knowledge, 2025. Verfasst von Prof. Hiroki Sakamoto. Die Arbeit nennt 19 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY-NC 4.0).
Abstract (Originalsprache)
Cities are increasingly deploying digital infrastructure to improve mobility, resource distribution and public services. This study introduces a multi-modal AI framework that integrates deep reinforcement learning with sensor, image and traffic flow data to optimize urban infrastructure management. The system uses a hierarchical action policy, where local agents control traffic signals, road segments and energy nodes, while a global coordinator learns long-term system-wide policies. A multimodal encoder merges video feeds, IoT sensors and geospatial data to enhance situational awareness. Testing on the CityFlow and METR-LA datasets showed a 21 percent reduction in congestion, 14 percent decrease in energy consumption and significant improvements in emergency routing response time. The framework also includes an interpretability module that translates agent decisions into natural-language explanations for urban planners. The findings confirm that multimodal reinforcement learning can provide significant benefits for managing complex city systems. Future work will focus on edge deployment and digital twin integration.
Zitieren
APA
Sakamoto, P. H. (2025). Multi-Modal AI Framework for Intelligent Urban Infrastructure Management Using Deep Reinforcement Learning. Journal of Multidisciplinary Knowledge. https://doi.org/10.36676/jmk.v5.i2.101
Vancouver
Sakamoto PH. Multi-Modal AI Framework for Intelligent Urban Infrastructure Management Using Deep Reinforcement Learning. Journal of Multidisciplinary Knowledge. 2025. doi:10.36676/jmk.v5.i2.101
BibTeX
@article{Sakamoto2025,
author = {Sakamoto, Prof. Hiroki},
title = {Multi-Modal AI Framework for Intelligent Urban Infrastructure Management Using Deep Reinforcement Learning},
journal = {Journal of Multidisciplinary Knowledge},
year = {2025},
doi = {10.36676/jmk.v5.i2.101}
}
EUVIMED