CPS・シミュレーション基盤

CPS & Simulation Foundation

CPS / Digital Twin / Urban OS

Core CPS, digital twin, and urban OS technologies at the Mobile Computing Lab that capture cities, people flows, transportation, and buildings in cyberspace and connect the real and virtual worlds

Cyber-Physical Systems (CPS) and digital twins are foundational technologies that reproduce real-world cities, transportation, buildings, and environments in cyberspace. By fusing simulation with real data, they enable policy evaluation and future prediction that would be difficult to test in reality. Our lab works on digital twin construction at various scales: traffic and pedestrian flow simulation, urban modeling using 3D point clouds and NeRF/3D Gaussian Splatting, and energy management simulation in smart homes.

We also pursue AI-integrated system design, including automatic generation of simulation agents using large language models (LLMs) and large-scale scenario evaluation through parallel simulation on HPC (High-Performance Computing) infrastructure. These technologies contribute to building urban OS-like platforms that support urban planning, transportation policy, and disaster preparedness.

Distributed Systems / System Software

Middleware and orchestration technologies that process and integrate large-scale data streams from IoT devices to the cloud

We research middleware and orchestration technologies for efficient processing, aggregation, and coordination of data streams in large-scale distributed environments spanning IoT devices, edge, and cloud. Key research themes include real-time data delivery via Pub/Sub messaging, ROS-based robot and autonomous system control, and online analysis of massive sensor data through IoT data stream processing.

We also work on federated learning for collaborative model training across multiple sites while preserving data privacy, mathematical optimization in distributed settings, multi-agent systems for cooperative decision-making, and real-time prediction with parallel and distributed computing from sensor data. These technologies serve as the computational foundation for our applied research in city-scale traffic/pedestrian simulation, disaster management systems, and smart agriculture.

Recent Related Publications

🏆JIP Specially Selected Paper
ICNN Based Distributed Optimization of 3D Point Cloud Quality for Real-Time Physical Space Sharing
Yui Maruyama, Tatsuya Amano, Hirozumi Yamaguchi
Journal of Information Processing, 2025, Volume 33, Pages 325-335
🏆MobiQuitous2024 Best Paper Award
Simulating Urban Pedestrian Flows by Fusing Wide-Area Location Data and Spot Pedestrian Counts
Masashi Uegaki, Tatsuya Amano, Hirozumi Yamaguchi
Mobile and Ubiquitous Systems: Computing, Networking and Services. MobiQuitous 2024, pp.550-569
🏆Best Short Paper Award
Generating Human Daily Activities with LLM for Smart Home Simulator Agents
Haruki Yonekura, Fukuharu Tanaka, Teruhiro Mizumoto, Hirozumi Yamaguchi
The 20th International Conference on Intelligent Environments (IE2024)
🏆Best Poster Award
One Model Fits All: Cross-Region Taxi-Demand Forecasting
Ren Ozeki, Haruki Yonekura, Aidana Baimbetova, Hamada Rizk, Hirozumi Yamaguchi
The 31st ACM SIGSPATIAL International Conference on Advances in Geographic Information Systems (ACM SIGSPATIAL 2023)
ROSを用いた編隊飛行制御アルゴリズムの高精度評価系の実装
天野辰哉, 廣森聡仁, 山口弘純, 森岡澄夫
第 13 回計測自動制御学会制御部門マルチシンポジウム
Agent-Based Modeling of Flood Evacuation Dynamics Driven by Large Language Models
Mizuki Funato, Fukuharu Tanaka, Hirozumi Yamaguchi
SCA/HPCAsia 2026 Poster
Global Indoor Positioning Powered by a Supercomputing-Driven Architecture
Fukuharu Tanaka, Hamada Rizk, Hirozumi Yamaguchi
SCA/HPCAsia 2026 Poster
High-Reliability Network-Distributed MoE-LLM System
Haruki Yonekura, Hirozumi Yamaguchi
ICDCN 2026 Doctoral Symposium
Challenges for Federated Crowd Management in Smart Cities
Daniela Nicklas, Leonie Ackermann, Debasree Das, Tatsuya Amano, Hamada Rizk, Hirozumi Yamaguchi
ICDCN 2026 Workshop: IWNDSC2026