Human-Flow Digital Twin for Predicting the Effects of Mobility Introduction on Visitor Circulation

2026-05-19 · Source: cs.MA updates on arXiv.org · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Data Science & Analytics, Public Transportation & Urban Mobility · Depth: Expert, extended

Summary

A new framework predicts the effects of introducing mobility services on visitor circulation using a human-flow digital twin. This digital twin integrates a multi-agent simulator that models how visitors choose destinations based on their current location and spot attractiveness. The simulator's decision model is trained using pre-intervention human-flow data, inter-spot distances, spot attractiveness, and travel volumes. By representing mobility introductions as changes to inter-point distances or spot attractiveness, the framework can simulate and quantify effects like changes in visitor counts and circulation. The method was evaluated using human-flow data from Wakayama Castle Park in Japan, demonstrating that a Multi-Layer Perceptron (MLP) decision model reproduced flows with mobility introduction with a cosine similarity of the spatial population distribution exceeding 0.7, confirming its ability to replicate flow changes.

Key takeaway

For urban planners and tourism managers evaluating new mobility solutions, this digital twin framework offers a cost-effective way to predict visitor circulation changes. You can simulate various deployment strategies by modifying environmental factors in the model, avoiding expensive field trials. Focus on MLP-based decision models for accuracy in smaller areas and consider explicit exit modeling to refine population distribution predictions.

Key insights

A human-flow digital twin predicts mobility introduction impacts on visitor circulation via agent-based simulation.

Principles

• Model mobility as environmental feature changes.
• Combine macro-level routing with micro-level dynamics.
• Calibrate demand using observed spot counts.

Method

The method involves an offline phase for data preparation and flow reconstruction using GMMs and a multi-agent simulation for trajectory refinement. An online phase then uses a trained neural network (MLP or GNN) as a destination-choice model within the simulator under modified environmental features to represent mobility introduction.

In practice

• Use MLP for destination choice in compact environments.
• Explicitly model exit as a decision class for accuracy.
• Visualize simulation results with 3D mapping tools.

Topics

• Human-Flow Digital Twin
• Multi-Agent Simulation
• Destination-Choice Modeling
• Mobility Introduction
• Wakayama Castle Park

Best for: AI Scientist, Research Scientist, Data Scientist

Related on AIssential

• Proposal-Conditioned Latent Diffusion for Closed-Loop Traffic Scenario Generation — A new diffusion framework improves closed-loop traffic simulation efficiency and control for autonomous vehicles.
• From Imitation to Alignment: Human-Preference Flow Policies for Long-Horizon Sidewalk Navigation — FlowPilot combines anchored flow matching with human-in-the-loop preference learning for robust, socially compliant sidewalk navigation.
• GraphWorld: Long-Horizon Planning with World Models for End-to-End Autonomous Driving — GraphWorld leverages latent world models and interaction graphs to achieve robust, long-horizon planning for end-to-end autonomous driving.
• From Seeing to Simulating: Generative High-Fidelity Simulation with Digital Cousins for Generalizable Robot Learning and Evaluation — WorldComposer generates diverse, high-fidelity simulation environments from real-world panoramas for robust robot learning.
• GPS-Enhanced Tourist Mobility Modeling with Seasonal Spatial Priors and LLM-Based Activity Chain Generation — The framework combines GPS data, demographics, and LLMs to simulate realistic, demographically-aligned tourist mobility for urban planning.
• PLAN-S: Bridging Planning with Latent Style Dynamics for Autonomous Driving World Models — PLAN-S enhances autonomous driving LWMs by explicitly modeling style-conditioned spatial costs for improved controllability and safety.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by cs.MA updates on arXiv.org.