AI Digital Twin Predicts Visitor Flow Changes from New Mobility
New simulation model predicts how adding transport options reshapes crowd movement in real city parks.
A team of Japanese researchers led by Chiharu Shima has developed a human-flow digital twin framework that uses multi-agent simulation to predict how introducing new mobility options affects visitor circulation in public spaces. The system, accepted at IEEE MDM 2026, is designed to help urban planners quantify the impact of interventions like adding shuttles, walkways, or changing spot attractiveness before any physical changes are made.
The framework works by first training a multi-agent decision model (using a multi-layer perceptron) on pre-intervention data, including inter-spot distances, spot attractiveness, travel volumes, and actual human-flow patterns. Once trained, each agent chooses its next destination based on current location and surrounding environmental factors. By representing mobility introductions as changes to inter-point distances or spot attractiveness, the simulator can reproduce post-intervention human flows. In tests at Wakayama Castle Park in Japan, the model achieved a cosine similarity of over 0.7 for spatial population distribution, confirming it can accurately replicate flow changes caused by mobility interventions.
- Multi-agent simulator trained on pre-intervention data including spot attractiveness and travel volumes to predict destination choices.
- Achieved >0.7 cosine similarity in reproducing actual post-intervention spatial population distribution at Wakayama Castle Park.
- Enables planners to simulate effects of changes to inter-spot distances or spot attractiveness before deploying mobility measures.
Why It Matters
City planners can now simulate mobility interventions to optimize visitor flow and reduce congestion without costly trial-and-error.