IEEE MASS2024
D3 Student Tanaka Presents Research at IEEE MASS 2024 Conference
Publications & Talks
International Conference · 2026
Rapid Policy Impact Estimation for Pedestrian Flow Using Neural Network Surrogates
Abstract
High-fidelity pedestrian microsimulators are essential for evaluating crowd-management policies but are too computationally costly for rapid, interactive policy screening. We propose using neural network surrogate models that predict simulator outputs conditioned on network topology, demand, and policy parameters, enabling fast estimation of policy impacts without running the full simulator. We evaluate the surrogates on synthetic networks, measuring reconstruction accuracy on pedestrian flows and prediction error on a composite policy effectiveness score. Results show that a spatio-temporal graph-based surrogate provided the most accurate policy-impact estimates.
Research Note
Crowd management measures such as making a corridor one-way, opening an extra entrance or shifting the timing of announcements are worth testing before they are applied. The pedestrian microsimulators that can test them, however, take long enough per run that interactive comparison of alternatives is impractical.
Surrogate models replace the simulator with a learned approximation of its input-output behaviour, returning an answer immediately. What makes this case harder than usual is that the input is not simply a vector of numbers. Pedestrian flow depends strongly on how the walkable space is connected, so a model that ignores that structure has little chance of predicting the effect of a policy correctly.
The proposed surrogates therefore take network topology, demand and policy parameters as conditions and predict the simulator output from them, allowing policy impact to be estimated without running the full simulation. Evaluation on synthetic networks measured both how accurately pedestrian flows were reconstructed and how well a composite policy effectiveness score was predicted.
Among the configurations compared, a spatio-temporal graph-based surrogate gave the most accurate policy impact estimates, indicating that representing the walkway network explicitly as a graph is what carries the predictive power.
Related Projects
セマンティック通信による多端末連携型の状況理解と消防システムへの適用
Ministry of Internal Affairs and Communications (MIC) FORWARD
デジタルインフラ構築部門
