16 Paper Presentations at ICDCN2026
16 presentations and 2 awards at ICDCN 2026 held in Nara, Japan (January 2026)
Publications & Talks Awards
International Conference · 2026
Bayesian Optimization Approach for Crowd Flow Modeling on Railway Station Platforms
Abstract
Accurately reproducing pedestrian behavior on railway station platforms is essential for safety management and congestion mitigation. This paper presents a digital-twin framework that integrates real video analysis and crowd simulation through Bayesian optimization. The method combines flow-level information from optical flow and person-level information from object detection and tracking, and reduces discrepancies between observations and simulation by automatically updating parameters. RAFT estimates pixel-level motion vectors, while ByteTrack provides individual positions and counts. From these data, differences in flow distributions, crowd counts, and positional matching are computed to form an objective function for parameter optimization. Using 360 frames of real footage over 100 iterations, the framework showed a clear reduction in error and achieved about one-third of ByteTrack’s MAE and RMSE, accurately reproducing real crowd density and movement patterns.
Research Note
A railway station platform concentrates a large number of people into a narrow space in bursts timed by arriving trains, which makes faithful reproduction of pedestrian behaviour directly relevant to safety management. Pedestrian simulators can express such behaviour, but they carry many parameters governing walking speed, personal spacing and similar factors, and tuning them by hand for a specific site is laborious.
Bayesian optimization addresses exactly this kind of problem. It is a method for finding good inputs to an expensive function using few evaluations, by maintaining a probabilistic estimate of the function shape from the trials so far and choosing the next trial where it expects to learn the most.
This work uses it to couple video analysis with crowd simulation in a digital twin. Two complementary sources of information are extracted from the footage. RAFT provides pixel-level motion vectors that describe the flow as a whole, while ByteTrack supplies individual positions and counts. Differences in flow distribution, in crowd count and in positional matching are combined into an objective function, and simulation parameters are updated automatically to reduce it.
Over 100 iterations on 360 frames of real footage the error fell clearly, reaching about one third of the MAE and RMSE of ByteTrack and reproducing the observed crowd density and movement patterns.
Related Lab Activities
Related Projects
LLM-Driven Urban Transportation Simulation Platform
Japan Science and Technology Agency (JST) JST Strategic Basic Research Programs (PRESTO)
[Social Transformation Platform] Co-Creation of the Transformation Platform Technology for Human and Society by Integration of the Humanities and Sciences
