Niwa, Ryo (University of Tsukuba, AIST), Takami, Shunki (National Institute of Advanced Industrial Science and Technology), Shigenaka, Shusuke (National Institute of Advanced Industrial Science and Technology), Onishi, Masaki (National Institute of Advanced Industrial Science and Technology)

Scenario Generation Based on a Gaussian Mixture Model for the Stochastic Optimization of Crowd Control

Scheduled for presentation during the Regular Session "Generating driving scenarios II" (FrBT10), Friday, September 27, 2024, 13:50−14:10, Salon 18

2024 IEEE 27th International Conference on Intelligent Transportation Systems (ITSC), September 24- 27, 2024, Edmonton, Canada

This information is tentative and subject to change. Compiled on October 3, 2024

Abstract

Several spectators moving simultaneously after a large-scale event can cause massive congestion, and thus far, crowd control methods such as route guidance and allocation to the destination have been employed as valid countermeasures to relieve the massive congestion. Onsite guides aid in crowd-control operations, and previous studies reported that incorporating crowd simulation and stochastic optimizations for considering uncertainty can help achieve crowd control adaptable to any scenario. Increasing the number of measured scenarios used for stochastic optimization can improve the robustness of the optimal solutions; however, the high cost of measuring scenarios, such as crowd flow, prevents securing a sufficient number of measured scenarios. Although various scenario-generation methods have been proposed to duplicate the measured scenarios, adapting stochastic optimization is impossible because of the specialized development of scenario generation required for each field. Therefore, a scenario generation based on a Gaussian mixture model is proposed in this paper for the stochastic optimization of crowd control. The proposed method achieves robust crowd control with a small number of measured scenarios, which was previously impossible without a sufficient number of measured scenarios. Experiments validated the performance of robust control in stochastic optimization with and without the proposed method.