Policy Optimization for Pedestrian Traffic Management by Surrogation of Simulation Models
2024 IEEE 21st International Conference on Mobile Ad-Hoc and Smart Systems (MASS), Seoul, Korea, Republic of, 2024, pp. 203-211
DOI: 10.1109/MASS62177.2024.00036
Abstract
This paper proposes a method to swiftly develop crowd control policies that enhance the smooth flow of pedestrian traffic. The method employs a novel, gradient-based black-box optimization technique, which uses a neural network to create a surrogate model replicating a multi-agent simulator. This approach facilitates the rapid generation of improved policies by leveraging gradient information, thereby significantly decreasing the time required to determine optimal strategies. By treating the simulator and its evaluation function as differentiable entities, the method allows for quick policy adjustments guided by gradient information. The proposed method quickly and accurately obtains policy parameters, adapting seamlessly to different optimization strategies in various scenarios. Our methodology was tested by applying it to actual pedestrian flow data from Koshien Stadium in Hyogo, Japan. The real-world application not only confirmed the practical utility of our optimization technique in effectively managing crowd scenarios but also marked a significant advancement in public safety measures for densely populated areas.
Sports events, concerts, and live performances attract large crowds, and the congestion during the return home after such events increases the risk of accidents, obstruction of emergency evacuations, and crime, posing a threat to public safety.
Therefore, appropriate crowd management, traffic planning, and security enhancements are required.
In particular, congestion mitigation through behavioral changes, such as phased departures and circulation promotion, has gained attention.
However, predicting the effectiveness and cost efficiency of such policies is challenging, and large-scale simulations require significant time, making traditional methods inefficient for optimal policy exploration.
To address this issue, this study constructs a surrogate model of a multi-agent simulation using neural networks and utilizes the gradient information of this model to efficiently search for optimal policies, enabling the rapid derivation of effective policies.
Published Papers:
- 田中福治, 天野辰哉, 内山彰, 廣森聡仁, 山口弘純, 中村佑輔, 小出英理, 勝間田優樹, “匿名化処理されたメッシュ間移動データからの人流再現手法の検討,” モバイルコンピューティングと新社会システム (MBL-105) , pp1-3, 2022年11月.
- 田中福治, 天野辰哉, 内山彰, 廣森聡仁, 山口弘純, 中村佑輔. (2024). ニューラルネットワークを用いた歩行流制御施策の最適化手法の提案と評価. 研究報告モバイルコンピューティングと新社会システム (MBL), 2024(15), 1-8. https://ipsj.ixsq.nii.ac.jp/records/232575
- 田中福治, 天野辰哉, 内山彰, 廣森聡仁, 山口弘純, 中村佑輔. (2024). 模倣学習型ニューラルネットワークを活用した歩行流制御施策の最適化. 第 86 回全国大会講演論文集, 2024(1), 41-42. https://ipsj.ixsq.nii.ac.jp/records/236338
- F. Tanaka, T. Amano, A. Uchiyama, A. Hiromori, Y. Nakamura and H. Yamaguchi, "Policy Optimization for Pedestrian Traffic Management by Surrogation of Simulation Models," 2024 IEEE 21st International Conference on Mobile Ad-Hoc and Smart Systems (MASS), Seoul, Korea, Republic of, 2024, pp. 203-211, https://ieeexplore.ieee.org/document/10723552