| 作成者 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 本文言語 |
|
| 出版者 |
|
| 発行日 |
|
| 収録物名 |
|
| 巻 |
|
| 開始ページ |
|
| 終了ページ |
|
| 会議情報 |
|
| 出版タイプ |
|
| アクセス権 |
|
| Crossref DOI |
|
| 権利関係 |
|
| 権利関係 |
|
| 概要 |
Urbanization in the Philippines presents challenges such as traffic congestion and road safety, particularly with the increasing number of vehicles on the road. This study proposes an intelligent traf...fic management system utilizing IoT and Convolutional Neural Networks (CNNs). Specifically, YOLO models were adapted to detect local vehicle types unique to Butuan City. A hybrid YOLO classifier and DeepSORT are integrated for real-time vehicle detection, classification, and tracking, with transfer learning on local datasets to enhance model precision. To dynamically control traffic signals, the system incorporates multiple deep reinforcement learning (RL) methods, including Deep Q-Learning (DQL), Covariance Matrix Adaptation Evolution Strategy (CMA-ES), and Advantage Actor-Critic (A2C). The RL-based controller learns and adapts real-time traffic signal timing based on vehicle density, waiting times, and phase duration at intersections. Among the methods, A2C demonstrated the highest efficiency, significantly reducing vehicle waiting times and enhancing traffic throughput. Comparisons showed that both DQL and CMA-ES provided robust performance, each contributing unique advantages in different traffic scenarios, though A2C emerged as the optimal solution in simulations. This comprehensive approach highlights the potential of combining IoT, CNN-based vehicle detection, and adaptive RL controllers, offering a scalable solution to improve urban traffic efficiency and support sustainable city planning initiatives.続きを見る
|
Mendeley出力
2025_p0762