Scholarlink

<p>Accurate, real-time traffic flow prediction is crucial for mitigating the substantial economic and environmental impact of urban congestion and ensuring the resilience of Smart Mobility systems. This study addresses the critical challenge of providing predictive intelligence for non-signalised intersections, a decentralized yet crucial infrastructure component within Urban Cyber-Physical Systems (CPS), which has often been overlooked by prior research focused on signalised roads. To achieve this, state-of-the-art Deep Learning (DL) approaches, including Recurrent Neural Networks (RNN), Long Short-Term Memory (LSTM) networks, and Convolutional Neural Networks (CNN), have been developed and comparatively analyzed specifically for this non-signalized context. Using historical and real-time traffic data, the models are evaluated against key metrics (RMSE and MAE). The LSTM model proved superior in capturing temporal dependencies, achieving optimal predictive performance while maintaining robustness across varying traffic volumes, vehicle speeds, and road conditions, as confirmed by a detailed sensitivity analysis. The models’ high accuracy provides actionable insights for real-time traffic management systems, enabling autonomous decision-making and enhancing the resilience and adaptability of smart mobility networks in dynamic urban environments.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Comparative evaluation of recurrent and convolutional neural network architectures for traffic flow prediction at non-signalised intersections

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年5月28日

查看全文
  • Thabo Matue,
  • Andronicus A. Akinyelu,
  • Mokotsolane Mase

摘要

Accurate, real-time traffic flow prediction is crucial for mitigating the substantial economic and environmental impact of urban congestion and ensuring the resilience of Smart Mobility systems. This study addresses the critical challenge of providing predictive intelligence for non-signalised intersections, a decentralized yet crucial infrastructure component within Urban Cyber-Physical Systems (CPS), which has often been overlooked by prior research focused on signalised roads. To achieve this, state-of-the-art Deep Learning (DL) approaches, including Recurrent Neural Networks (RNN), Long Short-Term Memory (LSTM) networks, and Convolutional Neural Networks (CNN), have been developed and comparatively analyzed specifically for this non-signalized context. Using historical and real-time traffic data, the models are evaluated against key metrics (RMSE and MAE). The LSTM model proved superior in capturing temporal dependencies, achieving optimal predictive performance while maintaining robustness across varying traffic volumes, vehicle speeds, and road conditions, as confirmed by a detailed sensitivity analysis. The models’ high accuracy provides actionable insights for real-time traffic management systems, enabling autonomous decision-making and enhancing the resilience and adaptability of smart mobility networks in dynamic urban environments.