A critical review of the available literature underscores the potential of DL to improve congestion management and provides important pointers for future development in order to make it more applicable to sustainable and intelligent transportation systems.
Abstract
Problem of managing and controlling traffic congestion in Intelligent Transportation System (ITS). Classical, statistical and model-driven methods usually fail to take into account the nonlinear dynamic properties of congestion especially for that which concern abrupt transitions between free-flowing and congested states. Deep Learning (DL) has recently shown effective results as it is able to process large scale, high resolution data and elaborate on complex patterns that traditional methods tend to miss. To this end, a comprehensive survey is presented to investigate the most recent literature on DL-based methods for congestion prediction, detection and control, focusing on methodological advances, model diversity and performance improvements with respect to accuracy, scalability and computational complexity. At the same time, it highlights remaining gaps such as lack of robustness support in real-world scenarios, haphazard metrics interpretations and trade-offs between predictive accuracy and scalability. Through a critical review of the literature, this paper. In reviewing the available literature, this work underscores the potential of DL to improve congestion management and provides important pointers for future development in order to make it more applicable to sustainable and intelligent transportation systems.
A key contribution of this study lies in its comparative synthesis of ML and DL models, revealing that hybrid and graph-based DL architectures consistently outperform traditional ML methods when handling large-scale, heterogeneous traffic datasets.
Thabo Matue, A. A. Akinyelu, Mase Mokotsolane· International Journal of Dat...· 0 citations
This systematic review critically examines recent advancements in deep learning-based traffic flow prediction models, emphasizing studies published between 2018 and 2026 and indicates that although graph-based and transformer-based architectures currently achieve state-of-the-art predictive performance, integrating explainability, real-time adaptation, and privacy-preserving learning remains a significant research priority.
Dr. P. K. S. Bhadauria· International Journal of Cre...· 0 citations
With the acceleration of urbanization, traffic congestion at multiple intersections has become a core bottleneck restricting urban operational efficiency. To address this, this paper proposes a traffic signal dynamic optimization algorithm, the Cross-Attention Mechanism and Dueling Double DQN (CAM-D3QN). This method utilizes a novel crisscross attention module to dynamically model spatial dependencies between intersections and incorporates the Dueling Double DQN architecture for robust Q-value estimation. Validated on CityFlow using Grid-4×4 and Hangzhou-real networks, CAM-D3QN significantly outperforms the state-of-the-art baseline, GPLight, achieving relative improvements of approximately 10.5% in average vehicle delay, 11.3% in average queue length, 3.3% in throughput, alongside notable reductions in stops (12.1%) and fuel consumption (7.2%). Ablation experiments further demonstrate that removing the cross-attention module increases queue length by 30.6% in sudden congestion scenarios. The proposed method achieves superior performance to the baseline across four typical traffic scenarios on the Hangzhou-real road network, demonstrating its generalization capabilities. By leveraging the coordinated optimization of dynamic spatial perception and robust value assessment, this paper provides an effective solution for efficient and robust coordinated traffic signal control. The framework can be combined with traffic states acquired from radar, roadside sensors or wireless communication units in intelligent transportation systems.
L. Chang, D. Wei· Advanced Electromagnetics· 0 citations
A hybrid deep learning framework that integrates Bidirectional Long Short-Term Memory and Gated Recurrent Unit networks with an attention mechanism within a federated learning paradigm is proposed, which enables decentralized model training across multiple data sources without requiring raw data sharing, thereby preserving privacy while maintaining predictive performance.
Apurba Nandi, Shaoni Banerjee, Avik Kumar Das et al.· Neural computing & applicati...· 0 citations
Accurate Traffic Flow Forecasting (TFF) is important for emerging Intelligent Transportation Systems (ITS) that support active traffic management, optimize routes, and reduce congestion. In this paper, Deep Learning (DL) methods for TFF, with an emphasis on models like Recurrent Neural Networks (RNN) reinforced with attention mechanism, Bidirectional Long Short-Term Memory (Bi-LSTM), as well as Stacked Autoencoder (SAE) is used for ITS. In complex traffic situations, these strategies improve prediction accuracy and remove nonlinear spatial-temporal networks. Bio-inspired optimization methods, such as the Fruit Fly Optimization Algorithm (FFOA), Philippine Eagle Optimization (PEO) and Kookaburra Optimization Algorithm (KOA) are reviewed for adaptive learning, weight initialization and optimal parameter adjustment in order to further improve model performance. The model architectures, optimization techniques, and assessment criteria discussed in recent research are compared in this review to show how they contribute to precise RMSE, MAPE, MAE traffic forecasts. With a focus on multi-source data fusion, real-time adaptability and interpretable AI frameworks for next-generation ITS, it concludes by identifying research gaps and future creativities.
V. Poornima, M. Subashini· International Conference on...· 0 citations
The Intelligent Traffic Management Systems (ITMS) have become an important feature of the smart city infrastructure because of the fast increase in the city population and the consequent urban traffic congestion, fuel use, and road accidents. Conventional methods of traffic management apply a lot on the operation of fixed-time control mechanisms and rule-based systems, which are not flexible to the dynamic traffic conditions. In the recent past, progress in the field of deep learning has resulted in the creation of data-driven systems of traffic management that can learn intricate spatial and temporal patterns of traffic through major sources of heterogeneous data. The paper provides a detailed research on designing, implementing and testing of an Intelligent Traffic Management System based on deep learning. The suggested system combines convolutional neural networks (CNNs) to estimate the traffic density, recurrent neural networks (RNNs) and long short-term memory (LSTMs) to predict the traffic flow, and reinforcement learning (RLs) to control traffic signals. Various data sources such as live video streams, sensor data and past traffic data are used to improve accuracy of predictions and effectiveness of decisions. The proposed system architecture includes a modular architecture that will include all the layers of data acquisition, preprocessing, model training, and real-time deployment. Numerous experiments on benchmark traffic datasets have shown that congestion is greatly reduced, the average vehicle waiting time is minimized and the traffic throughput is much improved in comparison to traditional systems. The findings confirm that traffic management systems based on deep learning can contribute significantly to the improvement of urban mobility, environmental impact, and road safety. This paper has given relevant information about the application of AI-driven traffic control systems in practice and opened up the prospects of future research in intelligent transportation systems.
Ibrahim A. Lawal· International Journal of Art...· 0 citations