Abstract
Traffic management is improved in cutting-edge smart cities using technologies such as machine learning and deep learning to streamline daily tasks and boost productivity. However, traffic management still suffers from challenging issues such as poor traffic congestion prediction, lack of traffic flow management, public transportation optimization, and emergency management. In this article, we provide a thorough understanding of the benefits, drawbacks, and practical implications of leveraging machine learning and deep learning in Traffic Management Systems (TMSs) by methodically reviewing and critically analysing various traffic management techniques. We also present a generic traffic management architecture that uses a set of assessment criteria to evaluate 20 recently proposed research prototypes (i.e., published since 2019). Finally, we highlight the ongoing challenges and prospective trajectories within the rapidly evolving domain of traffic management, underscoring the need to address emerging issues and directions in its dynamic development. This survey article offers insights that might help in efficiently tackling the issues posed by traffic management while maximizing the potential of machine learning and deep learning techniques. This survey can be a significant resource for researchers, policymakers, and practitioners.
Introduction
The orchestration of automotive and pedestrian mobility over public thoroughfares, including roads, streets, and highways, is referred to as “traffic management”. At its core, traffic management aims to promote the safe and effective circulation of traffic, relieve congestion, and lower the number of collisions on the roads. These goals require the use of a wide range of strategies, including traffic engineering, the use of traffic control equipment, and careful transportation planning. All of these actions work to regulate vehicle movement and ensure the safety of all road users (Nellore and Hancke, 2016).
In modern civilizations, traffic management is crucial, especially in urban areas where the pervasive problem of traffic congestion looms large. Significant economic consequences of this congestion include lost productivity, wasted fuel, and increased air pollution. By improving traffic flow and reducing the amount of time commuters spend stuck in traffic, effective traffic management can lessen these effects, reducing stress and improving the well-being of both drivers and passengers (Alsrehin et al., 2019). People in general and tourists alike struggle with the effects of increased traffic and interruptions brought on by population development. Based on elements like accidents, road work, special events, and unlawful parking, it is possible to distinguish between regular and non-recurring traffic obstacles. Accidents stand out among them due to their direct effect on human life and their disproportionately high rate of fatalities among children.
Our society is experiencing the integration of intelligent technologies that improve our daily activities in an era of developing technology. A significant technical development known as the Internet of Things (IoT) connects many smart gadgets, allowing for smooth connection and data exchange. IoT technology is now more widely used in the transport sector, where it has a significant impact on how effectively transport systems operate. To further improve these systems, information technology must be integrated more. As travel demands continue to diversify, traditional traffic management methods are no longer adequate. Governments, such as China, have recognized the need for scientific and advanced traffic management approaches, which necessitate the integration of emerging Artificial Intelligence (AI) technologies (Cheng et al., 2021). To detect and prevent traffic-related concerns, researchers have resorted to Machine Learning (ML), Deep Learning (DL) techniques, and the quantity of geo-tagged data available on social media platforms (Azhar et al., 2023, Abdullah et al., 2023) which improved traffic management strategies.
Congestion prediction is given top priority by effective traffic management centres globally to provide a quick incident reaction. Due to the interdependencies of traffic flow in both time and space, this work is difficult. The ever-growing amounts of traffic data generated have been used to train ML predictions, but DL approaches have shown promise as alternatives and outperformed conventional models. Nevertheless, there are still questions about their applicability, accuracy, and ideal parameter adjustment (Mihaita et al., 2020). In fact, as traffic control centres work to maintain regular travel patterns and assure prompt incident clearing on a daily basis, traffic congestion continues to be a major concern (Jovanović et al., 2022). The use of deep learning in IoT applications is still in its infancy despite substantial efforts in recent years (Ma et al., 2019).
The research questions we aim to address are as follows:
- •
What machine learning and deep learning techniques have been used in traffic management systems?
- •
What are the advantages, disadvantages, and practical implications of using machine learning and deep learning in traffic management systems?
- •
How can a generic traffic management architecture be evaluated using predefined assessment criteria to handle traffic management challenges?
In this work, we overview the benefits, drawbacks, and practical implications of leveraging ML and DL in TMSs by methodically reviewing and critically analysing various traffic management techniques. In addition, we present a generic traffic management architecture that uses 12 different assessment criteria among 3 layers to evaluate 20 recently proposed research prototypes. Furthermore, we highlight current research issues and future research directions in the field of traffic management. This survey article is considered a guideline for researchers and practitioners who are planning to leverage ML and DL techniques to enhance their traffic management systems.
The rest of this article is organized as follows. Section 2 overview the background. Related Work is discussed in Section 3. Section 4 illustrates the design of the traffic management framework architecture. Section 5 presents the research prototype evaluation. Section 6 provides the traffic management open issues and future directions. Finally, concluding remarks are presented in Section 7.
Section snippets
Background
We give a brief overview of traffic management in this section. The core ideas and guiding principles of machine and deep learning are then introduced. We also go through popular deep-learning models that are useful for analysing, detecting, and predicting traffic data.
Related work
Manibardo et al. (2021) investigate how well deep learning approaches work for predicting traffic on roads. On traffic datasets, the authors assess how well three deep learning models LSTM, CNN, and a mixture of the two perform in comparison to more conventional machine learning models. They deduced from the experimental findings that shallow modelling techniques frequently outperform deep learning models in terms of performance. Additionally, they outline research opportunities for the given
Traffic management framework architecture
To understand the benefits and drawbacks of machine learning and deep learning techniques in traffic management, we propose a generic traffic management architecture and distinguish a set of assessment criteria to evaluate recently proposed research prototypes. The results of this evaluation will help us in identifying possible open issues in the field of traffic management. Fig. 1 depicts a generic architecture for traffic management that we propose in this section.
Research prototypes
This section provides a comprehensive analysis of 20 representative traffic management research prototypes. These research prototypes include a variety of studies that have undergone quantitative analysis for a full examination. The research prototypes were chosen to represent a wide set of Machine Learning (ML) and Deep Learning (DL) techniques used in the subject of traffic management. We ensure a rigorous and objective assessment of their findings and consequences by exposing them to
Traffic management open issues and research direction
Traffic management is a complex problem that can benefit from the use of Deep Learning (DL) techniques. However, several issues need to be addressed when applying DL to this field. Some of these challenges and future research directions include:
1. Poor Traffic Congestion Prediction: With traffic congestion being part of the focus in the evaluated research prototypes (with only 40% of the overall attention dedicated to it), the findings reveal the need for effective measures to alleviate
Conclusion
Traffic management has developed into a popular research area in recent years, attracting the interest of several researchers to tackle issues including flow prediction and traffic congestion. In this article, we thoroughly evaluated and analysed a wide range of strategies, methodologies, and technologies used in traffic management and discussed future research directions. Moreover, we present a generic traffic management architecture that has three different layers including the data
CRediT authorship contribution statement
Hanan Almukhalfi: Data curation, Formal analysis, Investigation, Methodology, Resources, Validation, Visualization, Writing – original draft. Ayman Noor: Conceptualization, Data curation, Funding acquisition, Investigation, Resources, Supervision, Validation, Visualization, Writing – original draft, Writing – review & editing. Talal H. Noor: Conceptualization, Formal analysis, Funding acquisition, Investigation, Methodology, Project administration, Resources, Supervision, Writing – review &
