Science AI Journal
Engineering · Research topic

Open research questions in Traffic control and management

41 unresolved questions extracted from the limitations and future-work sections of 656 Traffic control and management papers in our library. Each links back to the study that raised it.

What the literature leaves open

  • In con- sequence, deployment in the real world, robustness in the presence of uncertainty in traffic conditions, computational efficiency,and compatibility with IoT-based intelligent transportation infrastructure are important open challenges. In addition, the study highlighted several key research gaps related to real-world deployment constraints, scalability issues, heterogeneous traffic modelling, and some explainability of RL systems, and the lack of standardised benchmarking procedures.

    A Systematic Review of Traditional and Reinforcement Learning-Based Traffic Signal Control Methods · 2026 · DOI

  • TSC systems have changed dramatically from the conventional fixed-time and optimisation-based decentralised systems to the intelligent adaptive systems that rely on reinforcement learning and multi-agent coordination. The reviewed studies showed impressive gains in traffic efficiency, delay, and adaptive signal control; however, there are several key research gaps that are not adequately addressed. The literature reviewed shows that future research should not only consider enhancing traffic performance but also scalability, transferability, sustainability, safety, explainability, and real-world deployment capability.

    A Systematic Review of Traditional and Reinforcement Learning-Based Traffic Signal Control Methods · 2026 · DOI

  • Whether the PGQL gain transfers to asymmetric networks, multi-lane roads with turn pockets, heterogeneous block lengths, real rush hour O-D patterns, incident disruptions, or networks of 49 or more intersections has not been established by this work. The PGQL framework directly addresses this concern at the deployment level: even if Q-learning convergence guarantees are violated in practice (because the environment is non-stationary, exploration is limited, or training data is sparse), the warm-started agent's behaviour is bounded near max-pressure as long as the learned overrides remain infrequent. A note on the operational significance of the observed gains is warranted.

    Pressure-Guided Q-Learning: a warm-started hybrid controller for signalised urban grids · 2026 · DOI

  • This is evaluated using a microscopic simulated traffic platform, which, although capable of modelling detailed vehicle interactions in the real world, does not account for all real-world uncertainties, such as heterogeneous driver compliance, communication latency, sensor degradation, and mixed traffic compositions. These results demonstrate that isolated, segment-level control is insufficient to manage expressway networks characterised by strong spatial dependence and high traffic volatility.

    An AI-Enabled Intelligent Traffic Guidance System for Expressway Networks Based on IoT and GIS Integration · 2026 · DOI

  • Despite the fact that the suggested RL-based Vehicle Actuated Control (VAC) system is still able to achieve good results in simulation, there are several dimensions that may be used in researches in the future to make the Suggested system more effective and appropriate in reality. The suggested future course of action is to expand the suggested model to big scale traffic networks that include a quantity of interconnected crossways through multi-agent RL strategies. This would help in coordinating and managing traffic of various road lights within the city in terms of complete road systems. Another direction of work that is very significant in the future is the introduction of the real-life traffic data captured by the associated vehicle sensors, cameras and technologies in the graphical presentation of the traffic states to make them even more precise. The Vehicle-to-Everything communication the Internet of Things (IoT) technologies and technologies can also be added in order to equip the system to respond to the real-time situation on the traffic. It is also possible to research the further work in the future on the algorithms of DRL and hybrid optimization techniques to achieve the better stability of the learning and results of the traffic control. Application of the Suggested system within the existing structure would eventually lead to the formulation of the intelligent transport systems and smart city traffic management systems. traffic management (V2X) REFERENCES F. Rasheed, Q. Ni, and H. Xu, “DRLfor Vehicle Actuated Control (VAC): A review,” IEEE Access, vol. 8, pp. 208016–208044, 2020. D. Ma, Y. Wang, Q. Guo, and Y.

    Reinforcement Learning for Dynamic Traffic Signal Control Using Advanced Machine Learning Techniques · 2026 · DOI

  • Further research on the area of traffic congestion mitigation is strongly endorsed. Finally, future studies can be utilized from the traffic distribution outcomes of this study for further investigation of travel patterns, funding infrastructure expansion, evaluating the impacts of new projects, and boosting traffic performance assessment strategies.

    Operational and Statistical Assessment of Checkpoint-Induced Bottlenecks on a Selected Segment in Expressway No. 1, Iraq · 2026 · DOI

  • GRJNST, Volume: 04 - Issue 3 (2026) / ISSN P: 2790-7643 Article ID: 2082 https://doi.org/10.53762/grjnst.04.03.03 G. 2082 Page 27 The study advised that urban officials and policymakers should focus on the development of Smart Transportation Systems by investing in the advanced digital infrastructure and intelligent traffic management technologies. To enhance traffic coordination in congested regions, governments ought to increase the use of real-time traffic monitoring systems and control mechanisms that can adaptively change signal timing to enhance traffic coordination. It is also suggested that models of artificial intelligence and machine learning should be integrated into traffic management systems to improve predictive abilities and help to make proactive decisions. To ensure a smooth communication between vehicles and infrastructure, as well as control systems, transportation agencies are advised to enhance IoT-based data integration systems. In addition, capacity-building programs and technical training should be provided to transportation professionals to effectively manage and operate intelligent systems. Awareness campaigns should be encouraged as well to ensure that the people adopt smart mobility solutions and to ensure that the technology-driven transportation systems are accepted by the people.

    Smart Transportation Systems: Enhancing Traffic Flow and Reducing Urban Congestion through Intelligent Solutions · 2026 · DOI

  • The next research should be conducted on how to integrate new technologies like autonomous vehicles, digital twins, and blockchain into Smart Transportation Systems to further streamline the efficiency and scalability of such systems. Research is also needed to understand how big data analytics and edge computing can enhance the responsiveness of the real-time traffic and minimize the system latency. Empirical investigations of different cities and regions may give more knowledge to the efficacy of Smart Transportation Systems in diverse conditions of cities. Also, further studies can be carried out in the creation of hybrid models that utilize intelligent transportation GRJNST, Volume: 04 - Issue 2 (2026) / ISSN P: 2790-7643 Article ID: 2082 https://doi.org/10.53762/grjnst.04.03.03 G. 2082 technologies with sustainable mobility solutions like electric vehicles and shared transportation systems. There is also the need to explore policy frameworks and governance models that can support the large-scale implementation of Smart Transportation Systems in the developing countries.

    Smart Transportation Systems: Enhancing Traffic Flow and Reducing Urban Congestion through Intelligent Solutions · 2026 · DOI

  • Extraction of Traffic Time Intervals Data extraction was performed manually by reviewing video recordings, demonstrating high accuracy in vehicle counting (İlyas et al., 2024). The data will be categorized by vehicle type: motorcycles (MC), Light Vehicles (LV), and heavy vehicles (HV). For time headway data, various combinations of vehicle convoys were also observed manually, such as LV-LV (light vehicle followed by light vehicle), HV-LV (heavy vehicle followed by light vehicle), and so on, including MC-MC, MC-LV, LV-MC, HV-HV, and LV-HV as shown in Figure 2. Calculation of the Passenger Car Equivalent (PCE) The HV PCE value is calculated by dividing the average HV time headway by HV by the average LV time headway by LV. The result will be accurate if the HV time headway is independent of the vehicle in front or behind (Alenzi et al., 2022). This condition is met if the average time headway of LV, followed by LV, plus the average time headway of HV, followed by HV, is equal to the average time headway of LV, followed by HV, plus the average time headway of HV, followed by LV. The above can be expressed in (1): where ta is the average time headway between a Light Vehicle(LV) and the Light Vehicle(LV) following it, tb is the average time headway between heavy vehicle (HV) followed by heavy vehicle (HV), tc is the average time headway between light vehicle (LV) followed by heavy vehicle (HV), and td is the average time headway between heavy vehicle (HV) followed by light vehicle (LV). The conditions required to satisfy the above equation are difficult to meet because every vehicle has different characteristics. Similarly, drivers have varying driving abilities. Therefore, adjustments to the average time-distance value are necessary using (2): (1) Aswar et al. | 85 tatdtbtc+=+ Borneo Engineering: Jurnal Teknik Sipil (2) (3) Where na is the number of LV time headway data points followed by LV, nb is the number of HV time headway data points followed by HV, nc is the number of LV time headway data points followed by H, and nd is the number of HV time headway data points followed by LV. Next, the average time headway of the vehicle pairs is corrected using (4), (5), (6), (7), and (8): Using the corrected average time headway value, it can be concluded from (8) (4) (5) (6) (7) (8) where tak is the corrected average time difference between LV and LV, tbk is the corrected average time difference between HV and HV, tck is the corrected average time difference between LV and HV, and tdk is the corrected average time difference between HV and LV.

    Estimation of Passenger Car Equivalent (PCE) Values in Heterogeneous Traffic Using the Time Headway Method on Timor Raya Road, Kupang City · 2026 · DOI

  • Declarations The approach itself is limited by distances and the connected vehicles’ communication technology: • If the communication distance is too short, the decen- tralized communication infrastructure lacks sufficient connectivity for the consensus and gossip algorithms; • if the actuation section is too long, estimation quality suffers from aged information; • if the sensing section too short, estimation quality suf- fers as the time that vehicles participate in the consensus algorithm is too short.

    V2VSL: Infrastructure-Free, Decentralized Variable Speed Limit Control · 2026 · DOI

  • The paper does not evaluate the LoRa communication range, interference tolerance, or packet loss rates in urban environments with multiple wireless systems, which is critical for reliable emergency vehicle preemption across a city-wide interconnected signal network.

    Vision-Based Adaptive Traffic Signal System with LoRa Emergency Vehicle Priority · 2026 · DOI

  • The automatic incident detection enhancement mentioned in future work lacks specificity regarding what types of incidents (accidents, stalled vehicles, debris) should be detected or what vision-based features (e.g., trajectory anomalies, vehicle clustering patterns) would distinguish incidents from normal traffic congestion.

    Vision-Based Adaptive Traffic Signal System with LoRa Emergency Vehicle Priority · 2026 · DOI

  • The vision-based system relies solely on Raspberry Pi Camera Module v1.3 for traffic density analysis, but no comparison or validation is provided against other sensor modalities (ultrasonic sensors, inductive loops, LiDAR) under adverse weather conditions (rain, fog, nighttime) or varying lighting conditions.

    Vision-Based Adaptive Traffic Signal System with LoRa Emergency Vehicle Priority · 2026 · DOI

  • Cloud-based data analytics for city-wide traffic optimization is mentioned as a future enhancement, but the paper does not address bandwidth requirements, latency constraints, or cloud platform selection (AWS, Google Cloud, etc.) for processing real-time video streams from multiple Raspberry Pi Camera modules across numerous traffic signals.

    Vision-Based Adaptive Traffic Signal System with LoRa Emergency Vehicle Priority · 2026 · DOI

  • The paper mentions deploying multiple interconnected intersections for coordinated signal control but provides no specific coordination protocol, synchronization mechanism, or network topology design for how ESP32 microcontrollers with LoRa modules would communicate across multiple junctions in a city-wide system.

    Vision-Based Adaptive Traffic Signal System with LoRa Emergency Vehicle Priority · 2026 · DOI

  • The paper proposes integrating deep learning models for vehicle classification and traffic prediction but does not specify which deep learning architectures (e.g., YOLO, Faster R-CNN, ResNet) should be evaluated or tested against the current image processing approach used with the Raspberry Pi Camera Module v1.3 for real-time traffic monitoring.

    Vision-Based Adaptive Traffic Signal System with LoRa Emergency Vehicle Priority · 2026 · DOI

  • The research suggests that future transportation policy should prioritize the implementation of adaptive traffic control systems to reduce travel and waiting times, as well as emissions. Real-time, AI-driven traffic light systems should be developed to dynamically respond to traffic conditions in congested areas, thereby enabling more efficient traffic management. Egypt should also invest in innovative country initiatives, particularly in deploying ITS for traffic monitoring, predictive analytics for congestion management, and real-time rerouting via user navigation apps. These policy directions align with Egypt's urban development goals and can be tailored for similar cities facing traffic congestion challenges. Future research could expand on this work by applying the simulation framework across different intersections and integrating socio-economic variables to achieve more inclusive policy outcomes.

    Scenario based traffic optimization in Egypt performance gains through simulation modeling · 2026 · DOI

  • coupling − Stronger between agents’ decisions. − Longer decision chains. − Delayed assignment. credit Overall, the increased expressiveness and compatibility inevitably results in a higher complexity of a problem, which becomes characterized by higher non-stationarity and multiple competing coordination equilibria. 1.5. Implications for scalability and coordination The proposed reformulation naturally supports semi-synchronous decentralized decision-making, where agents act independently but remain coupled through shared environmental constraints and observations. Coordination emerges as agents adapt their local policies to the dynamically changing set of feasible joint signal configurations. By focusing agents on localized control regions rather than assigning full intersection control to a single agent, the framework improves robustness to local failures and distributes the difficulty of decision-making. This becomes increasingly important for intersections with complex geometries or high-dimensional traffic patterns, where centralized control is less robust. Overall, this reformulation reframes single-intersection traffic light control as a testing ground for decentralized coordination under uncertainty, providing the foundation upon which the learning framework, introduced in the subsequent sections, is built.

    MULTI-AGENT DEEP REINFORCEMENT LEARNING FRAMEWORK DESIGN FOR EFFICIENT SINGLE-INTERSECTION TRAFFIC LIGHT CONTROL · 2026 · DOI

  • RL gives substantial benefits in the application of transportation systems, where real-time adaptive control is critical to increasing efficacy and efficiency. Traffic Control approaches that rely on prespecified models of these processes are perceived to have a substantial disadvantage compared to the ability to learn through dynamic interaction with the environment. This paper introduces an innovative RL technique utilizing the DQL algorithm to minimize traffic congestion effectively. The system is structured based on an intersection-centered traffic model, emphasizing its ability to optimize waiting times and improve reward systems. This study’s findings represent a significant advancement in traffic management, creating an effective method for decreasing traffic congestion. Our current method effectively manages road intersections and makes optimal decisions to reduce traffic congestion. This advancement shows significant potential and is a crucial addition to traffic control. In the future, our model will be enhanced to work with real-time traffic data and optimization. This development will enable our system to connect to the internet, allowing the model to receive real-time data. In this capability, the agent can make informed decisions and adopt the optimal lane for vehicle movement. To address the high computational complexity in DQL for large-scale traffic networks, first, feature extraction, and dimensionality reduction techniques will reduce the state and action space. Secondly, more efficient neural network architectures will be used to improve processing efficiency. Additionally, techniques such as experience replay and target networks will stabilize learning and reduce redundant computations. Parallel computing and distributed learning will also be utilized to manage large-scale data by distributing the computational load across multiple processors, thereby cutting computational costs.

    A reinforcement learning approach for reducing traffic congestion using deep Q learning · 2024 · DOI

  • The primary objective of this study was to determine the effect of carriageway width, the radius of the horizontal curve, and gradients on Passenger Car Unit (PCU) values as well as on capacity of two-lane undivided Highways, and more importantly, to develop a multiple linear regression model to determine the capacity of the highway when all of these factors are present, which has not been previously reported.

    Two-Lane Highway Capacity Estimation Based on Geometric Features · 2023 · DOI

  • Although SSD is generally sufficient to allow skilled and alert drivers to the stop their vehicles under regular situations, this distance is insufficient when information is difficult to comprehend.

    Empirical modeling of the relationship between decision sight distance and stopping sight distance based on AASHTO · 2018 · DOI

  • CONCLUSION: The proposed method can diminish one of the challenges in front of transportation engineers, which is to identify high WWD crash locations due to insufficient information in crash reports.

    Prediction of Potential Wrong-Way Entries at Exit Ramps of Signalized Partial Cloverleaf Interchanges · 2014 · DOI

  • However, the model is limited by a linear assumption of tire degradation and deterministic race conditions, which may not fully capture stochastic on-track interactions.

    A unified graph-based optimization model combining shortest paths and Hamiltonian constraints for motorsport strategy · 2026 · DOI

  • Existing approaches address racing line optimization and pit stop decision-making as separate problems, leaving the formal integration of trajectory-level and strategy-level optimization within a single graph-theoretic architecture as an open problem.

    A unified graph-based optimization model combining shortest paths and Hamiltonian constraints for motorsport strategy · 2026 · DOI

  • Future research will focus on expanding the road element sets to include complex topologies such as roundabouts, grade-separated interchanges, and bridges.

    Complexity controllable road network generation for virtual testing of autonomous driving · 2026 · DOI

Most-cited papers in Traffic control and management

Most recent work

Find a gap in your own Traffic control and management sub-topic

This page shows what the Traffic control and management literature already flags as unresolved. To narrow it to your specific question, run the guided finder — it searches the gap library on demand and checks candidates against 250M+ OpenAlex works.

Open the Research Gap Finder →

Related topics in Engineering

Traffic and Road Safety
143 open questions
Spatial Cognition and Navigation
95 open questions
Smart Cities and Technologies
89 open questions
BIM and Construction Integration
87 open questions
Design Education and Practice
86 open questions
Engineering Education and Curriculum Development
82 open questions
Digital Transformation in Industry
81 open questions
Perovskite Materials and Applications
78 open questions
41 open questions have been extracted from the limitations and future-work passages of 656 Traffic control and management papers in our library. Each one below links back to the study that raised it, so you can read the original claim in context.

Tools for your next paper

Gaps & Literature SearchFind open questions in any field — 120,000+ open gaps mined from our own library.Pre-CheckIs your paper ready? A Tier 1–5 acceptance probability in seconds.Journal FinderA ranked shortlist of target journals from a 17,500-venue index.Duplicate Publication CheckerCheck prior-publication & salami-slicing risk across 8 sources.Citation GeneratorFreeA DOI or title → APA, MLA, Vancouver, Chicago, BibTeX, RIS. No signup.Graphical Abstract MakerFreeTurn your findings into a colorblind-safe graphical abstract.

Compare the category — Honest roundups of the AI research tools, ours listed alongside the alternatives.

Command palette

Jump anywhere, run any action.