Table 1 Overview of related work.
From: Scenario based traffic optimization in Egypt performance gains through simulation modeling
Ref | Main Target / Objective | Methodology | Dataset / Simulation | Metrices | Key Results | Pros | Cons / Limitations |
|---|---|---|---|---|---|---|---|
Andrea et al | Enhanced comprehension of traffic signal scheduling | Cellular Genetic Algorithms (CGAs), in-depth solution analysis | Synthetic traffic scenarios | Convergence velocity, solution variability, transit duration | CGAs outperform conventional GAs in both convergence and diversity | Offers a profound understanding of the solution space and is adaptable to complex intersections | Characterized by high computational demands, predominantly reliant on simulations, and may lack applicability to real-time scenarios |
Seyit et al | Real-time adaptive traffic signal control | Swarm Optimization: PSO, ACO | SUMO traffic simulator | Average waiting time, queue length, travel time | Substantial decrease in wait times and queues | Capable of real-time operation; straightforward implementation; adaptable | Vulnerable to algorithmic settings; efficacy may diminish under inconsistent traffic conditions |
Tuo et al | Improve GA-based traffic control using ML | Boosted Genetic Algorithm guided by Machine Learning | SUMO simulation | Travel time, delay, throughput | Hybrid GA-ML reduces transit time and delays, thereby accelerating convergence | Combines ML guidance with GA to increase convergence and efficiency | Significant complexity; necessitates training data; extended configuration duration |
Muzamil et al | Compare fuzzy logic strategies for traffic signal timing | Fuzzy logic controllers with state inputs | SUMO simulation | Throughput, delay, stops | State-based fuzzy controllers improve traffic throughput | Flexible and interpretable; can model nonlinear traffic patterns | Manual rule creation involves specialized knowledge, and performance is susceptible to adjustment |
José et al | Adaptive traffic control using fuzzy logic + classical formulas | Fuzzy logic combined with Webster and Modified Webster formulas | SUMO traffic simulator | Delay, stops, and queue length | decreased stops and delays in comparison to fixed-time control | Combines traditional and sophisticated methods; adaptable | Depends on precise traffic estimation; simulation-based |
Luow et al | Distributed cooperative intersection control | Multi-agent Reinforcement Learning (MARL) | SUMO simulation | Delay, travel time, throughput | Cooperation increases intersection efficiency and lessens traffic | Decentralized; scalable to multiple intersections | Training stability issues; high complexity |
Dimitrius et al | RL-based traffic signal optimization with dual agents | Dual-agent Double Deep Q-Network (DDQN) | SUMO simulation | Travel time, waiting time, convergence rate | Faster convergence and reduced delays vs standard DQN | Addresses overestimation; efficient learning | Requires a large training dataset; computationally intensive |
Xiaoyi et al | Compare reinforcement learning agents for traffic signal optimization | Comparative study of RL agents (DQN, DDQN, etc.) | SUMO simulation | Travel time, delay, throughput | Identified the strengths and weaknesses of different RL agents | Provides benchmarking insights; practical guidance | Results scenario-dependent; may not generalize universally |
Carvalho et al | Hierarchical traffic signal control using RL | Hierarchical Reinforcement Learning (Options Framework) | SUMO simulation | Travel time, waiting time, and learning efficiency | Improved learning efficiency; better long-term decision-making | Scalable; handles complex decision sequences | Increased algorithmic complexity; longer training required |
Ilhan et al | Validate the accuracy of SUMO traffic flow generation | Statistical comparison of simulated vs. real traffic | Real-world traffic data; SUMO simulation | Traffic volume, speed, and flow patterns | SUMO closely replicates real traffic behavior | Validates SUMO for research; important for simulation-based studies | Not a control method; limited to flow validation |
González et al | Estimate road pollution from vehicle movements using video surveillance | Deep Convolutional Neural Networks for vehicle detection and tracking; trajectory-based emission estimation | Surveillance traffic videos; UNLV dataset; traffic flow analysis | Recall, vehicle trajectory accuracy, and turning movement counts (TMC) | Achieved a recall ≈ of 0.62; reliable vehicle trajectories enabling pollution estimation | Non-intrusive pollution monitoring uses existing cameras and supports environmental traffic analysis | Sensitive to lighting and weather; limited real-time performance; struggles in dense or complex scenes |
Al-Zoghby et al | Real-time adaptive traffic management to reduce congestion and emissions | YOLOv11-based vehicle detection; traffic density estimation; dynamic signal optimization | Real-time camera feeds; traffic simulation for performance evaluation | mAP, F1-score, waiting time, fuel consumption, emissions | High detection accuracy (mAP 92.4%, F1 89.7%); significant reduction in waiting time and emissions | Real-time capable; high accuracy; integrates traffic efficiency with sustainability goals | Performance degrades under severe weather and very low-light conditions |
Biramo et al | Evaluate the impact of automated vehicles on fuel consumption and emissions | Traffic microsimulation under different AV penetration and automation scenarios | 22-km simulated test track; five AV penetration scenarios | Fuel consumption, CO₂ emissions, traffic flow indicators | 25% AV penetration reduced fuel and CO₂ by up to 8.35%; higher penetration yields greater benefits | Quantifies environmental impact; beneficial for transport planning and policy analysis | Entirely simulation-based; assumes predefined driving behavior; no real-world validation |
