How does a traffic light decide when to change, and what is a smart intersection?

Numbers match the citations in the article. A working link isn't proof that a page supports a claim; check the quoted passage and date.

1. 1

   Traffic light control and coordination (Wikipedia)

   WikipediaPublished Oct 15, 2026Checked Oct 17, 2026

   “The normal function of traffic lights requires more than sight control and coordination to ensure that traffic and pedestrians move as smoothly and safely as possible. A variety of control systems are used to accomplish this, ranging from simple clockwork mechanisms to sophisticated computerized control and coordination systems that self-adjust to minimize delays for people using the junction. In the United States, traffic signal timing is traditionally operated to minimize vehicle delay at traffic signals. This often affects the safety and mobility of people walking and riding bicycles.”

2. 2

   IoT-Simulated Digital Twin with AI Traffic Signal Control for Real-Time Traffic Optimization in SUMO.

   Sensors (Basel, Switzerland) (Ceapă et al.)Published Mar 17, 2026Checked Oct 17, 2026

   “Urban traffic congestion leads to longer travel times, economic losses, and increased pollution. Recent advances in the Internet of Things (IoT) provide detailed real-time traffic data, yet testing adaptive control strategies directly on live networks remains costly and risky. To address this challenge, we propose an IoT-driven digital twin framework for the design and evaluation of AI-based traffic management systems. The framework is implemented in the Simulation of Urban MObility (SUMO) and uses its Python 3.14.2 API to emulate a dense network of IoT sensors that stream real-time information on vehicle density, queue lengths, and waiting times. This simulated IoT data feeds an AI agent that adapts traffic signal control in real time. The agent is trained with a composite reward function to jointly minimise vehicle waiting times and emissions. Its performance is compared with fixed-time and vehicle-actuated control under varying traffic demand scenarios. Results demonstrate the effectiveness of combining IoT-based simulation with AI control, providing a safe and scalable pathway towards the real-world deployment of intelligent traffic management systems.”

3. 3

   Adaptive traffic signal control using deep reinforcement learning: Toward smarter and safer urban mobility.

   PloS one (Alanazi et al.)Published Mar 30, 2026Checked Oct 17, 2026

   “Thus, the need arises to develop intelligent systems capable of optimizing traffic flow in real time, reducing delays, and addressing the growing challenges of intelligent transportation systems. To address these requirements, a novel deep reinforcement learning framework that combines the Twin Delayed Deep Deterministic Policy Gradient (TD3) with prioritization-based Intelligent Traffic Control (P-ITC) is proposed for real-time traffic signal optimization using stability techniques. The module focuses on TD3's stability-enhancing techniques, including clipped Q-learning, delayed and targeted policy updates, and smoothing. The system ensures robust signal timing decisions across intersection networks. PER prioritizes critical traffic signal experiences, ensuring the system learns from key events that influence real-time traffic flow. The proposed TD3P-ITC framework achieves maximum reductions in queue length (up to 22 at transport hub intersections and 25 at highways) and a 17.9 percent decrease (compared to baseline approaches) in simulated accident rates.”

4. 4

   Beyond static snapshots: gated recurrent units and graph attention networks for smarter multi-agent traffic control.

   Frontiers in artificial intelligence (Mohita et al.)Published Sep 18, 2026Checked Oct 17, 2026

   “The final phase decisions of the signal are derived using 3DQN. The proposed system is extensively evaluated using SUMO simulations on two synthetic grid networks and two real-world traffic networks. Experimental results demonstrate that HC-STGRL consistently reduces average trip times and delays compared to state-of-the-art MARL baselines, achieving up to a 11.3% reduction in travel time and a 14.9% reduction in waiting time relative to the strongest baseline, CoLight, on real-world road networks. Furthermore, the framework exhibits notable robustness, maintaining efficient operation even when 20% of sensors malfunction, a condition that significantly degrades the performance of existing methods: HC-STGRL incurs only a 25.4% performance degradation under 20% sensor failure, compared with 44.7% for CoLight. Importantly, HC-STGRL incorporates built-in mechanisms to prevent major arterial roads from starving adjacent intersections, thereby overcoming a well-documented limitation of conventional throughput-maximizing control strategies.”

5. 5

   Traffic light (Wikipedia)

   WikipediaPublished Oct 15, 2026Checked Oct 17, 2026

   “Traffic lights, traffic signals, or stoplights – also known as robots in South Africa, Zambia, and Namibia – are signaling devices positioned at road intersections, pedestrian crossings, and other locations to control the flow of traffic. Traffic lights usually consist of three signals, conveying meaningful information to road users through colors and symbols, including arrows and bicycle symbols. The usual traffic light colors are red to stop traffic, amber to signal a change, and green to allow traffic to proceed. These are arranged vertically or horizontally in that order. Although this is internationally standardized, variations in traffic light sequences and laws exist on national and local scales. Traffic lights were first introduced in December 1868 on Parliament Square in London to reduce the need for police officers to control traffic. Since then, electricity and computerized control have advanced traffic light technology and increased intersection capacity. The system is also used for other purposes, including the control of pedestrian movements, variable lane control (such as tidal flow systems or smart motorways), and railway level crossings.”

6. 6

   Dynamic Traffic Signal Control for Isolated Intersections: Enhanced SARSA Reinforcement Learning with Expectation Prediction and Eligibility Traces.

   Sensors (Basel, Switzerland) (Gao et al.)Published Jul 9, 2026Checked Oct 17, 2026

   “To overcome these limitations, this paper develops a two-branch cooperative signal control framework integrating Expected SARSA and SARSA(λ) for isolated signalized intersections, constructing a traffic-adaptive Markov Decision Process (MDP) dynamic decision model using real-time traffic flow data. Expected SARSA reduces policy variance and Q-value overestimation by calculating expected action values. SARSA(λ) adopts eligibility traces for multi-step temporal difference error backpropagation to significantly boost sample utilization efficiency and overall model robustness. Experiments cover algorithm comparison, framework generality and hyperparameter robustness. Experimental results reveal that under the framework proposed in this paper, the two improved algorithms outperform the baseline algorithms in quantitative evaluations conducted from three dimensions. Scenario and parameter tests validate its generality and robustness. This work offers an efficient and reliable solution for real-time adaptive signal control of isolated intersections.”