How many times have we waited for the green traffic light thinking, “This has to be longer than necessary”? Traffic lights often feel outdated, inefficient and frustrating — but what if they didn’t have to be?
Traffic signals are a set of visual aids that use coloured lights to regulate the flow of vehicles and pedestrians, by managing right-of-way for multiple directions of traffic flow at a time. Well-coordinated signals can improve travel times, enhance road safety and optimise traffic throughput.
However, when traffic signals are poorly timed, they can have the opposite effect — causing congestion, increasing emissions and even making intersections more dangerous. In some cases, rigid timing cycles that don’t account for real-time conditions, like an unexpected surge in traffic, can lead to inefficient flow, exacerbating delays instead of reducing them.
Cities across the world, including many Indian cities have been experimenting with artificial intelligence (AI)-based dynamic traffic signals. These systems utilise AI, machine learning and real-time data from sensors, cameras and multiple connected devices (internet of things or IoT) to optimise traffic flow. Unlike static systems, these signals adapt timings based on live traffic conditions, reducing travel time and pollution.
Bengaluru launched its Adaptive Traffic Control System (ATCS) in May 2024 to ease congestion. Goa is set to instal dynamic traffic signals at 91 locations to improve traffic management and curb violations. Meanwhile, Gurugram’s Metropolitan Development Authority is upgrading 111 existing signals into smart signals and plans to instal 32 more, including pedestrian signals, in the next phase.
Let’s assume that a simple four-arm, two-phase traffic signal system alternates right-of-way between two competing traffic streams, such as north-south and east-west movements.
In Phase 1, north-south traffic gets a green light while east-west traffic stops; in Phase 2, the flow reverses. Each phase consists of a green time (when vehicles can move), a yellow time (a short warning before stopping) and a red time (when traffic must wait), sometimes with an all-red clearance period to ensure the intersection is empty before switching phases. A phase is a specific period within a signal cycle when a particular movement or group of movements is given the right-of-way.
A traffic signal controller governs the phase duration.
In a traditional system, the signal controller has "fixed-time" with pre-set cycles. For example, if the cycle time is 120 seconds, the total time for one full sequence of signal phases (green, yellow and red) is 120 seconds, including all directions of traffic movement. In our example, one phase has a 55-second red light, a 5-second yellow light and a 60-second green light, making the total cycle time 120 seconds.
The timing can be adjusted for peak and non-peak hours, with traffic engineers using traffic volume data to decide on durations. However, these timings don't adapt in real-time and require manual updates based on surveys and analysis. This process is cumbersome and often leads to outdated cycle times, as it demands frequent manual updates across multiple intersections.
Static vs dynamic traffic control systems in a simple two-phase traffic light module
In an AI-based system, inductive loop sensors, cameras or radar are used to detect vehicles waiting at the stop line. The sensors are embedded in the pavement and detect the presence of vehicles through magnetic field disruption, while radar sensors use radio waves to gauge vehicle speed and distance. Cameras with computer vision (such as automatic number plate recognition) help monitor traffic and pedestrian flow, adjusting signal timings based on visual data.
These sensors feed real-time information into an AI that analyses incoming data using machine learning algorithms trained on past traffic patterns. It identifies congestion hotspots, unusual traffic surges, or empty roads and calculates the optimal signal timing to improve flow. The AI also prioritises emergency vehicles, public transport, or high-density lanes based on set policies in the system.
Finally, the AI plugs this analysis into the signal controller, which determines the optimal green, yellow and red light durations. AI systems learn from past decisions, improving their predictive accuracy over time.
There is no pre-set cycle time, or a phase time. So, if east-west has fewer vehicles at any particular time of day, its green time may reduce automatically, allowing the north-south phase to start earlier. If no vehicles are detected in a lane, the system may even skip that phase.
AI-based dynamic signals or “smart signals” are economical, attributable to congestion reduction and emission abatement. Smart signals can save cities $277 billion dollars in four years globally, according to a study by Juniper Research, a UK-based market research company for the digital marketing sector. Out of this, more than 95 per cent will be due to congestion reduction, which in turn reduces productivity losses and emissions.
Pittsburgh, Pennsylvania deployed the adaptive traffic control system called Scalable Urban Traffic Control (SURTAC), which resulted in a 25 per cent reduction in travel time, a 40 per cent decrease in wait times at intersections and a 21 per cent drop in vehicle emissions.
In Fremont, California, an AI-based system is used to communicate with ambulances, snowplows and public transportation. These vehicles have access to smarter, safer routes through a platform that cuts the number of red traffic lights they meet, decreasing average travel time across the city from 46 to 14 minutes.
Several other examples can be seen worldwide, such as the Sydney Coordinated Adaptive Traffic System (SCATS) in Sydney, Australia, which is now also in use in other global cities such as New Zealand, Hong Kong, Shanghai and Guangzhou to name a few; the Yutraffic FUSION system in Prague, Ellwangen, Germany and London, UK; the Hikvision system in Xi’an, China and so on.
In Bengaluru, the ATCS has led to noticeable improvements. By January 2025, authorities reported an increase in average speeds by at least 16 per cent and up to 61 per cent on certain central stretches. The city, ranked among the world’s most congested in 2024 by Dutch location technology company TomTom, has long struggled with gridlock, seeing peak-hour speeds drop to just 18 km / h.
The release of the ChatGPT series between 2018 and 2024 by tech giant OpenAI marked a turning point in AI development, before which the AI research was largely confined to niche applications with limited impact. ChatGPT played a crucial role in democratising AI, making advanced language models more accessible to the public.
This widespread accessibility spurred innovation across industries, leading to the development of specialised AI applications in fields such as healthcare, finance, transportation and customer service.
AI trains on vast amounts of data, enabling models to learn complex patterns and improve decision-making across various applications. Ironically, the very data that fuels AI has sparked debates over ethics, bias from historical inequalities in datasets and privacy concerns.
Therefore, while AI-powered smart traffic signals offer significant benefits, they also come with several riskss. Data privacy and surveillance risks, for instance, may arise as these systems rely on cameras, sensors and license plate recognition, potentially leading to unauthorised tracking or misuse of personal data. Data collection should be anonymised, encrypted and strictly regulated, with clear policies on storage and access to prevent misuse.
Then, bias in AI algorithms can result in unfair traffic management, where certain areas receive more efficient signal control than others, reinforcing infrastructure inequalities. AI models should be trained on diverse datasets, regularly audited and include human oversight to ensure fair traffic management.
Cybersecurity threats are another major risk, as hackers could manipulate traffic flow or disable signals, leading to disruptions and accidents. Traffic management systems should be secured with encryption, multi-layer authentication and real-time threat monitoring to prevent cyberattacks.
Finally, ethical and legal issues remain unresolved, particularly in determining accountability when AI malfunctions cause accidents. Clear legal frameworks, accountability guidelines and fail-safe mechanisms should be established to define responsibility in case of AI failures.
India’s Union Budget 2025-26, presented by Finance Minister Nirmala Sitharaman on February 1, 2025, underscored a strong commitment to urban transformation. The new Rs 1 lakh crore Urban Challenge Fund, for instance, aims to make cities more efficient and livable by fostering innovation in infrastructure and governance.
AI-powered traffic signals, which can dynamically adjust to real-time congestion patterns, fit well within this framework, addressing the long-standing inefficiencies of static signal timings.