Case for using AI for Traffic Management

If you have ever driven a vehicle in India, there is no way that you haven’t been stuck in traffic. In certain cities, it is a way of life. No one likes it. Traffic adds a lot of stress to our already stressful life.

In India traffic is managed through three things

  1. Traffic Police
  2. Traffic Signals
  3. Road Signs

It is extremely rare to see Indian drivers just follow road signs. No one cares about speed limits or lane driving. It makes India one of the most dangerous places to drive. Every year almost a hundred fifty thousand people die in road accidents. [1] Traffic police do the best they can with the resources they get and the training that they receive.

That leaves us with traffic signals.

In India traffic signals are mainly two types: a) One with preset timers, and b) One which are manually controlled by traffic police. There are pilots being done for other types of traffic signals but nothing has changed on a larger scale.

There are certain challenges with the present system:

  • Preset timing signals result cars piling up at peak hours and cars having to wait even when there is no traffic during lean hours.
  • Preset traffic signals are not updated as the areas around the signal develop and traffic patterns change.
  • Traffic policemen can only gauge/guess how to manage traffic based on how far they can see. They rarely use information from adjoining areas in making decisions about managing traffic.
  • Manually controlled traffic signals will only use the judgement of the traffic policemen in regulating traffic. Its effectiveness would depend entirely on the policeman controlling it.
  • The system cannot adopt to emergencies or special circumstances.


Creating an Artificial Intelligence (AI) system that manages traffic.


a) AI can be taught easily and it improves continuously – Preset signals are based on logic. Wait time is calculated based on the number of vehicles passing through a crossroad and the direction that it is going towards. AI can be taught the logic model. It can use it to manage traffic. It can also automatically notice changes in traffic patterns and can make modification to signal timings. It can adapt during special circumstances for example public holidays, an events which might impact traffic flow like concerts or political rallies. Given that it is basically a computer system, it can be ‘taught’ about any changes in rules that need to be made like new one way streets or no car streets. It will build it into the system and find the best possible way to manage traffic with that information.

b) AI can see the whole picture – Traffic policemen make decisions based on the status of a single crossroad. This decision is mostly a gut decision rather than a decision which takes into account multiple parameters. AI will be one system managing all the traffic signals across the city. It would be able to see which areas are more crowded and facilitate the diversion of traffic from there to areas where there is less traffic. If it ‘sees’ an ambulance at one crossroad, it can ensure that the ambulance gets as much space as possible to reach a hospital quickly. If it ‘sees’ an accident which has blocked the road, it can circumvent that part of the road and ensure that all vehicles don’t get stuck in a jam because of the accident.

c) Traffic police can be focus on enforcement – Traffic policemen currently have two responsibilities. One is to manage traffic and second is to enforce traffic rules. Now that AI would manage traffic, they can focus solely on enforcing traffic rules. Certain cities in India have already started using cameras to capture photos of people not following traffic rules and charging them fine. This reduction in responsibility for traffic policemen will mean that their training can become targeted to one skill. AI cannot check for PUC or for documents of ownership or ask if the person has a legitimate reason for not wearing a helmet. We would all be better served if policemen focus on the more human aspect of the job.

d) Collect accurate data for traffic management interventions – Traffic management interventions sometimes include infrastructural changes to be made. It is important to collect information about the current traffic flow and the changes in patterns to be able to decide which intervention is the best to choose. AI can help us collect this seamlessly. It can count the number of vehicles plying on the road at different points of time. It can help us find out which routes are more often used by vehicles and what kind of vehicles. Ofcourse all this can be found by humans as well but AI would be able to do this along with doing its job of managing traffic.

e) Overall security can be improved – AI would be able to ‘watch’ all the roads and hence it would improve security on the streets at all times in the day and night. Road side thefts, harassment, vandalism and other crimes that happen on the streets can be monitored. It would make our streets safer.

Overall, using AI system to manage traffic would reduce human errors, would reduce the dependency on individual human judgement, reduce work load for an already stressed traffic police workforce, and increase management efficiency by many folds.

It would need significant investment and expertise but all the investment would be recovered by saving invaluable hours that people spend stuck in traffic. Technology should be used to solve problems that we currently face and not just to make the problems more bearable.

Add a Comment

Your email address will not be published. Required fields are marked *