How AI Can Help Enforce Traffic Rules Better

Posted by Sukant Khurana in Sci-Tech, Specials
February 20, 2018

Co-authored by Saikat Bhattacharyya

Things have changed since 1912, when Lester Wire developed the first traffic light, hoping that it would signal pedestrians and vehicles alike to follow and abide by the newly developed rules and regulations of road transport. While the signalling did turn out to be successful, the other half did not.

In today’s era, traffic has been a regularly developing issue, more so a dangerous one that all mainstream cities which are highly inhabited are facing. Every day, the population increases exponentially, whereas the road sizes and the personnel managing them do not, converting cities of lakes, gardens and calming atmosphere like Bangalore to now be an over-burdened city that cannot handle the seething empowers.

So have the traffic experts fallen asleep, that has led to this mess?

No, that is not the issue everywhere. For example, as stated in the article “How Mizoram’s Motorists Follow One Simple Rule to Avoid Traffic Jams & Ensure Congestion-Free Roads” we can infer from the video how better traffic is managed there than the rest of chaotic India, which involves the four-wheeler’s and the two-wheeler’s having their own side where they are supposed to drive in. There is no rush, as all the vehicles maintain a steady speed, without overtaking one another, which is enough for any motorcycle from another lane to cautiously come over to the right-lane, which belongs to the motorcyclists. Such a display of understanding is one way we can tackle this global issue.

But, can this really be implemented everywhere?

Let us take the case of India, with one of the highest population densities and lack of traffic discipline. In cities like Delhi, Bangalore, Kolkata, Mumbai and many more, with people from different cultures, the communication and co-operation gap remains and everyone has different ways of thinking, and different issues to solve. To develop a centralized system and solve this issue, we would require more power than simple human police management or education for people.

Let us think for a minute about the issue here, because it’s bigger than just “traffic management”. We are talking about Criminal Law Enforcement which has very well been lacking in the cities today. Just recently, we read a news article — “Car stolen from outside Wawa with two kids inside”. A woman had left her car in the parking lot with two children inside to quickly run in a store, and when she arrived five minutes later, both the car and her children were gone.

In a recent traffic rules enforcement survey, 83% of the people from Gujarat do not bother following the rules of traffic, like wearing a helmet, which would in time lead to them “burying their heads in the sand”. In 2016, 8,136 people from Gujarat lost their lives in street accidents, huge numbers of the fatalities being a consequence of the casualties not wearing helmets or safety belts.

While having a bigger population is disadvantageous in some fields, that might not be the case for the field of traffic management. Till recently, we were trying to solve the traffic management problem with human power, and while we have been successful in some places, we have mostly failed. However, taking an approach with Big Data and Artificial Intelligence in this field can provide us with different results.

Photo by Richard Lee on Unsplash

We humans believe that we are individually rational, however in scenarios of collectiveness irrationality seems to play the major role, and traffic management issues are the proof. A single selfish overtaking for personal benefit or perhaps even a brake merge and there it sends out a cascade of brakes, forming a chain and engulfing all of the traffic in it. However, Artificial Intelligence (AI) can help.

In an article by Michael Byrne titled “New AI Algorithm Beats Even the World’s Worst Traffic”, he wrote that only 10% of the cars would have to be connected to the network for this to work. But how exactly is this working?

Judgement affects driving significantly. People are greedy, and they tend to do good mostly for themselves rather than the whole mankind. This is one of the reasons why traffic jams happen in the first place. But, what if one day we completely remove the human mind from this system, and introduce another complicated network, which would be connected to each and every car running on it, making decisions optimal for all of them combined, providing the best solution with every fraction of a second? Won’t that be amazing?

Most of the times accidents happen because some people fall asleep while driving, under the influence of alcohol or drowsiness. However, suppose we move the entire system to be automated such that it takes proper decisions based on certain circumstances and learns based on previous experiences, these problems would never arise.

Artificial intelligence needs Big Data to make judgements. Suppose, there is an autonomous car running on an AI, which can plan its route and accordingly move safely in the streets. How will this autonomous vehicle make judgements, like where to turn left from, how to avoid collision, what’s the status of the traffic signal, should it stop or keep moving or speed up depending upon the traffic density on the road? For these scenarios and many others, the AI needs input of lots of data — mainly from the sensors, from the network it’s a part of, a feedback mechanism, and many more. All these data are not in the same format, as we’re taking in data from different sources, and hence cannot be processed in the same way. To analyze such amounts of unstructured varieties of data, becomes a Big Data problem.

As the world grows, the data grows too, and AI is a technology which empowers connected machines to learn, evolve and grow, and this can only be achieved by “reiterating and consistently updating the data bank through recursive experiments and human intervention”.

Let’s look at some of the successful attempts so far. is an autonomous self-driving vehicle software startup using deep learning. It concentrates on the “software necessary for cars to understand their surroundings and make decisions”. With deep learning, the working of the human brain can be copied in order to produce programs that can solve complex problems.

But how is all of this related to crime at all? Let’s look at the potential of AI when combined with big data. We are dealing with a traffic that does not follow rules, and this is because we give them the option of following or not following the rule. What if we were to remove that option, and make everything autonomous, so that the people would just have to accept what is happening, with systems such highly detailed that chances of traffic problems would tend to zero. Now such a situation might only be a dream for now, but not for long. Some might even consider it a nightmare.

You might ask, how can we force people to just use autonomous cars, what if some of them don’t agree and they want to drive the cars on their own. Yes, that might be the case. People mostly don’t accept changes quickly. It has to be proven to them that the new system that if build, is better than the existing system, such that it would prevent all sorts of unforeseen incidents and reduce the overall crime rates. It might not just be for driving, with AI and big data, we can make all parts of a vehicle intelligent. What will happen then is, suppose someone is trying to break into your vehicle and they break the window, your car would automatically send you a warning signal or some kind of message or would activate some sort of emergency protocol which would stop the criminal from achieving what he/she wants.

The cars would be trained to run on all types of terrains with proper reinforcement learning hence eliminating the hassle for us to drive it.

Now imagine what could be achieved with such a system. If we can build autonomous police cars or bots, we could stop almost all crimes throughout the city. That is far away in a utopian or dystopian future but an interesting thought. With proper data mining, we could analyze the crimes that have happened over the past few years and are happening right now in real-time, a system could be constructed to analyze and find what types of crimes happen at what regions in the city. Allocation of police force to appropriate locations is not as futuristic as our above thought and something that can be easily implemented. Here is an idea start-ups should grab or we might do it ourselves.

It’s not just crimes that this system can be used in, we could integrate this AI into an ambulance. Thousands have died over the past years due to Ambulance delays. What this AI could do is, increase the priority of the ambulance when it’s on the street, and hence all the other vehicles around it will slow down and efficiently position themselves to construct an unobstructed path for the ambulance to reach the hospital way quicker than usual. This would save hundreds of live every year in crowded cities like Kolkata, Mumbai, Bangalore and etc.

The next generation is coming, whether we want it to, or not. The population increases exponentially every month, and there is only one way to optimize our survival — through intelligence, even if it is artificial. A lot of us won’t accept these changes immediately, because of what we’ve been accustomed to. But we all have to agree — humans are unreliable. Now I’m not saying machines are not, but changing the behaviour of all the humans and changing the behaviour of few machines — we both know which one of them is feasible.

This post was first published here.


Mr Saikat Bhattacharyya is a researcher working with Dr Sukant Khurana on application of artificial intelligence and big data in fighting crime.

Dr Sukant Khurana runs an academic research lab and several tech companies. He is also a known artist, author, and speaker. You can learn more about Sukant at or and if you wish to work on biomedical research, neuroscience, sustainable development, artificial intelligence or data science projects for public good, you can contact him at or by reaching out to him on LinkedIn.

Featured image source: Igor Ovsyannykov on Unsplash