If you own a smart device, you must know about Google Maps. It is our 'Go to' app for searching destinations and directions.
According to Google's blog, people drive over 1 billion kilometers in over 220 countries every day.
And they do it with Google Maps.
Google Maps are known for its accurate traffic predictions and expected time of arrival (ETA). Ever thought, how is that happening?
Well, it is the magic of DeepMind's AI tool.
It uses machine learning tools to study how much time it takes for a person to arrive at a destination. Then it stores the data to show it to a different person wishing to visit the same place.
Along with that, it uses an algorithm to predict the ETA and live traffic on a particular route. With DeepMind’s AI model, Google Maps made sure we get only the accurate data!
DeepMind is an Alphabet research AI lab. Its research team tests and studies various AI programs. If you check their blog, you will find works on AI games, health, and more.
It makes one wonder how many things can Artificial Intelligence do!
The team of engineers, scientists, AI experts, and more believe that Artificial Intelligence could be humanity's most useful invention. So, they study various things that can make people's life easier.
Last year, Google got DeepMind onboard on Google Map project. The main goal was to make Google Map prediction top-notch and provide better service to users.
For traffic prediction, the AI collects data from various sources and feeds them to the algorithm. It thus stores the data and learns how to use it to predict the traffic flows.
It collects several data from our phones such as live location, past traffic memory, construction sites location, road size and more.
Additionally, it uses public information to find the location of road damage, size, and direction, road quality, and more.
For example, the AI algorithm predicts the chances of traffic jams if there is an accident. It receives data of any damaged road from the local Government. Then uses it to predict the traffic around it.
It collects all these data from various sources and feeds them into DeepMind's neural networks. The AI studies and predicts the traffic. However, during the COVID-19 lockdown, there was a change in the traffic pattern.
The scientists added new data into the neural networks to predict traffic during lockdown periods.
Google Maps product manager Johann Lau said, “We saw up to a 50 percent decrease in worldwide traffic when lockdowns started in early 2020. To account for this sudden change, we've recently updated our models to become more agile — automatically prioritizing historical traffic patterns from the last two to four weeks and deprioritizing patterns from any time before that.”
The AI algorithm divides the maps into several segments. They are called "Supersegments".
There is no distinct size of these supersegments. However, they change with the traffic. One segment collects about terabytes of data.
So, you may assume the size!
The supersegment is a group of adjacent streets that have the same traffic volume. Each group has its neural network which does traffic prediction.
It uses a specific kind of neural network called ‘Graph Neural Network’. Google considers this as the most suitable AI algorithm for traffic prediction.
Before partnering with DeepMind, Google already had a 97% accuracy rate of the expected arrival time for a destination.
For ETA predictions, DeepMind uses the Graph Neural Networks. The accuracy rate of the expected arrival time was up by 50% in major cities around the world.
Using the traffic data from various segments (as mentioned above), it collects current traffic information. It won't be able to show future traffic like condition of traffic after 20-50 minutes.
For showing future traffic conditions, it uses historical data from various sources. The AI algorithm combines the current traffic and the historical data to predict ETA to a destination.
For example, the algorithm has data of the rush hour of a route in Sydney for a random Monday. But it turns out that a Monday is a holiday but it will show it is a rush hour for future predictions. So, it will also show that the ETA is much longer than the reality.
However, the current traffic and the ETA will be accurate on-road, thanks to the Graph Neural Network. Moreover, data on road quality, accidents, and more are added to the model to raise the accuracy rates for ETA and traffic prediction.
Let's say you are heading towards the airport. When you leave your house, the traffic shown in Google Maps is normal and free-flowing. There is no indication of any disruption or bad roads.
However, as you are on the move, Google Maps will start predicting live traffic details. It will tell you the exact condition of traffic as you move through the routes.
So, 5 minutes into the drive, it predicts that you may get stuck in a traffic jam. It also tells how much time it might take to get out of it. Therefore, it starts suggesting routes that can help you reach the airport on time.
Thus it will help you avoid the traffic and also reach on time.
Google Maps partnering with DeepMind's AI lab has brought a revolutionary change in the application. It made the predictions 50% more accurate. Now anyone can learn about traffic jams and ETA without relying on Radio FM or assumption.
The Graph Neural Network delivers accurate data for the user. Moreover, it will also suggest roadways that can lessen the travel duration. So, don't forget to use Google Maps the next time you are on the road.