Sep 04, 2020 12:00:00

Accuracy of Google Maps traffic information improved with DeepMind's machine learning tool

Google Maps is used by more than 1 billion kilometers every day. In particular, the estimated arrival time of the navigation function has been factored in due to traffic congestion, and the error has been considerably suppressed. In order to improve the accuracy, it seems that machine learning using artificial intelligence / DeepMind is being performed.

Traffic prediction with advanced Graph Neural Networks | DeepMind

https://deepmind.com/blog/article/traffic-prediction-with-advanced-graph-neural-networks

Google Maps 101: How AI helps predict traffic and determine routes

https://blog.google/products/maps/google-maps-101-how-ai-helps-predict-traffic-and-determine-routes/

Google Maps is used by more than 220 countries and territories to travel more than 1 billion km daily. The navigation function displays the estimated time of arrival at the destination, but when this time is displayed, traffic congestion information on surrounding roads is processed on the back side.

According to Google and DeepMind, accuracy is improved by combining elements such as traffic information collected anonymously from Android terminals, construction information and restriction information provided by local governments and authorities, and quality information on the road itself. I heard that In particular, unpaved roads and roads covered with gravel, dirt, and mud may be difficult to drive, so it is rarely included in the recommended route.

According to Google's example, the use of DeepMind tools has improved the accuracy of traffic information around the world, such as 51% in Taichung, 43% in Sydney, 37% in Osaka and 34% in Orlando. I will.

In addition, due to the influence of the new coronavirus, the traffic pattern has changed dramatically worldwide, and when the lockdown was done in various places in the beginning of 2020, the traffic volume decreased by up to 50%. thing. Since this reduced traffic information is a special situation, Google has updated its forecasting model to automatically prioritize the history for the past 2-4 weeks and lower priorities.