How many times do we use Google Maps instinctively to find the shortest or best routes to a particular destination? Probably countless times! This is what everyone else does too. Millions of people use Google Maps daily during their commute, whether they are walking or driving a vehicle.
According to Google, users in over 220 countries and territories have logged more than 1 billion kilometers traveled using Google Maps. When you get behind the wheel of a car or motorcycle and turn on the navigation system, a number of details become immediately apparent. These details include:
The direction to take
The traffic density along your route
The amount of time it will take to get there
The estimated time of arrival (ETA).
Although it seems effortless, a great deal of work goes into providing this data in such a short amount of time. So how does Google's AI do this?
When users' location data from Google Maps navigation is added up, it can show how traffic flows around the world. The data is useful for determining the current traffic situation, such as whether or not a traffic jam will delay your trip, but it doesn't give any indication of how the traffic will be in 10, 20, or even 50 minutes.
This is where we see the greatest impact of cutting-edge technology. Google Maps analyzes past traffic patterns for roads over time to estimate traffic with advanced machine learning techniques and a little bit of history to predict what traffic will look like in the near future.
Google and DeepMind, an Alphabet AI research facility, recently collaborated to increase the precision of their traffic forecast tools. The threshold for the accuracy of the ETA estimates is already pretty high. Over 97% of trips have been estimated correctly.
By collaborating with DeepMind, Google was able to further reduce the number of inaccurate ETAs using Graph Neural Networks, a machine learning architecture. By using this method, Google Maps is better equipped to anticipate whether or not you will be impacted by a slowdown that may not even have started yet!
Johann Lau, Product Manager, Google Maps states the following in a blog post:
Google Maps revealed an intriguing trend: global travel patterns have drastically changed since the COVID-19 epidemic. Google has also recently revised its models to become more flexible, automatically prioritizing historical traffic patterns from the last two to four weeks and deprioritizing patterns from any other period of time. This is done to account for this quick change.
Google Maps chooses driving routes using forecast traffic algorithms. If the AI predicts that traffic in one direction will become congested, it will offer you a less congested alternative right away.
While doing that, the condition of the roads is considered too. Factors such as, "Is the road paved?" or muddied with gravel or other debris affect the efficiency of the driver. Google also considers the size and directness of a route because it is frequently more efficient to drive down a highway than it is to take a smaller road with several stops.
Two other sources of information are important to make sure we recommend the best routes:
Authoritative data from local governments
Google Maps can get official information about speed limits, tolls, and which roads are closed because of COVID-19 or construction.
Real-time feedback from users
Incident reports from drivers help Google Maps swiftly display information about restricted lanes, nearby construction, damaged vehicles, and other roadside obstructions.
Both sources are also used to help us comprehend when road conditions change unexpectedly owing to mudslides, snowstorms, or other forces of nature.