Many habits have changed in response to COVID-19 lockdowns, and that is reflected in large services like Google Maps. The company saw an “up to a 50% decrease in worldwide traffic,” which has required road predictions to be retooled. Meanwhile, Google Maps has “significantly” improved ETAs by leveraging DeepMind AI.
Google Maps predicts “what traffic will look like in the near future” by combining historical patterns with live traffic conditions. This yields predictions that are accurate for 97% of trips.
This process is complex for a number of reasons. For example — even though rush-hour inevitably happens every morning and evening, the exact time of rush hour can vary significantly from day to day and month to month. Additional factors like road quality, speed limits, accidents, and closures can also add to the complexity of the prediction model.
To get more accurate ETAs, the Google Maps team partnered with Alphabet AI lab DeepMind. Results improved by double digits in some cities: Berlin, Jakarta, São Paulo, Sydney, Tokyo, and Washington D.C. Taichung, Taiwan, particularly benefited from this new implementation at 51%.
To do this at a global scale, we used a generalized machine learning architecture called Graph Neural Networks that allows us to conduct spatiotemporal reasoning by incorporating relational learning biases to model the connectivity structure of real-world road networks
DeepMind has previously worked with Google to improve text-to-speech and Android app recommendations in the Play Store.
Meanwhile, given the “sudden change” in traffic due to COVID-19, Google Maps has updated its prediction models to focus on the last two to four weeks of traffic patterns. Data older than that is deprioritized, given how there’s wide variance in which parts of the world have returned to normal.