TheMassachusetts Institute of Technology (MIT),in the United States, and the Qatar Computer Research Institute (QCRI) have developed a GPS system with Artificial Intelligence (AI) to update the maps and give more precise directions to drivers, such as locating them in the correct lane of the road.
The developed model, called‘RoadTagger‘(‘ road label ‘, in English) uses AI and satellite imagery to label the characteristics of routes in places with limited map data, which will allow drivers to provide more details about their roads.
MIT says that creating detailed maps is quite expensive and time consuming. This is usually done by large companies, such as Google, which sends vehicles with cameras to capture images of the roads.
He also points out that combining these images with other types of data could create accurate and updated maps. Therefore, it indicates that it would be easier to do it using machine learning models with satellite images, since they areEasier to obtain and update.
‘RoadTagger’ uses acombination of neural networksto automatically predict the number of lanes and types of roads, be it a street or a highway.
When testing the system on blocked roads on digital maps of20 US cities, ‘RoadTagger’ counted the number of lanes with 77 percent accuracy.
In addition, ‘RoadTagger’ differentiated the types of roads with 93 percent accuracy, according to a document published by the MIT and the QCRI.
“The most up-to-date digital maps come from the places that most concern large companies. If you are in places that do not care much, you will not have advantages regarding the quality of the map,” said Sam Madden, professor at theDepartment of Electrical Engineering and Computer Science (EECS).
“Our goal is to automate the process of generating high quality digital maps, so that they can be available in any country,” he added.
TheresearchersThey have also claimed that they are planning to use this system to predict other features, such as parking spaces and bicycle lanes.