Today at CES, GM announced that it’s “exploring” the use of cameras on board its cars to automatically build high-definition mapping data — the kind of data needed for fully autonomous driving. The company would use the cellular connection in its OnStar modules, already built into most vehicles it sells, to upload a relatively low-bandwidth stream of differential mapping data back to the mothership, updating a master database of information like lane markings and precise location information.
GM could crowdsource your car’s data to make better maps for self-driving cars
High-def maps are key for autonomous cars
The project is in collaboration with Mobileye, a large supplier of semi-autonomous sensors and systems to the automotive industry; GM already uses Mobileye sensors for features like lane departure warning and collision alert, so firing up a system like this in new cars — which can generate the data using existing camera systems — is a relatively minor undertaking. Mark Reuss, GM’s product boss, says it could be integrated into new model launches as soon as “later this year.”
Anonymized, crowdsourced mapping data is seen as a possible boon for autonomous systems, which require higher-definition roadmaps than humans do in order to operate correctly in a broad spectrum of scenarios. In rolling out its Autopilot feature, Tesla announced that it would be collecting mapping data from its customers; for GM, which dwarfs Tesla, high-quality mapping from customers in the field could roll in and be continuously refreshed far more rapidly.
Follow topics and authors from this story to see more like this in your personalized homepage feed and to receive email updates.
