It seems like the autonomous vehicle industry agrees that they will require high definition maps in order for the vehicle sensor packages to compare the derived location with what is supposed to be ground truth. The only problem is, nobody knows what the specification of a high definition map is supposed to be.
Let’s take one rather large example. Ushr claims they are producing high definition maps of 125,000 miles of interstate in the U.S. and Canada for General Motors. These maps from a single source could be a tremendous example of collect once, use many. Ushr claims the mapping has been done to “under 10cm absolute accuracy.” They do not say whether that is vertical or horizontal, or with what statistical uncertainty.
Genesys builds HD maps. They are based in India. Genesys describes the HD map has having high-precision navigational geometry and 360-degree panoramic imagery leading to a unique “fingerprint” of every stretch of the road. That’s as close as they come to a specification.
In the old surveying days – steel tapes and plumb bobs, the saying used to be “close enough for highway work,” as compared to vertical building construction where tolerances were generally tighter. Well that is all changed in large part because of GNSS and the emergence of autonomous vehicles.
It would seem that ASTM, or SAE would be interested in working on a standard for the HD map. That way the autonomous vehicle industry would be able to rely on the data in the HD map to meet their navigational requirements. At the very least there should be a horizontal and vertical accuracy spec at 2 sigma.
What do you think?
Note – If you liked this post click here to stay informed of all of the 3D laser scanning, geomatics, UAS, autonomous vehicle, Lidar News and more. If you have an informative 3D video that you would like us to promote, please forward to editor@lidarnews.com and if you would like to join the Younger Geospatial Professional movement click here
Seems to me that for autonomous vehicles, HD maps only need to be accurate enough that the vehicle can identify its surroundings. This should only require accuracy to within a few cm – mobile mapping lidar should be accurate enough. It would rely on the geometric relationship between objects, SLAM style. However, things change on the ground quickly. Even Google Maps Navigation falls apart when a big construction project appears at an intersection. How it deals with cases like this is the real challenge.