Alphabet’s Waymo, widely considered to be in pole position in the race for fully automated vehicles, has publically released a set of data that is now available to third-party researchers. The data consists of 1000 individual 20-second driving segments involving its cars on the test drives that have now covered millions of miles of road. In doing so, Waymo is following a trend already established by driverless technology competitors including Lyft, Uber, GM’s self-driving unit Cruise and Argo AI.
While it’s only a small sample set that has been released, Waymo believes that it will be of particular value to researchers because different elements such as bikes and pedestrians have been labelled and other signals also marked. Doing this themselves would be very time consuming. The dataset is also deeper and more nuanced than others that have been released because it includes not only data from the vehicles’ cameras but also readings from radar and lidar sensors.
The data comes from drives made in California’s San Francisco and Mountain View, Phoenix in Arizona and Kirkland in the state of Washington. Waymo has already begun to pilot a limited commercial ‘robotaxi’ service in Phoenix. Routes are fixed and vehicles still also have a human ‘safety’ driver ready to take over the controls.
Many voices in the technology community have called on the various companies developing self-driving technology and conducting test drives to make more of the data they are collecting public. Because reliable self-driving algorithms able to make autonomous vehicles safe need to ‘learn’ from huge volumes of data to compensate for their lack of human-like intuition, there is obvious value in pooling data.
However, that data was also extremely expensive for the companies that own it to gather and is a key part of their competitive advantage when it comes to developing technology that might one day be approved by regulators. Which leads to a conflict of commercial interests it will be difficult to overcome in the interest of the greater good. A full data-sharing memorandum would almost certainly mean self-driving vehicles would more quickly become a reality for all of the companies pursuing the technology. But it would also close the gaps separating companies and having a head-start on the competition could be worth billions to those companies.
Commenting on the move, Drago Anguelov, Waymo’s head of research said:
“To me, it’s a bit of a labor of love. I think that also creating such a data set is a lot of work. And it takes many months to label the data, ensure that all the relevant parts are to the highest standards that one expects, making sure that the right utilities are available for researchers to be able to make progress without being hamstrung.”
The MIT Technology Review publication opined that while Waymo deserves some credit for the release of the dataset:
“if the industry wants to overcome concerns about autonomous vehicles’ safety, the businesses in it will have to become far more transparent about what they’ve learned”.