Autonomy is an account of the development of driverless car technology from the earliest days of the DARPA Grand Challenge through the present. The author, a former GM executive and current advisor to Waymo, is decidedly optimistic about technology from the first page of this book to the last. Consequently, he seems to understate the policy controversies and social upheaval that could accompany the shift to autonomous vehicles.

Another criticism of the book is that it is too focused on the U.S. passenger vehicle market. Autonomous vehicles are already commonplace in industrial applications, such as in Australian mining operations. This is a far cry from driving safely on public roads, but autonomous driving in the U.S. is also quite different from driverless operation in the “hyper-multimodal” crowded urban centers of India or major African cities.

The strongest aspect of the book is its clear identification of major schools of thought amongst the developers of autonomous vehicles. One such divide is whether to use custom hardware or off-the-shelf materials (p. 54). Similarly, companies that have made substantial progress on autonomy systems also debate whether to partner with automobile manufacturers or develop new vehicles from the ground up (pp. 257, 265).

The authors also explain for a general audience the important role that maps play in autonomous driving systems. The decision to use maps was one of the critical “hacks” that facilitated the success of Carnegie Mellon’s entry in the first DARPA challenge:

If Red Team members could give Sandstorm an accurate map of its surroundings before the race, they could remove a time-intensive step from the computational task…. The team had assumed they were trying to build a robot that could sense the world so well, it could discern a road in the desert and navigate it safely for 150 miles. Using maps meant that the robot could be told in advance where the road was, and how to drive it. The method had the potential to allow Sandstorm to travel much faster than it otherwise might. (p. 33)

The development of Google Maps–and especially its Street View component, with images of virtually every road in markets where Google operates–was also a major contributor to the early success of the Chauffeur project (which was eventually spun out of Google as Waymo during the Alphabet re-organization):

Google’s development of the Street View service stands as a carographic achievement unmatched since the days of Vasco de Gama and Magellan….

Chaffeur’s approach to the safe and reliable deployment of self-driving cars required highly accurate maps of the roads its vehicles navigated…. 3-D map data provided the Chauffeur vehicles with the ability to locate themselves in the world, just as preexisting maps did for Thrun’s Minerva robot tour guide and Levandowski’s pizza-delivering Pribot. (pp. 172-73)

Although there is potential for the development of mapless or “maps-light” autonomy systems in the future, for now mapping provides an incredibly useful prior about the state of the world: “[G]etting to full autonomy, the kind that allows you to think it’s a good idea to make a car without a steering wheel or brakes, requires really good maps–a high resolution 3-D scan of every road the self-driving car is going to ride, ever.” (p. 285)

Despite its shortcomings listed above, this book is a useful history of autonomous driving systems. It is a useful reference for both the general public and AV practitioners to understand how this technology got its start and became part of the “adjacent possible” in just a few short years.