Philip Koopman, Ph.D., is a professor at Carnegie Mellon University and an expert in the field of autonomous vehicles. The subtitle of his book is Measuring and Predicting Autonomous Vehicle Safety.
I started reading this book with questions. After reading the wonderful book Autonomy by Lawrence Burns several years ago, which promised a new world of autonomous vehicles that would reshape our world, I wondered, what’s taking so long? The Burns’ book promised a future that was and is compelling. Why hasn’t it arrived?
This book answers that question. For autonomous vehicles to become ubiquitous, they must be safe enough for widespread adoption, which begs the question, how safe is safe enough?
Most people answer, when autonomous vehicles are safer than human drivers, that’s safe enough. It turns out, it isn’t that easy. As safe as which human drivers? Under what circumstances? And how will we know?
Most of this book discusses these questions. Like many technical fields, there is very specific terminology. Measurement of the data is very complex. Doing testing on public roads and then comparing with a control group can’t be done.
The author is an engineer and the book, while interesting, is a very difficult book to read. About halfway through, it just became painful, and I quickly skimmed the second half. The point had been made already: This is way harder than we imagine.
What seems to be happening is that autonomous cars are often frustrated by humans doing stupid things. Compound that with societal values that seem to be holding autonomous vehicles to aviation standards rather than landbound vehicle standards. With aviation standards being, essentially, zero deaths.
That’s simplifying it way too much, but that’s where the industry is. Nevertheless, progress continues and Waymo, probably the leader, continues to roll out its autonomous taxi service. Tesla continues to tout its system, but it is far from perfect.
The most advanced systems still require the ‘driver’ to be ready to assume control of the vehicle on a moment’s notice in the worst of circumstances. That just isn’t practical. The systems need to be good enough that the driver can sleep while the vehicle is underway. That’s why this is taking so long.
Do I recommend this book. Oh, hell no! As I said above, it is painful to read. It is a 352-page engineering paper on a very technical topic. However, if you’re an engineer or just enjoy this sort of thing, have at it! Otherwise, the conclusion is above.
For those of us looking forward to a world with safe and efficient autonomous vehicles serving us, I hope I live long enough to see it.
No comments yet.