Waymo's official blog on its self-driving cars, covering mileage, disengagement data, mapping, machine learning, trucks, and its legal dispute with Uber.
HN commenters cite it for technical details on autonomous driving, such as disengagement rates, hand signal recognition, emergency vehicle sounds, and the retired Firefly prototype.
Depending upon your statistical sources, US traffic fatalities are around 1.25-1.50 per 100 million miles. [1] All forms of real-world autonomous driving across all manufacturers across the world are still somewhere below 200 million miles, conservatively estimated. [2] [3] Between the Tesla and Uber fatalities, by these rough back-of-the-envelope numbers, autonomous of various grades is still roughly 2X higher than human drivers. Maybe 1X if you squint at the numbers hard enough, but likely not orders of magnitude lower. I don't anticipate rapid legislative and insurance liability protections for autonomous systems until we see orders of magnitude differences on a per 100 million miles driven basis, and that will take time.
Waymo racks up about 10,000 miles per day across about 600 vehicles spread in about 25 cities. [4] Roughly 3.6 million miles per year if they stay level, but they're anticipated to rapidly add more vehicles to their fleet. In the US alone, about 3.22 trillion miles were driven in 2016. [5] Don't know what a statistically valid sample size is based upon that (I get nonsensical results below 2000 miles, so I'm doing something stupid), though. If Waymo puts two orders of magnitude more cars out there, they'll still "only" rack up about 365 million miles per year, and not all the miles on the same version of software.
[1] https://en.wikipedia.org/wiki/Transportation_safety_in_the_United_States
[2]…
Rodney Brooks is misleading. He says things like, "It was actually done in Germany in 1987." but neglects to mention that the road had no other cars on it.[1] He also claims that the first autonomous coast to coast drive happened in 1995, but actually the project was autonomously controlling steering, not gas or brakes, and 150 miles of the trip were driven by humans.[2]
Modern autonomous vehicles are much more impressive. They were successfully navigating urban environments 12 years ago.[3] Waymo's cars have driven over 10 million miles and disengagement rate is once every 11,000 miles.[4] We're at the point where no major breakthroughs are needed, just incremental improvements. That's what's different from earlier eras of autonomous vehicles.
1. https://en.wikipedia.org/wiki/History_of_self-driving_cars#1980s has this quote:
> In the 1980s, a vision-guided Mercedes-Benz robotic van, designed by Ernst Dickmanns and his team at the Bundeswehr University Munich in Munich, Germany, achieved a speed of 39 miles per hour (63 km/h) on streets without traffic.
2. http://www.cs.cmu.edu/afs/cs/usr/tjochem/www/nhaa/nhaa_home_page.html
3. https://en.wikipedia.org/wiki/DARPA_Grand_Challenge_(2007)
4. https://medium.com/waymo/an-update-on-waymo-disengagements-in-california-d671fd31c3e2
Will they be the current cars, or cars with the new next-gen hardware that are substantially safer? The cars that end up driving could signal a lot to the industry.
New lidar was spotted a few months ago and more recently on the Jaguar cars and the CEO said it’s an order of magnitude better: https://thelastdriverlicenseholder.com/2019/04/21/new-waymo-lidar-spotted/
https://medium.com/waymo/waymo-iaa-frankfurt-2019-b3cca36d8479
One would imagine this early announcement comes in reaction to their recent “valuation” cut: https://www.cnbc.com/2019/09/27/waymo-valuation-cut-40percent-by-morgan-stanley-to-105-billion.html
If Waymo starts driving without safety drivers en-masse, even on small fixed routes, it will signal certain takeover metrics have been hit. It will be interesting to see if Waymo uses existing cars (the ML and planners just got better) or if they use new hardware and signal a certain lidar improvement is necessary for achieving a viable takeover rate.
The signal won’t be clear but should be there. And then it’s up to the rest of the industry to show how fast they can react to the change.
1) Tesla does not report their numbers, though one person who went for a ridealong in the autonomy day demo reported one takeover in 20-minutes [1], so at least once every 10 miles.
2) Waymo does testing on a lot of types of roads. They reported 1.26 million miles in CA last year [2]. It's not possible to apples-to-apples compare this to Tesla's numbers because Tesla does not share their data. Musk even turned down an analyst's request the other day for details backing up his claim of autopilot safety [3].
Tesla/Musk are being unreasonable when they claim stats show their system is 40% safer or 2x safer than normal driving because they are comparing to NHTSA stats which include all classes of cars (including motorcycles), all road types (whereas AP is mainly used on divided highways), and all weather conditions (nobody activates AP in the snow, I hope, and I doubt many do in rain).
[1] https://youtu.be/2BZHXh1nbWc?t=450
[2] https://medium.com/waymo/an-update-on-waymo-disengagements-in-california-d671fd31c3e2
[3] https://www.youtube.com/watch?v=HqvatzjHGyk&t=47m17s
The technology for autonomous cars does exist. Latest data from Waymo is that they can drive 5.000 miles on average before disengagement (https://medium.com/waymo/accelerating-the-pace-of-learning-36f6bc2ee1d5).
They are not ready to be widely deployed but are close, which is why we're talking about them. It's an exciting technology with great potential impact on the future and a lot of that impact is positive.
As far as bikes go: no one is complaining about them. It's all in your mind.
As far as public transit: autonomous cars will improve it. By my calculations, autonomous car will be cheaper, unsubsidized, than a $2.50 bus ride subsidized by the city from our taxes (SF has ~$1 billion budget to fund money-loosing bart, buses etc.).
And that's cars. Things will be even cheaper if we get buses to be autonomous. At which point it's likely bus service will be cheaper and provided by a commercial company (the only reason public transport is government owned is because it looses money, so it wouldn't exist as private company).
Waymo says they have proof of #1 and #2.
We found that six weeks before his resignation this former employee, Anthony Levandowski, downloaded over 14,000 highly confidential and proprietary design files for Waymo’s various hardware systems, including designs of Waymo’s LiDAR and circuit board. To gain access to Waymo’s design server, Mr. Levandowski searched for and installed specialized software onto his company-issued laptop. Once inside, he downloaded 9.7 GB of Waymo’s highly confidential files and trade secrets, including blueprints, design files and testing documentation. Then he connected an external drive to the laptop. Mr. Levandowski then wiped and reformatted the laptop in an attempt to erase forensic fingerprints.
Ref: https://medium.com/waymo/a-note-on-our-lawsuit-against-otto-and-uber-86f4f98902a1#.3vm7k8r6n