Navigant Research Leaderboard Report: Automated Driving
navigantresearch.com
navigantresearch.com
> In 2016, Waymo logged more than 635,000 miles while testing its autonomous vehicles on California’s public roads compared to just over 20,000 for all its competitors combined.
https://www.forbes.com/sites/chunkamui/2017/02/08/waymo-is-c...
> Tesla’s sensor suite has so far been much lighter than Google’s, but that is now changing with the second generation sensor suite making its way to customers now. As of the end of October 2016, Google had accumulated about 3.5 million miles of data (2.2 million autonomous and 1.3 million manual miles) and it is adding about 25,000 miles per week with a fleet of just over 50 prototypes.
> On the other hand, Tesla is now adding millions of miles every day from its first gen Autopilot. The second generation, which aims to be fully self-driving like Google’s, should catch up with the first generation in term of data gathering during the second half of 2017, but with a much greater scope of information with now 360-degree camera coverage powered by ‘Tesla Vision’, as well as better ultrasonics and Tesla’s new radar technology.
https://electrek.co/2016/11/13/tesla-autopilot-billion-miles...
It feels to me like Waymo is trying to go from standard cars all the way to level 4 in one leap, but with a tiny number of test vehicles and locations. Tesla is going at it piecemeal instead, but are increasing their data they collect and the capabilities of each generation exponentially; and as their customer base grows, their data collection grows in scope and usefulness well beyond what Waymo can deploy out of pocket.
Google because they are competent at software engineering and have the internet / maps advantage to build really "smart" self driving cars that can talk to each other, learn and send data back to server for other cars to use, etc.
Competency developing software, not competency building vehicles will be the key factor here. Google will most likely find success licensing their technology to anybody rather than making their own cars. I could see Ford and GM getting some basic self driving software working by themselves but as the software gets more advanced and complex they'll really need a partner like Google.
While I always liked Tesla's demo videos the posted I learned all I wanted to know about its inability in imperfect conditions by just surfing Youtube for user demonstrations of AP2.
Has anything come out of Google lately? Also, how many programs are still operating on "if we just can map it all" mentality. Level 4+ are going to have to have cars that adapt to ever changing conditions.
I am still curious about one long term question, can an autonomous car speed and if it does who is at fault?
I would also imagine self driving a bus would be simpler, as unlike cars, they have mechanisms to change the flow of traffic themselves (some buses have stop signs, plus bus only lanes, bus stops, etc.).
Examples:
They'll lose on cost, efficiency, and convenience to one or the other in almost every category of travel.
These metrics seem absolutely stacked to favor incumbents with "wait and see" approach.
Why not simply have self driving car algos compete in a random and controlled computer environment and see which performs the best from a clearly objective standpoint? Why such completely subjective metrics in this scale when these types of technologies can be measured so easily with a single computer?
This is simply to win over new consulting clients and has very little bearing on reality
http://www.theverge.com/2016/12/28/14100278/ford-new-self-dr...
https://medium.com/@ford/building-fords-next-generation-auto...
But they also have the ability to be producing vehicles with that tech, at volume, in a couple of years. Waymo has essentially 0 manufacturing capacity and Tesla is still spooling up to get their annual production comparable to monthly production for GM or Ford.
On the other hand, in looking for a new crossover SUV this year, I was surprised at how many of the pieces of autonomous driving the 2017 Ford Escape actually had in production. It's not as headline-grabbing as autopilot for sure, but perhaps it's a relevant approach.
I've been pretty bullish about Volvo's approach (and put the brakes on that SUV search to wait for the 2018 XC60), so it's a little disappointing to see them relatively far down that list, while VW/Audi, which I see as weaker in these areas, gets a better position.
The report itself is not available without logging in, apparently. The claims seem rather vague to me, and (admittedly, as a techy) it would seem easy to undervalue experience with software engineering at large in this endeavour.
If you log in it just tells you the price: $3,800.