The 14 things wrong with SAE Levels of Autonomy [Part #5]
January 27, 2020
What SEQ will look like in 2032 – Infrastructure
March 13, 2020
Last year I posted a series of articles about road authorities and their responsibilities in supporting autonomous vehicles followed by a breakdown of the SAE’s Levels of Autonomy and why we should just stop thinking like that. It is so 201x-ish.
Being a new decade, let’s start afresh with autonomous vehicles (AV), take stock, survey the landscape, and see what’s hype and what is legitimate.
In my mind there are two big players that we need to watch very closely - Waymo and Tesla - and I will describe the reasoning in upcoming articles. But let’s have a look at the other players and see what they have been up to and who should be watched and who should be ignored.
Uber
The tragic accident that cost the life of Elaine Herzberg in Tempe, Arizona in March 2018 was a watershed moment for Uber and the entire industry. There were numerous contributing factors behind this event and shadow has been cast over the driver and their level of attentiveness but to be clear, the car saw the pedestrian 1.3 seconds before impact and determined that emergency braking was required but the system had no mechanism to either alert the driver or to enact the emergency braking of its own accord.
(Pause and read that last sentence again and then I would encourage readers to internalise a string of expletives)
Uber immediately suspended the testing of their cars but resumed scaled down trials (point to point between two of their offices limited to 25mph) in Pittsburgh in December 2018.
Suffering from technical and cultural problems meant that Uber did not appear to progress in 2019 other than to have duct taped their sensor rack onto yet another type of car that was announced at their 2019 Summit to some fanfare.
The accident occurred in a Volvo XC90 and this new news is that Uber are now using a… Volvo XC90.
A purchase of 600 acres of land in late 2019 to build a test city would suggest they are going back to school (even though they already had one).
Reports are that they keep failing track validation tests.
Uber had a disappointing IPO in 2019 but their market capitalisation is an eyewatering US$53B.
We probably don’t need to revisit Uber until 2025.
Zoox
Just because some rich person invests in a company, doesn’t mean that it’s a good idea. Mike Cannon-Brookes (of Atlassian fame) and Zoox is one such example. Zoox is based in California and founded in 2012 by Australian Tim Kentley-Klay and American Jesse Levinson.
Zoox is targeting the robotaxi market. They are creating not only the software behind autonomous vehicles but also the vehicle itself… but it’s not road legal so they stick their sensor suite on Toyota Highlanders.
They shun the media for the most part with the exception of this interview with Bloomberg in July 2018 which includes a great video.
Zoox includes the stereotypical founder’s story and cultural eccentricities that are expected of a company worth over $3B with no product and no revenue and a head count over 1,000 and growing.
A recent review Jesse Levinson says:
Zoox CTO and cofounder Jesse Levinson told me that the company’s vehicles are better than human drivers at following the rules of the road, but worse at handling unusual events.
Oh dear.
Zoox may be entirely legitimate and it may “win” (whatever winning is) but they release so little information as to be a curiosity that needs to be carefully monitored.
Mobileye
Mobileye is an Israeli company acquired by Intel for $15B two years ago. They have a great roadmap initially focused on ADAS technology (Advanced Driver Assistance Technology), digital mapping the road network then on to Robotaxis and autonomous vehicles.
It appears that they are going to be providing the technology to the car manufacturers, rather than trying to build one themselves like Zoox is doing.
Mobileye have partnered with Volkswagen, BMW, Honda, General Motors, Nissan, Opel, Renault, Fiat Chrysler, Nio and Volvo and Telstra are their Australian distributor.
Contrasting Mobileye with Zoox is interesting. Whereas Zoox appears to be doing a ‘lunge for the line’ including the software and the car, Mobileye has a clearly articulated plan to build products, make revenue, reinvest, incrementally build their capabilities and partner with existing car companies. Zoox is funded by venture capitalists and acts like cowboys and Mobileye was bought by an 800lb gorilla and is behaving like a mature team with a plan.
People catching buses in Brisbane may have heard the bus beeping in a new way in the last 12 months. You can see the Mobileye Shield+ System installed at the bottom of the windshield. It beeps whenever it detects potentially dangerous situations.
Yandex
Yandex is a Russian search engine that is now attempting to build self-driving cars (just like Google and Waymo).
Yandex has partnered with Hyundai.
The company already runs Russia's largest ride-hailing service, Yandex.Taxi, which saw revenue in the first half of 2019 and turned profitable in the second quarter.
The thing that stands out is that their demo of automated driving appears to be as good as, if not better than Waymo. It is amazing.
They had a big presence at CES 2020 doing unprotected turns and driving with great confidence.
They have clocked 1.5M miles of autonomous driving.
From watching the CES videos, the car appeared to be doing the same route so it could all be smoke and mirrors.
THE YEAR OF SELF-DRIVING CARS
WHEN WILL SELF-DRIVING CARS GET HERE
Yandex has been keeping a low profile but needs to be closely watched.
Lyft, Aptiv and NuTonomy
Lyft, a ride hailing company like Uber, also has an autonomous vehicle initiative.
They appear to have a social conscience.
They are creating their own tech, partnering with others, sharing their app with Waymo and Aptiv and the data they are collecting is open sourced and available on Github.
According to the company, 96% of people who try hailing a driverless vehicle in the Lyft app say they want to do so again.
There is scant information about the quality of their self-driving cars; they release information very carefully. They have an autonomous vehicle lab called “Level 5” and Ford Fusions kitted out with an array of sensors in a unique configuration.
Lyft operates the “largest public self-driving commercial programme in the US having completed over 75,000 rides”.
They appear to have had a very big 2018 but were strangely quiet in 2019.
They partner with Aptiv, a company with a colourful past who also make autonomous cars through their acquisition of NuTonomy in 2017 and appear to be positioning themselves as ‘platform creators’ that will licence their tech to the auto manufacturers. Aptiv also had a very big CES 2020. They have been operating in Las Vegas for about 18 months with BMWs 5 Series – but unlike other manufacturers the sensors are very discreet and it looks just like a normal car rather than the Ghostbuster’s Hearse.
The line where Lyft stops and Aptiv starts is confusing. How many cars Lyft is operating autonomously is vague but Aptiv has been running a fleet of 75 of them in Las Vegas since 2018.
Of the 75,000 rides, none appear on YouTube…. Really? Who are they driving around?
There is something going on here but they are not releasing enough information to know the true lie of the land.
Are there engineers behind the wheel?
Does it work in all weather conditions?
Does it work all over the city?
Comma AI
Founded by George Hotz, an IT prodigy, hacker and roboticist, the Comma Two hardware powered by Openpilot open source software fits to about 60 types of cars. It is attempting to compete with Tesla Autopilot and GM Super Cruise.
It has a rudimentary sensor suite but in a series of simple tests it appears to have behaved adequately.
Will it be ever able to navigate complex city streets?
Probably not. This tech looks like it will be little more than “level 2” and if you use it, please pay very close attention to the road and check whether it will void your insurance.
