Why does Google Waymo need to put more than 3,000 autonomous vehicles on the road in 2018?

Yole Développement’s (Yole) release of Sensors for Robotic Vehicles 2018 comes just as dramatic developments in this field have happened in Arizona and California. But we, at Yole would like to give another, different, view of the autonomous vehicle landscape at the start of 2018. We have made a few graphs from the 2017 Californian reports for when autonomous mode has been disengaged, and our analysis follows.

Disengagements are actions from a human or the vehicle itself which interrupts the autonomous mode.

Miles driven per disengagement in 2017 correlates well with the cumulated miles driven by vehicles from each company in that year. But let’s be careful, we are only talking about California. Waymo vehicles have driven around 5,000,000 miles in total, of which just 300,000 were in California in 2017. All players are tending to increase the number of miles their vehicles have driven each month except Waymo, which had a slight decrease at the beginning of 2017, before it rose again.

The cumulated miles driven:miles per disengagement ratio indicates the aggressiveness of players’ tests. Does each new mile traveled improve or degrade the number of miles per disengagement? Cruise, for example, softened its behavior during summer 2017, as its aggressiveness reduced, and then became more confident again, accepting fewer miles per disengagement. By contrast Waymo was more aggressive in the summer and finally this metric increased as it got more miles per disengagement at the end of the year. The aggressiveness can be driven by taking fewer risks or by significant system improvement, and we can’t know which is the case.

However, 100 seems to be the upper limit for the cumulated miles driven:miles per disengagement aggressiveness ratio players are targeting. They seem to either upgrade their system, correcting for any bugs or hardware upgrades, or drive in less dangerous conditions when they near this figure.

To explain why they’re targeting this number, remember that the aggressiveness metric is the total number of miles their vehicles have driven divided by the miles per disengagement figure. If companies want to progress they have to go to the limits of their system. In a sense, they have to create disengagements – but not too many. One goal is to solve any issue that has occurred in order to raise the miles per disengagement score. Yet it appears that no matter how big the company or how advanced they are in their program, they seem to have this same goal, to near 100 disengagements per year. If they were to generate more than 100 per year Yole’s analysis is that their engineering teams would not be able to keep up with solving the problems raised.

The disengagement data also show that for Waymo, the number of miles traveled in California no longer allows for a fine measure of actual performance, as the company had only one disengagement in November 2017.

Waymo is not so far from reaching 100,000 miles per disengagement, so it will need to drive at least 10 million miles in 2018 if it wants to keep its aggressiveness score below 100. This is three to five times the mileage it’s doing right now with 500 vehicles. It will therefore be forced to change scale and put 1,500-3,000 vehicles on the road in 2018. This is what Waymo needs just to get to the next level of improvement.

Source:   

Author:

Pierre Cambou joined the imaging industry in 1999. Following an engineering degree from Université de Technologie de Compiègne in parallel to a master of science from Virginia Tech in 1998, as well as graduating from Grenoble Ecole de Management’s MBA, Cambou took several positions at Thomson TCS, which became Atmel Grenoble in 2001 and e2v Semiconductors in 2006. In 2012 he founded Vence Innovation, now called Irlynx, in order to bring to market a disruptive man-to-machine interaction technology. He joined market research and strategy consulting company Yole Développement as imaging activity leader in 2014.

RELATED WEBCAST

Core Technologies for Robotic Vehicle – Get access to the replay

Get ahead of the robotic vehicle wave and get to know the key technologies for this transportation revolution to happen.
Yole Developpement and KnowMade have organized a special webcast on April 17^th 2018 – you can still register here and get access to the replay, do not miss it!

After reviewing the key insights of our latest report Sensors for Robotic Vehicle 2018 they had presented two special focuses: the first one on computing technology for AV and the second one on the LiDAR IP landscape. The complete presentation part last 30 minutes and is followed by a 30 minutes Q&A with the analysts.

RELATED REPORT

Sensors for Robotic Vehicles 2018

High end industrial sensors will win in the emerging robotic vehicle industry.
Get more here.