Google’s Waymo ahead in driverless race
BY RON HAMMERTON | 2nd Feb 2017
According to reports published by the California Department of Motor Vehicles (DMV), Waymo’s fleet of 60 test vehicles racked more than a million kilometres on Californian roads last year while recording just 124 system disengagements.
This compares with 341 disengagements in 680,000km in 2015 – a drop of 75 per cent per kilometre travelled.
Disengagements – when the driver has to take control of the vehicle because the automated systems get into difficulty or the driver errs on the side of caution – fell from an average of 0.8 per 1000 miles (1600km) in 2015 to 0.2 in Waymo test cars in 2016.
The DMV, which has issued 11 autonomous vehicle testing licences to automotive and tech companies such as Tesla, Ford, Bosch, Delphi, BMW, Volkswagen and General Motors, requires the companies to self-report such episodes, along with distances travelled.
Unlike Waymo, most of the companies involved in Californian testing clocked far smaller distances, with some reporting only hundreds of kilometres of autonomous driving on public roads in 2016. VW and Honda did none, at least not in California.
Tesla – a California-based company with a stated intention of launching autonomous vehicles as soon as possible – reported only 850km of California roads autonomous testing, all in October in a fleet of four vehicles.
While Tesla reported one disengagement every 3.5 miles (5.7km), Waymo’s rate was just one every 5000 miles (8000km).
In 673 miles (1080km) of autonomous driving on urban streets, Mercedes-Benz recorded 341 disengagements, or one every two miles.
GM affiliate Cruise was one of the more active test companies, logging almost 10,000 miles (16,000km) in a fleet of 20 electric Chevrolet Bolts that averaged a disengagement every 50 miles (80km).
The problem with comparing such data is that there is hazy official standard for what constitutes a disengagement.
Most companies have been taking a conservative approach in order to avert disaster, instructing test drivers to take over at the slightest hint of trouble and programming systems to alert the driver if, for example, sensors are having trouble reading line markings.
What is undeniable, however, is that Waymo has launched into public-road testing on the United States west coast with a passion, fielding a much bigger test fleet than others and logging much higher mileage than any other.
In all, the company racked up 3.7 million kilometres in public roads testing in the US in the 12 months to the end of November – up 50 per cent on the previous reporting period, which was 14 months.
In its report to Californian authorities, Waymo says disengagements are a natural part of the testing process, allowing engineers to expand the software’ s capabilities and identify areas of improvement.
“During our testing, our objective is not to minimise disengagements,” it says. “Rather, it is to gather – while operating safely – as much data as possible to enable us to improve our self-driving system.
“Therefore, we set disengagement thresholds conservatively, and each is carefully recorded.” Waymo says its drivers are required to err on the side of caution and take manual control if they have any doubt about the wisdom of continuing in autonomous mode.
Some of these situations have been triggered by a failure of back-up systems in the computer control, while others are caused by behaviour of other road users, including cyclists.
One of the major causes of disengagements is a discrepancy in guidance software – presumably where the mapping does not match the real road surface.
Waymo says each incident is recorded automatically by the driving system and later “replayed” by technicians with input from the drivers involved.
As GoAuto reported in December, Waymo has forged a partnership with Fiat Chrysler Automobiles to develop autonomous driving technology.
The dealer involves a fleet of 100 Chrysler Pacifica Hybrid vehicles, each fitted with a roof-mounted sensor system and test computers.
