Disengage, disengage! Cali DMV reports show how often human drivers override robot cars

Driverless tech not ready, to shock of nobody

Mercedes' driverless cars need human intervention approximately every 2.08km (1.3 miles), and other makes are totally reliant on frequent switching to manual, according to figures out this month from the Californian Department for Motor Vehicles.

The "disengagement reports" (the times an autonomous car was taken over by the human tester) of the major autonomous vehicle companies that test on the US Golden State's highways and byways are available on California's DMV's website. Driverless tests take place in several other US states, but there is no equivalent law in these places.

Twenty companies were required to produce reports, and 19 did so (class dunce Faraday Future is being followed up by the DMV for not handing in its homework). Seven – including BMW, Volkswagen, Ford, Honda and Tesla – reported that they did no testing on public roads, or at least none in California.

At the top of the list was Waymo, which completed 567,365km (352,544 miles) with 75 cars, and 63 disengagements. That works out at a fairly respectable one disengagement every 9,005km (5,595 miles), which is roughly between a third and half of the average individual's total annual driving distance.

Second was General Motors' Cruise, driving 86 cars a total of 211,910km (131,675 miles). It had 105 disengagements, or one every 2,018km (1,254 miles). However, GM's fleet was involved in 22 collisions last year, and two already this year, which doesn't look fantastic compared to Google/Waymo's three crashes in 2017.

The other manufacturers on the list drove far fewer vehicles and total miles, but many had far more disengagements. Mercedes-Benz, for example, drove three vehicles 1,749km (1,087 miles), but required 842 disengagements, equal to one disengage every 2.07km (1.29 miles) (although only 240 of these were due to the test driver's decision, the rest being automatic). Electronics behemoth Bosch also fared poorly, disengaging its three cars 598 times over 2,340km (1,454 miles), or once per 3.9km (2.43 miles).

John Simpson, of US non-profit Consumer Watchdog, points out that the Californian data is "only a partial picture, because other state regulators aren't protecting their citizens" by requiring similar reports. He added: "Despite the self-serving hype of the manufacturers, robot technology simply isn't ready for our roads without hands-on, behind-the-wheel engagement and supervision by a human driver."

While the above data shows that driverless tech is indeed still a while away from being fully reliable, none of the autonomous vehicle makers that we can think of are claiming to be so, and certainly don't intend to be any time before 2020-ish. We might have reached the "Kitty Hawk" moment for self-driving cars, but still regular interventions by test drivers and the occasional crash can be expected for some time yet. ®

Other stories you might like

  • The future: Windows streaming through notched Apple screens

    Choice is the word for Jamf's Dean Hager

    Interview As Apple's devices continue to find favour with enterprise users, the fortress that is Windows appears to be under attack in the corporate world.

    Speaking to The Register as the Jamf Nation User Conference wound down, the software firm's CEO, Dean Hager, is - unsurprisingly - ebullient when it comes to the prospects for Apple gear in the world of suits.

    Jamf specialises in device management and authentication, and has long been associated with managing Apple hardware in business and education environments. In recent years it has begun connecting its products with services such as Microsoft's Azure Active Directory as administrators face up to a hybrid working future.

    Continue reading
  • There’s a wave of ransomware coming down the pipeline. What can you do about it?

    AI can help. Here’s how…

    Sponsored The Colonial Pipeline attack earlier this year showed just how devastating a ransomware attack is when it is targeted at critical infrastructure.

    It also illustrated how traditional security techniques are increasingly struggling to keep pace with determined cyber attackers, whether their aim is exfiltrating data, extorting organisations, or simply causing chaos. Or, indeed an unpleasant combination of all three.

    So, what are your options? More people looking for more flaws isn’t going to be enough – there simply aren’t enough skilled people, there are too many bugs, and there are way too many attackers. So, it’s clear that smart cyber defenders need to be supplemented by even smarter technology incorporating AI. You can learn what this looks like by checking out this upcoming Regcast, “Securing Critical Infrastructure from Cyber-attack” on October 28 at 5pm.

    Continue reading
  • Ransomware criminals have feelings too: BlackMatter abuse caused crims to shut down negotiation portal

    Or so says infsec outfit Emsisoft

    Hurling online abuse at ransomware gangs may have contributed to a hardline policy of dumping victims' data online, according to counter-ransomware company Emsisoft.

    Earlier this month, the Conti ransomware gang declared it would publish victims' data and break off ransom negotiations if anyone other than "respected journalist and researcher personalities" [sic] dared publish snippets of ransomware negotiations, amid a general hardening of attitudes among ransomware gangs.

    Typically these conversation snippets make it into the public domain because curious people log into ransomware negotiation portals hosted by the criminals. The BlackMatter (aka DarkSide) gang's portal credentials (detailed in a ransom note) became exposed to the wider world, however, and the resulting wave of furious abuse hurled at the crims prompted them to pull up the virtual drawbridge.

    Continue reading

Biting the hand that feeds IT © 1998–2021