One man dies with Tesla Autopilot turned on and there’s talk of scaling back autonomous driving. Google is said to have jumped straight to work on Level 4 self-driving, the highest level (where cars wouldn’t even need a steering wheel), because drivers at lower levels of autonomous have shown their inability to stay attentive. It’s possible safety zealots and regulators seek to reign in self-driving cars. That’s unnecessary.
Here are five things automakers, regulators, car dealers, and drivers can do to make self-driving safer. Otherwise, we risk losing the very real advantages of partial self-driving, such as adaptive cruise control (ACC) and lane centering assist (LCA) that can be soothing on big city highway commutes and brain amplifiers on long, tiring vacation trips.
Cars today with adaptive cruise control and lane centering assist, possibly with blind spot detection and forward emergency braking, are effectively self-driving on interstates, until they encounter a situation they can’t handle. They require the driver to have his or her hands lightly on the wheel. Take them off, the car senses it, sounds a warning beep after 10-15 seconds, and disables lane centering assist about 5 seconds later. Some automakers set the timeout to a minute or more, enough time to crawl in the back seat and make a YouTube video, or for the average driver to really lose attention and possibly nod off.
If the timeout is 15 seconds, that’s enough. It’s enough to plug in your smartphone, turn around, and grab a bag from the back seat (not that you should) or type a quick text (not that you should).
“Tesla Autopilot” is a great marketing phrase, but it promises more that it delivers (for now), even if Tesla appends the word “beta.” What’s needed is not an owner lawsuit charging false promises, but a serious automaker / auto dealer training program for buyers, and their families, that explains all technology features of their new cars: the cockpit controllers (BMW iDrive, Audi MMI, Mercedes-Benz Comand), the LCDs and navigation systems, and the alphabet soup that makes up assisted driving: ACC, AEB, BSD, and LDA/LKA/LCA.
Here’s the problem: Auto dealer sales forces have no clue, generally speaking, what they’re selling when it comes to technology. If they understood technology, they’d be employed somewhere else that didn’t require working 60-hour weeks to earn $50,000. Some automakers have created regional geek squads to educate the sales forces who educate the consumers; some dealers designate one or two smart guys to explain tech (until they get hired away elsewhere). Give the buyer a $50 accessories department gift card for sitting through the training (carrot, not stick). Back that up with online videos and embedded videos that play in the car (when it’s stopped) for continuing education. It’s a start.
When the driver in an adjacent lane cuts into your lane, it takes adaptive cruise control about a second to lock in on the intruder and slow your car. This can be a scary moment for driver and passengers, and it’s hard to drive 50 miles on the highway without it happening.
The camera in the windshield that handles lane keep assist as well as (some cars) auto high beams and forward collision warning could track cars ahead of you in adjacent lanes. It could work with a centralized controller and ACC to get the car braking a couple fractions of a second earlier. This would raise a car from Level 1 autonomous driving, meaning one or more driver assists working on their own, to Level 2, which calls for multiple driver assists — especially ACC and LCA — that work together. For the next couple years, that’s what drivers really want and will find useful for highway commuting (handling stop and go traffic while staying in lane) and the occasional all-day trips where it’s hard to pay attention.
It’s normal for your attention to wander if the car helps with driving. But until cars get to the next level, where you get enough time to resume controlling the car (say 10 seconds to a minute), you have to stay more-or-less alert. Drowsy driver monitors can help. Originally conceived as a complex system with cameras and eye motion tracking, sometimes also measuring breathing and heartbeat, automakers have found they can track driver attention by tracking micro-adjustments people make when driving, as long as their hands are on the wheel. The car also tracks events such as drifting out of lane, which lane centering assist will catch.
With one technology or the other, the car should be able to catch a driver on the verge of drifting off. Then the car sounds an alert and suggests a coffee break. Researchers have also experimented with Let’s Play a Game, where the car voice module asks the driver questions and listens for response via speech-to-text conversion. That can keep the driver going for another hour or so. It’s a possible antidote to inattention.
Inattentive drivers is one reason Google back in 2013 went straight to working on driver-free cars: Drivers don’t remain vigilant and lack the “situational awareness” to quickly return to full attention.
In the 1990s plan, self-driving cars and trucks would follow embedded transponders in a limited access roadway. Now the car orients itself on any roadway based on visual recognition of nearby cars, traffic signs, and most of all, roadway markings. They all need to be kept in good condition, especially the lane markings. What’s marginal on a clear day can’t be read in the rain at night by the car’s lane centering assist camera (or by most drivers, for that matter). Some cars get speed limit cues from road signs, but only if they’re readable. Potholes force drivers to swerve, and that endangers a self-driving car nearby. Potholes also cost individual drivers more in tire, wheel, and suspension damage than it costs the state to fix the potholes for everyone.
The money is there, in the form of auto registration fees and state and federal gasoline taxes. Problem is, much of it has been siphoned off to other parts of government budgets, and now needs to be recaptured.