Learning to Let the Car Drive

In anticipation of autonomous vehicles hitting the road – and with advanced driver assistance systems (ADAS) such as adaptive cruise control and collision avoidance already taking some control from drivers – the federal government and General Motors worked with the Virginia Tech Transportation Institute (VTTI) to conduct a study of driver behavior when they aren’t actively driving.
Image may contain Vehicle Transportation Car Automobile Human Person Driving and Cushion
GM's Jeremy Salinger tests Cadillac's Super Cruise System. Photo: GM.

What will drivers do when autonomous cars become a reality: text or tweet while behind the wheel, relax and take in the scenery, catch up on some reading or even some sleep? While you may be able to kick back once the car takes over, don’t expect to check out. And letting go of the wheel – and full control of the car – will likely happen in stages to better prepare drivers for when cars inevitably switch to autopilot.

In anticipation of autonomous vehicles hitting the road – and with advanced driver assistance systems (ADAS) such as adaptive cruise control and collision avoidance already taking some control from drivers – the federal government and General Motors worked with the Virginia Tech Transportation Institute (VTTI) to conduct a study of driver behavior when they aren’t actively driving. The goal of the Limited Ability Autonomous Driving Systems study, conducted in 2011 in a driving simulator at Indiana University-Purdue University in Indianapolis and with VTTI on a GM test track in Michigan, was to learn how drivers react when a car takes over primary tasks they’re used to performing.

“Learning from the advance safety features gets people familiar with this type of interaction and is a step towards automated driving,” GM’s Innovation Program manager Jeremy Salinger told Wired. While not widespread, ADAS features found on higher-end cars have started to trickle down to more modestly priced cars, and they'll help more drivers become familiar with the feel of the vehicle driving itself. “We certainly do see these systems migrating to the mainstream," says Ian Riches, director of global automotive practice for Strategy Analytics. “In 2009, over 70 percent of ADAS technology was fitted to premium vehicles. By 2019, we’re forecasting only around 40 percent of ADAS will be on these premium vehicles. To put this in some context, premium vehicles make up less than 10 percent of global vehicle production."

Though most drivers typically have some understanding of the capabilities of ADAS technology, most don’t grasp the systems' limitations, says Dr. Eddy Llaneras, the principal investigator for VTTI on the study. Using the example of adaptive cruise control – which maintains a set gap between the vehicle it’s installed in and those in the same lane in front so that the driver doesn’t have to repeatedly disengage and engage the system – Llaneras notes that some drivers don’t always understand the system’s restrictions. “They get comfortable with it, but then they forget that it’s on,” he says, such as when a system’s line-of-sight radar sensor “loses” a vehicle ahead in a curve and the host car speeds up.

Cadillac’s semi-autonomous Super Cruise system, which combines lane-centering technology with adaptive cruise control and could be available by 2015, was the primary technology used in the study.

Llaneras points out that even when the vehicle can steer itself and stay in its lane, the driver still has to deal with unexpected obstacles or events. “With autonomous vehicles,” he says, “drivers may think the system can steer around obstacles, like when a truck loses a tire tread on the freeway. But it won’t – at least not yet. The systems that we evaluated did have that capability, but it’s an example of how there’s always limitations of what it can do and can’t do, and drivers need to learn these things over time. We’re talking about what I consider for low-level interventions,” he adds. “People also need to understand the car is able to maintain the lane only if given clear markings and good weather. And it’s not going to be able to respond to obstacles that come in from the side. If a deer runs in front of the vehicle, you may be able to see it before the sensors do.”

“The initial deployment of this kind of technology where the drivers doesn’t need to steer anymore will still require the driver to supervise the system and pay attention,” notes GM’s Salinger. Llaneras adds that a key factor will be keeping drivers engaged even if the car is doing the driving. “If you give people a vehicle that will self-steer, chances are they’ll start doing things like texting or making phones calls. We’ve had people read magazines and watch videos. So we’ll need to come up with strategies that get the driver to look forward. The system that we tested monitors the driver,” he adds. “It makes an assessment of whether they’re looking forward toward the road or away. One of the key findings is that if you put a rudimentary system in a vehicle that will achieve those two functions – steering and maintaining spacing – you need to have another system overlaid on this that will monitor the driver.”

When the vehicle detects that drivers are looking away from the road for too long, they’ll get an alert. Lexus already offers a Driver Monitoring System that uses an infrared camera to detect if a driver is looking away from the road. And Mercedes-Benz prototype Traffic Jam Assistant takes it one step further, requiring the driver to at least have one hand on the wheel when the auto-steer function is turned on.

According to a GM spokesperson, the safeguards the automaker is developing will be an extension of the driver alerts that are already on vehicles today, such as those used for forward collision warning and blind-spot monitoring. As for specifics, Llaneras says he’s “not at liberty to discuss some of the things GM is looking at. But I can say that … they’re doing things to keep drivers engaged in making high-level, strategic decisions."

Llaneras adds that the findings of the study – and GM’s use case for autonomous cars – are based on freeway driving, not surface streets with traffic signals and complex roadways. “Limited environments are a good place to start and for drivers to get their feet wet, to get them to appreciate what the system can and can’t do,” he says.

GM’s Salinger believes that “these new safety features and convenience features will change the way people interact with their cars and how they use theirs cars.” He says he’s looking forward to a time when drivers will let the car do most of the driving tasks, although not all the time. “People will still want to drive themselves sometimes,” he adds. “But there are a lot of times when driving is very boring.”