Two recent studies by the Insurance Institute for Highway Safety (IIHS) found drivers are more likely to multi-task with partial automation, even while following the system’s safety rules. 

“These results are a good reminder of the way people learn,” said IIHS President David Harkey, in a press release. “If you train them to think that paying attention means nudging the steering wheel every few seconds, then that’s exactly what they’ll do.”

The first study found participants driving a Volvo S90 with adaptive cruise control (ACC) and Pilot Assist (PA) were more likely to show visual-manual disengagement after a couple weeks of using PA. 

ACC is described in the report as lane centering that provides continuous longitudinal and lateral vehicle control. PA is described as a partial driving automation system that pairs the longitudinal of ACC with continuous lateral control. 

“Our findings suggest that drivers show reliably less attention to the road with the versions of partial automation we tested compared with manual driving that develops with time or is evident relatively early during drivers’ initial trips with the automation,” the report says. “The results support arguments for driver-monitoring solutions that ensure adequate attention to the road.”

The report says that advanced driver assistance systems (ADAS) are designed for crash avoidance or mitigation and driver comfort.

“In contrast, ADAS designed for comfort and convenience typically seek to alleviate the stress or boredom of driving,” the report says. “While instances of stress and boredom may also impact driving safety, systems may be more frequently designed for, and engaged by, the driver as a matter of convenience rather than a concern for safety.”

It notes that field research suggests that ACC use is associated with increased headway and lower G-force when braking relative to manual driving. However, it is also associated with increased secondary activity by the driver and complacency can reduce the driver’s readiness to react to hazards the technology can’t handle. 

The study used 29 participants in the greater Boston area, 20 of which commuted for at least 30 minutes on highways each way to work. The other nine drivers traveled 100 miles of highway three to four times per week. Each of the participants said they had never driven a vehicle with ACC. 

The drivers were separated into three groups. The first, Group A, was before any software updates that increased driver awareness of needing their hands on the steering wheel. 

The second group, Group B, drove the vehicle with one of Volvo S90’s updates that included software to warn drivers to return their hands to the wheel when the steering wheel torque sensor detected no hands on the wheel after some time, the study says. This included a visual component and text instructions for the driver to “Apply Steering.” 

A third group, Group C, drove vehicles that had the first update and a second update that added an escalation of color and sound including repeated tones that increased to pulsing tones as a warning. 

Each of the participants was trained on the vehicle. 

The data set included a total of 138 hours with a mean speed of more than 62 miles per hour on 8,619.5 miles driven. 

Group C, which drove vehicles with both software updates, drove twice as long as Group A and B. The group also significantly used more ACC and PA, the report says. 

The study found drivers were engaged in visual-manual behavior while driving between 5.7-33.7% of the time. 

The more the drivers used ACC and PA, the more they started to engage in visual-manual activity such as checking their phone, grooming, or eating, the report says. 

Distracted driving activity by Group A (exposed to no software update) and Group B (exposed to the first software update) increased by three times from the first two weeks to the second two weeks, the report says. The group also engaged in more activity during the first two weeks than Group C. 

However, Group C engaged in more activity in the second two weeks than Group A and B. Group C was engaged in distracted driving activity 33% more during the last two weeks, the report says. That’s about twice as much as driving manually in either period. 

“Our findings suggest that driver disengagement when using partial automation is greater than when driving without it, with differences evident despite software updates that changed system characteristics and wide variation in the samples’ commuting patterns and propensity to use automation,” the report says. 

The second study from IIHS investigated how 14 drivers with no partial automation experience evolved over a month of exposure to the Tesla Autopilot system in the 2020 Model 3. 

“We found that drivers learn to internalize safeguard sequences and discover windows of opportunity to do non-driving-related activities,” the study says. “People learned to respond quicker to alerts, leading to fewer escalated sequences in the latter half of the study. However, drivers also spent more time engaging in non-driving-related activities and glancing off-road, which corresponded with more initial alerts of the attention reminder sequence as time went on.”

The 14 drivers received training on the vehicles upon delivery, according to the study. 

The driver-monitoring system on the Tesla model tracks steering torque through sensors, the study says. The driver is required to physically move the wheel for the sensor to register any hand-on-wheel input. 

The Autopilot escalation sequence starts with a visual icon display, text notification, and flashing blue light. The second phase changes the grey icon to red, sends a second text notification, and the blue light increases its speed of flashing. This phase also includes one short audio tone burst.

A third phase includes the red icon, another text notification, and the blue light increasing in frequency with two short audio tone bursts. The fourth phase includes a larger red icon, and a text notification alerting that Autosteer will be unavailable for the rest of the drive. The audio tone burst starts to be continuous and the vehicle slows down and locks the driver out of the Autosteer. 

The study observed 3,858 attention-related alerts during 12,161 miles of driving with Autopilot engaged. 

It found that drivers received the first phase alert about 98% of the time. This increased by 39% from the first week to week four, the study says. Escalated alerts decreased by 76% over the same time. 

Overall,  72 phase one alerts escalated to other phases, the report says. Also overall, the phase one alerts lasted 13 seconds, phase two lasted 3 seconds and the third and fourth phases lasted an average of 8 seconds. 

Drivers used behaviors such as nudging the steering wheel, toggling the scroll wheels, and pressing the turn stalk/gear shift stalk to stop alerts, the report says. It says drivers did not always use their hands to nudge the steering wheel; sometimes they used other body parts, such as their knee. 

It notes that phase one alerts increased significantly over time relative to the first week. Drivers were 8% more likely to experience the phase one alert in the second week. This likelihood increased by 25% in weeks three and four. 

“When we began this study, we did not expect to see many, if any, emergency-slowdown-to-lockout events in the data; yet we observed 16 lockout events among four of the 14 participants,” the report says.

It notes one driver had 12 lockouts and three of the drivers attempted to reactivate the system after being locked out.

“The learning curve we observed with attention reminders also exists for the last resort countermeasures, as lockouts occurred predominantly within the first half of the study, with only one lockout event happening in the fourth week,” the report says. “The driver who had the most lockout events also grew quicker to respond to them over time.”

The data shows that drivers learned how to find windows of opportunity for non-driving activities, the study says. 

“Part of the safeguard solution may include adaptive activation and escalation upon the detection of persistent driver disengagement,” the study says. “Continuous monitoring and adaptation of countermeasure onset and escalation may be instrumental in addressing the idiosyncratic disengagement tendencies of a diverse driving population.”

IMAGES

Photo courtesy of LeManna/iStock