ADAS: A Game of 'What Does This Button Do?'

ADAS features offer great potential to assist drivers in reducing crashes, injuries, and fatalities when used properly -- in theory. Ron Wilson shares his recent ADAS experiences and some thoughts.

(Image: iStock)

I had the opportunity recently to spend some quality road time with a new sedan—whose maker shall go unnamed—and to explore its ADAS features. I made some specific notes about the behavior of the vehicle under the control of ADAS. And on reflection, the experience has led to some speculative thoughts about the direction of the technology.

My first discovery was that unless you are one of those people who habitually reads the entire manual before opening the door of an unfamiliar car, most of the ADAS features on this vehicle will be bits of an enigma. Individual features are activated by buttons on the steering wheel bearing rather cryptic ikons. Decipherable, but hardly a puzzle you’d choose to undertake while on a busy San Franscisco Bay Area freeway. There are also apparently ADAS modes, which were completely undecipherable, and so not explored.

So commenced a game of “what does this button do?” My first exploration, actually, didn’t require a button at all—the blind-spot warning indicators are apparently always on. The theory is that if there is a vehicle in either your left or right blind spot, a warning will illuminate on the instrument panel, and another will illuminate in the corner of the corresponding side mirror.

This sounds intuitive, but I found quickly that it added to, rather than reduced, the time required to check blind spots. One checks the mirrors, out of habit as much as anything else. Then one glances down at the instrument cluster to verify what one saw or didn’t see in the mirror. The bright red light is a strong reinforcement for the habit, and checking the indicators quickly begins to replace actually looking in the mirrors.

But I also learned that this is not a good habit. For some reason, when one passes a long truck at a low relative speed, the blind-spot software seems to decide that the truck is no longer really of interest and cancels the warning indicators. So you still must use the mirrors to verify that you are clear to change lanes in front of the truck you just passed. Trusting the red lights could lead to disaster.

Non-tailgating

My second adventure was with the smart cruise control. This system uses sensors to detect traffic in the lane in front of you, slowing the car before you approach closer than a preset, speed-dependent distance. In theory, a very good idea.

In practice, the system works exactly as specified, slowing the car to match the speed of whatever is in front of you in the lane to prevent tailgating. But there are issues with this. First, the sensors have a rather limited range. So if you are approaching another vehicle—a farm truck, say—at high relative velocity, once the smart cruise control identifies the vehicle it will brake rather decisively to prevent you from closing to what it considers an unsafe distance. This sudden maneuver may not be entirely welcome to the drivers behind you.

In California, their most likely reaction would be to angrily pull around and pass you, and then cut back in front. This, of course, would create a new object well within the smart cruise control’s critical distance, causing it to once again brake aggressively. And so on.

Waiting for a car to your right to pass, and then pulling out to pass the slow vehicle that had been in front of you brings an even more dramatic reaction. The smart cruise control identifies the car you just pulled in behind—the one that is currently moving away from you at a significantly higher speed than yours—as another target, and brakes aggressively. But now you have accelerated, changed lanes into faster moving traffic, and suddenly landed on your brakes. This behavior may also prove somewhat unwelcome on a California freeway.

Even if everything goes well, you are still likely to be a nuisance. The minimum following distance set by the smart cruise control is, shall we say, not in keeping with conventional driving practice in California. People will pass you simply because there is what appears to be about a five-mile gap between you and the car in front of you. Causing the smart cruise control to slow your car, etc.

Lane Departure

My final experience with this car’s ADAS complement was with the automatic lane-departure software. This module attempts to identify the lane edges from sensor data, and then takes control of the car’s steering—and, rather disconcertingly, produces tactile feedback to the steering wheel—to keep the car centered in the lane. Again, in principle this is a great idea, especially for fatigued or distracted drivers.

But once, again, the system suffers from limited sensor range. The system appeared to model curves as a series of line segments of about the length of the sensor range, and to select a straight course through each segment. This leads to the car taking a piecewise-linear path through a curve—a series of small but abrupt changes in direction that together keep the car in the lane. For drivers who are actually paying attention, this can be disconcerting.

But you had better pay attention. Because at times, the system appears to lose track of the lane edges during a turn. If this happens, the car will give no indication that it is abdicating. It will simply continue along the straight-line segment that it most recently chose, which, if you are in the middle of a curve, will take you straight off the road onto the shoulder. So if you intend to rely on the lane-departure feature, you had better be neither fatigued nor distracted, but able to very quickly identify what is happening, select a safe trajectory through the turn, and wrest the steering wheel away from the tactile feedback that the lane-departure algorithm is applying to it.

Some thoughts

My first impression is that each of the ADAS features I explored is in fact doing just what it was designed to do, within the limitations of the sensors upon which it relies. One rather suspects that the sensors were, however, selected on the basis of cost rather than on the basis of the performance required to produce a good driving experience.

My second impression revolves around that phrase, “driving experience.” I am not sure that anyone from the vehicle engineering team ever actually drove one of these cars on California freeway at 5 pm. Because with all the systems acting properly, the result is not an improved driving experience, but an additional set of challenges for the driver. Bad habits may form, and be strongly reinforced by the systems themselves. Fellow drivers will become angry. And in some cases, the ADAS systems may direct the car to do something actively dangerous, in the confident hope that the driver will intervene.

The problem here, if I am guessing accurately, is certainly not malice, or any attempt for marketing to mislead customers, or engineers to shirk their responsibilities. It is simply a yawning gap between the specifications for the individual ADAS functions and the real driving experience when the functions are operating together. Making the car perform a prescribed action in response to an identified set of inputs is very different from safe driving. The nature of that difference will depend on local driving habits, on road conditions, and on the style of the individual driver. But there will be an occasional discontinuity—sometimes a jarring one—between what the ADAS modules do and what the driver desires.

Editor’s note: Ron Wilson, a former colleague of Junko Yoshida when they both worked at EE Times, is today an independent technology analyst, writer and contributor to both Junko’s Tech Probe on Substack and Junko’s Talk to Us YouTube channel.