self-driving car

Children in Halloween costumes walk near a parked self-driving Google car to give its sensors practice. (Photo from Google’s blog)

Until the day comes (far in the future) when all vehicles are completely autonomous, even the best-designed self-driving cars will most likely need a human to take over when a situation occurs that the cars are not programmed to deal with, according to Willie D. Jones, who was discussing work being done by researchers in Munich, Germany, and the Netherlands, in an article for IEEE Spectrum.

Tactile Warnings

In the April 2016 issue of IEEE Transactions on Intelligent Transportation Systems, the teams, at Technische Universität München and the Delft University of Technology, wrote about their literature survey of possible “vibrotactile” ways an autonomous vehicle could alert a human that he or she needs to take the wheel. If a person is talking on a phone or perhaps watching a movie, it is possible that person could miss an audible or visual alert to take the wheel quickly to avoid a car accident.

Seats that give tactile warnings can be effective, said Sebastian M. Petermeijer, lead author of the article. These vibrotactile displays can be effective because they are hard to ignore. Petermeijer said:

In my opinion [they are] a potentially valuable addition to visual and auditory displays in cars, especially when drivers will be engaged in other tasks during highly automated driving.

Researchers have found that humans react more quickly to tactile warnings to apply the brakes than to such alerts as flashing lights, beeping sounds, or a computer voice. But even so, it takes about 0.8 seconds for a human to change his or her focus from, say, reading e-mail, to looking at the road. In addition to their effectiveness in getting humans to change their focus quickly, vibratory warnings (as opposed to visual or auditory alerts), also get humans to grab the steering wheel more quickly, the researchers found.

In their journal article, the researchers said:

Results showed that a distinction can be made between four dimensions for coding vibrotactile information (amplitude, frequency, timing, and location), each of which can be static or dynamic. There is a consensus that frequency and amplitude are less suitable for coding information than location and timing.

Transition Time

Interestingly, the teams found literature saying that the transition time a human needs to move from non-driving to driving can be shortened even more by vibrotactile technology that alerts the person to the direction of the danger. In other words, certain areas on a person’s body can be aligned with certain places on the road. In that system, perhaps vibrations on the seat under the person’s right hip would notify him or her that the danger on the road is coming from behind, on the passenger side of the vehicle, Jones writes. In another scenario, “a similar buzzing under the left knee could be directing attention out of the front windshield toward a broken tree branch hanging precariously over the left side of the driving lane.”

Too Much Warning? Or, Too Little?

Despite the benefits of tactile cues, researchers found that if there is a real danger or emergency on the road, relying on only one modality to alert the human is not best. However, there is also a phenomenon called an “alarm flood,” which means there are too many alerts at once, which could confuse a person.

The researchers suggested that each type of alert could be used in sequence. For example, a vibration might be first, then an auditory alert, and then a visual cue. These would correspond to directives to shift attention to the road, to reposition, and then to pay attention to the problem area and take action to avoid it.