On supporting science journalism
If you're enjoying this article, consider supporting our award-winning journalism by subscribing. By purchasing a subscription you are helping to ensure the future of impactful stories about the discoveries and ideas shaping our world today.
What if your car knew what you were going to do before you did it? What if, say, the car slowed down by itself in an emergency in response to your reaction, even before you had a chance to hit the brakes? Such technology could shave precious instants off of response times, possibly averting collisions and saving lives.
A group of German scientists have now shown that such a predictive system is possible, if not yet practical. In a study published July 28 in the Journal of Neural Engineering, the researchers reported that electroencephalography (EEG) brain wave sensors and electromyography (EMG) muscle sensors on the leg can be used to detect braking intention before the driver depresses the brake pedal. Although some crash-avoidance systems already in existence can apply the brakes when the car's sensors detect an imminent collision, the new study was designed not to automate the braking process but to close the gap between the driver's decision to brake and the car's response. In a driving simulator, the sensors shaved about 130 milliseconds, or about 20 percent, from drivers' reaction times.
"Using this technology is a real gain over the normal behavioral response, using the brake pedal," says study co-author Benjamin Blankertz, a researcher in the Machine Learning Group at the Technical Institute of Berlin. "In a way, 130 milliseconds is not very large, but related to the reaction time, which is on average 660 milliseconds in our study, it's quite a large fraction. If you relate it to a car traveling at 100 kilometers per hour, that amounts to about 3.6 meters, or about one car length."
To predict a driver's intent to brake, the experimental system recognized a specific sequence of brain components: activity in the visual cortex identifying a potential hazard, followed by cognitive processing of the emergency situation, and finally culminating in motor-system activity preparing to move the right leg from the accelerator to the brake pedal.
The group found that a monitoring system of EEG electrodes on the scalp and EMG sensors on the muscles of the lower leg could accurately predict braking intent, but Blankertz says that more work needs to be done to prevent dangerous false alarms. For instance, a driver might recognize a hazard, process the situation and prepare to apply the brakes, all before deciding at the last moment that it would be safer to avert the hazard by swerving or by coasting into a gentler deceleration. "In this case, the brain signals would first indicate that there's an intention to do the emergency braking, but then another decision would be made," Blankertz says. Therefore, the brakes should only be activated after the driver has fully committed to braking, and where that decision point lies in the cascade of brain activity is still a bit fuzzy.
"How do we make sure that we do not detect premature states in the decision process?" Blankertz asks rhetorically. "Currently, we cannot decide from the brain signal where is the point of no return. We cannot really say when the final decision to press the brake took place."
So at the moment the research is merely a proof of concept, one that awaits further research on the cognitive signatures of decision-making. The concept also begs for a less cumbersome way of monitoring the brain. The experimental EEG setup in the study involved 64 electrodes, each of which must be attached to the scalp in a time-consuming process—hardly the sort of thing anyone would want to do before driving a few miles to the supermarket. Each electrode also requires a layer of conductive gel on the scalp, which introduces an additional hassle after the simulated drive is over. "The other effect is that after using it you have to wash your hair," Blankertz says.
The research is just the latest in a series of experiments into brain–machine interfaces; a few years ago Blankertz and his colleagues worked on a somewhat more lighthearted problem. "We tried to control a pinball with mind control," he says, adding that the experiment was a qualified success. "It works, it's better than chance," he says. "But it doesn't work well enough that it's very much fun."
