This research is not quite ready for the market, but has a lot of potential to improve driver safety.
The ability to predict what a driver is going to do in the near future and to be able to prepare the car’s system for this sounds a little bit like science fiction, and it would naturally be a dream come true for the safety departments at car manufacturers. The dream is now one step closer to becoming reality.
Researchers at Chalmers University have made an order of magnitude improvement in the modeling of driver steering using continuous and linear control model. What makes this so revolutionary, according to the authors, is that it includes jerkiness in the control signal. Previous research has modeled velocity and acceleration, but not jerk to this level of precision.
It can identify critical events, model what the driver is doing, and predict what will happen next. Examples they discuss are drivers falling asleep or overcorrecting from a swerve. If the predicted result is unsafe, it can apply some kind of intervention. The intervention is not part of the research, but it could be anything from an alert, to a dampening of the movement, to a reversal of the movement.
This one is apparently closer to market ready. The Seeing Machines system uses cameras on the driver’s eye and facial movements to model drowsiness and attentiveness. They are currently targeting heavy machinery but plan to hit the consumer market this year. Unfortunately, they were negotiating with Takata (of exploding air bag infamy) so I am not sure if that will delay things.
I highlighted that this isn’t ready for prime time because no matter how good the model is, there are some affective challenges. If you are driving and make a jerky movement, would you be OK with the car taking over? If you had fallen asleep? If you had swerved to avoid an animal in the road?
And then we also have the liability question. If the intervention leads to some kind of damages, who is responsible? You, the car manufacturer, or the automation vendor? And what if it saved you from a worse crash, but resulted in a different crash? This is America, you can sue anyone for anything. The lack of case law creates uncertainty that companies shy away from .
What do you think? I don’t know enough about the human performance modeling science to know just how groundbreaking the model is, but the article indicates it takes us pretty close to a version that would be technically feasible as a steering assist feature for a car. Are we ready?
Image credit: Shuets Udono