An EU-funded project has created a car that learns from its driver – all very well, we reckon, as long as the driver is halfway competent.
The Drivsco program has used state-of-the-art sensors, image processors, and learning algorithms to create a prototype vehicle that tracks a driver’s actions, matches them with what it ‘sees’, and learns how its driver deals with situations such as curves or other vehicles.
Thanks to infrared headlights, stereo cameras and advanced visual processing the system can see better at night than a human driver, say its creators. In tests, it provided early warnings of hazards the driver hadn’t seen or reacted to.
“What we wanted was a system that learns to drive during the day by correlating what it sees with the actions a driver takes,” says its creator, Florentin Wörgötter.
“Then at night the system could say, ‘Slow down, a curve is coming up!’ – a curve the human didn’t see. Now we have a prototype that does this.”
The researchers drew their inspiration from animal visual systems. A key feature turns out to be constant two-way feedback between higher- and lower-level visual areas.
High-level visual areas store complex perceptions such as ‘car getting closer’ or ‘person crossing the road’ and send feedback that interacts with signals representing more basic features such as edges, colours, and movement.
Meanwhile, the car learns to drive by building up a huge database of correlations between what it sees and what the driver does, checking 20 times per second.
Currently, for obvious legal reasons, it limits itself to being a bit of a back-seat driver and warning about potential dangers. But in future, Wörgötter says, it could actually take control.
In tests, the system was able to produce consistent real-time predictions of how a particular driver would handle most highway or country road situations. City driving is still a bit too much for it.
“The ability to learn from a driver is quite new,” Wörgötter says. “I think it has great potential as a commercial product.”