Scientists from MIT taught robocars to get along with fools on the road

In Western countries, autonomous cars are already slowly appearing on the roads, but AI is still experiencing

difficulties trying to get along with ordinary drivers. The computer does not distinguish between the types of behavior of drivers, collecting them all “under one comb” - however, anyone who has ever driven a car knows that this is not so.

What can now?

New study from the computer labSciences and artificial intelligence at MIT solves the problem of the interaction of robotic machines with ordinary people at the wheel. Scientists forced the computer to predict their behavior on the road, using some tools from psychology. The system categorizes all drivers into two categories: selfish and safe.

Now the system studies the behavior of the setdrivers, constantly improving. The main parameters by which the system draws conclusions are how drivers keep their lanes and how they turn (do they assess the safety of the maneuver). The accuracy of the determination increased by 25% compared with previous experiments.

What should you work on?

Some inside look at the humanPsychology is simply necessary when robots and ordinary people share the road. Last year, for example, the Uber robocar knocked down a pedestrian to death due to the lack of systems to recognize extremely careless pedestrians. Now MIT is actively working on this issue.

</ p>