What do self-driving cars 'think' before changing lanes?
Unlike humans, self-driving cars perform countless calculations before reacting.
Changing lanes is easy for human drivers. But cars are different. Instead of brains and brawn, self-driving cars make decisions using programming, artificial intelligence (AI), and on-board perception systems like lasers, cameras, and radar.
MagazinePopular Scienceasked four different companies to find the answer: drive.ai (a startup based in Stanford, US), nuTonomy (from MIT), Uber, and Waymo. The question was: "What are cars 'thinking' when deciding to change lanes?".
How self-driving cars deal with traffic situations is different from humans. Photo:Sinelab. |
Read traffic lights
Signals scattered throughout the environment provide important clues for a robot’s brain. A drive.ai car can understand that a car stopped at a green light is probably broken down. Algorithms can also deduce that turning on the hazard lights on a car ahead means it won’t be going any further.
Deduce
Cars and trucks parked on curves often block part of the road. With self-driving cars, the rule is that a parked car will stay within the lane limits. Smart cars from nuTonomy and drive.ai take that into account when deciding whether to pass. If the car in front is closer to a turn than the center of the road, the robot car may choose to go around it.
Know the way
NyTonomy, which has tested its self-driving cars in Singapore and Boston, has a knack for building up knowledge of urban streets. There are places, for example, where a nuTonomy car will be inclined to overtake a parked truck. But if it suspects the car is just stopping temporarily, it will wait behind it—a cautious approach that is quite typical in real life.
Share the street
The roads are not just filled with cars and trucks, but also with motorcycles, bicycles and pedestrians. A self-driving car can brake automatically if someone cuts in the road carelessly. Waymo cars can avoid bicycles thanks to a system of cameras, lasers and radar that recognize them, analyzing all the relevant data such as distance and speed to decide what to do.
Change lanes
Navigation is one of the best reasons for a self-driving car to change lanes. The team at Uber programmed its cars to check if they had enough distance from other cars before they acted. One key factor is the “two-second rule” for calculating safe distances.
Self-contact
A large vehicle can block the view of the road ahead for both human and robot drivers, so a self-driving car behind a truck can stay in place until a human intervenes. If the big car in front changes lanes, or flashes its turn signal, the self-driving car will try to pass. But technology now allows vehicles to communicate with each other, and in this way, a parked truck can share its status with nearby cars, like saying, “I’m parked, go ahead.”