Self-driving vehicles need to be more like animals than machines to better adapt to unexpected situations on the road.
That’s the thinking behind a new type of autonomous truck programing developed by researchers at the Chalmers University of Technology in Gothenburg, Sweden.
“Biological systems are the best autonomous systems we know of,” said Ola Benderius, the lead researcher on the autonomous vehicle team at the Swedish university.
“A biological system,” he said, “absorbs information from its surroundings via its senses and reacts directly and safely, like an antelope running within its herd, or a hawk pouncing on its prey on the ground.”
After studying the behavior of human drivers, the researchers programmed a Volvo FH16 heavy-duty truck to mimic the reactions of an animal rather than function as a technical system.
Human perception and reaction differ greatly from how engineers have typically designed and constructed autonomous vehicle systems, Benderius said.
“We are trying to design a system that adapts to whatever happens, without pointing to specific situations,” Benderius said. “This is something that even the simplest animals can usually do better than existing vehicle solutions.”
The autonomous systems on self-driving vehicles are typically developed to handle specific issues. Their sensors observe a problem and a specific function is developed to treat the problem, he said.
Additionally, Google and other companies that work with autonomous vehicles try to create a digital map of the car’s surroundings to scan all potential problems. Benderius estimates this strategy captures about 99 percent of all potential dangers.
But it is that remaining 1 percent that has Benderius and his research team worried.
“You can cover a large number of different cases, but sooner or later the unexpected occurs, and that’s when an accident could happen,” Benderius said.
Their technology is meant to emulate how humans and animals observe the environment. Animals absorb information from their surroundings via their senses and reacting directly and safely to whatever is happening.
Information that the truck compiles from sensors and cameras is converted into a format that resembles the way in which humans and animals interpret the world via their senses, Benderius said. This enables the truck to adapt to unexpected situations in its basic design.
“Humans and animals can get a feeling of their speed and rotation by just looking at how the world flows around them while moving,” Benderius said. “Reflex responses are triggered whenever an object is quickly growing in our visual field.”
This way of processing information is crucial “to our own ability to be autonomous,” he said. “This is why we think autonomous vehicles also require the same possibilities.”
The project is being conducted in collaboration with Swedish company Volvo Trucks and Chinese-owned Volvo Cars, and is receiving financial support from the regional government.
Autonomous vehicles are set to play an increasing role in freight movement, said Jonny Morris, spokesman at Peloton Technology, a Mountain View, Calif., company that designs connected vehicle and platoon software.
“There is a growing realization that existing and emerging technology can improve safety and efficiency in the massive and economically critical freight trucking industry, both in the U.S. and Europe,” Morris said.
Research of the type that is underway at Chalmers is important for the industry, he said.
“Machine learning and reaction to sensor data are key building blocks to increasing automation in heavy trucks and other vehicles,” Morris said.
Benderius will get to test his self-driving truck programming theories against other autonomous vehicle software this weekend.
Chalmers researchers are entering the truck in the Grand Cooperative Driving Challenge in the Netherlands. It is a European Commission project where about a dozen teams will demonstrate how autonomous vehicles combine self-driving technology, vehicle-to-vehicle connectivity and vehicle-to-infrastructure communication.
On a stretch of the A270 highway between Helmond and Eindhoven, the vehicles will have to cross and exit at junctions and merge in and out of a line of vehicles, known as platoons.
Truck platoons are tightly contained, digitally connected packs of two to five vehicles driving in formation to reduce wind resistance and increase fuel efficiency. The pack is controlled by a “captain” in the front vehicle and the rest follow autonomously.
The teams also will demonstrate automated space-making for emergency vehicles in a traffic jam.