Experts say driverless cars will require artificial intelligence
Automakers intending to bring driverless cars to market need to work as much on software design as mechanical engineering, the researcher leading Nissan Motor Co.’s automated-vehicle program said.
Making cars that are “deliberative” in assessing road conditions, rather than just reactive, requires artificial intelligence, Maarten Sierhuis, director of Nissan’s Silicon Valley research center in Sunnyvale, California, said in an interview. The carmaker, which aims to sell vehicles that can drive themselves by 2020 or sooner, is developing software to read and filter sensor data much as a human brain does, he said.
“What the auto industry has to come to is a shift from thinking about the car as a physical, mechanical system,” Sierhuis said in an interview Tuesday during the Automated Vehicles Symposium in San Francisco. “Autonomy, autonomous systems, is about understanding how humans do that, and then replicating it with software.”
Carmakers and parts manufacturers including Nissan, Toyota Motor Corp. and Daimler AG, along with technology companies such as Google Inc., are accelerating research into systems that can make driving partly or fully automatic. Benefits include eliminating traffic accidents and congestion, reducing fuel use, and opportunities for people to use the time in transit for activities other than driving.
Traffic accidents kill more than 30,000 people annually in the United States. While U.S. regulators see potential benefits for improved safety from autonomous vehicles, regulatory and legal issues such as liability in accidents have yet to be addressed.
President Barack Obama Tuesday highlighted U.S.-backed research on technology being developed by carmakers to allow vehicles to communicate with each other to reduce traffic jams and accidents.
Automation features such as adaptive cruise control, blind- spot monitoring and lane-keeping assistance aren’t sufficient, said Sierhuis, a computer scientist who designed software for NASA space missions.
“It’s a matter of also understanding the roads, understanding the situation, understanding other objects and knowing what to do with that information,” he said. “To plan your path around it needs deliberation — A.I. thinking.”
Nissan’s lab in Sunnyvale, with about 20 engineers, finds small Silicon Valley startups that can supply new types of sensors and components needed for the Yokohama, Japan-based company’s autonomous-vehicle program, Sierhuis said.
Google said in May it would put at least 100 autonomous cars it designed in trials starting this year. The two-seat cars have a top speed of 25 miles an hour and no steering wheels, brakes or accelerator pedals.
While Google hasn’t said if it will sell such vehicles, “I’m confident Nissan can have autonomous vehicles ready by the end of the decade,” Sierhuis said.