PERFECTING THE PARTS
Pittsburgh at center of race to develop self-driving car components
The few miles of roadway between Pittsburgh’s Shadyside and Strip District neighborhoods are dotted with robotic cars — but it’s difficult to know how many.
The public recognizes Uber’s fleet of gray Volvo XC90s. Each is outfitted with a spinning top on the roof that resembles an oversized coffee urn. Inside the contraption, a laser constantly scans the environment around the car.
But other self-driving cars are unassuming vehicles that operate under the guise of a typical sedan or SUV, with hidden sensors. The British technology company Delphi confirmed that it has been testing its fleet of self-driving Audi SV5s in Pittsburgh. It would not say where or how many cars are being used.
The difference between the seen and the unseen is in the inner workings of an autonomous vehicle — its sensor suite using cameras, radar ,and light detection and ranging technology, called lidar, mimic human perception.
In Pittsburgh — a neurocenter for autonomous vehicle technology — companies vie to create the perfect suite of sensors. The right combination could be the difference between a metallic hunk atop a car and an expertly hidden array.
Uber’s self-driving efforts here are part of a crowd that includes Ford and its $1 billion investment in Strip Districtbased artificial intelligence startup, Argo AI; an emerging startup called Aurora Innovation led by former employees of Tesla, Google and Uber; and tech company Delphi.
“Pittsburgh is one of the core centers of talent right now for this industry,” said Sterling Anderson, a founder of Aurora Innovation.
At stake is not just the market for self-driving cars but also the ancillary markets in which companies hope to license custom hardware and software to tech companies and vehicle manufacturers.
The mother of all sensors
Competitors in the nascent industry each have a proprietary approach to computer vision, but most agree lidar is a mainstay for the ability to accurately scan a vehicle’s position in a three-dimensional space.
Lidar is not only expensive — about $75,000 for a manufacturer’s “off-the-shelf” product — but difficult to seamlessly integrate into a car prototype. Some companies, like Google’s self-driving car group Waymo, are designing custom hardware in an attempt to bring the technology to market quickly and at a lower price.
The lidar technology is so vital that Waymo took Uber to court in February, beginning a legal battle over trade secrets. Waymo contends one of its former engineers, Anthony Levandowski, downloaded 14,000 confidential files before joining Uber, according to a recent court order.
Judge William Alsup ordered Uber to return the stolen files by noon May 31 but did not grant a preliminary injunction suspending Uber’s research and development activities related to lidar. The judge said Waymo’s claims to the technology were baseless because lidar development involved wellknown principles of physics.
The case will head to trial later this year.
With the exception of Tesla, most players in the autonomous vehicles landscape use lidar, which uses pulses of light to send waves into the environment, reflecting off of objects and returning feedback on potential barriers. Lidar can define specific objects in any light condition, down to the lines on a street. But it is impacted by weather conditions.
A complete picture