General Motors is set to dramatically expand its fleet of autonomous Chevrolet Bolts in San Francisco, Detroit and Scottsdale, Arizona, according to documents filed with the Federal Communications Commission (FCC).
New short- and medium-range radars will be deployed in 300 self-driving Chevrolet Bolts, perhaps beginning as soon as next month. That will give GM and its autonomous technology subsidiary Cruise Automation the largest fleet of self-driving cars in the world.
Right now the largest such fleet belongs to Waymo, which says it has around 80 vehicles on public roads in the United States, with nearly 100 more arriving shortly. GM says it currently has more than 50 Bolts with autonomous technology in the three cities. The remaining 27 manufacturers testing in California together account for just 89 vehicles, although Ford, Uber and start-up nuTonomy at least are also testing elsewhere.
Here’s how IEEE Spectrum pieced together the story. Any organization that wants to use a radio frequency device that has not been certified by the FCC needs to apply for what’s called Special Temporary Authority (STA). This can include the millimeter-wave radar systems that many autonomous vehicles rely on to detect cars, pedestrians, and other road users.
On 17 March, Alps Electric, a Japanese supplier of automotive electronics, requested an STA for an experimental vehicular radar called Ukaza, operating at 76-81 GHz in the millimeter band. It wanted permission to test 3000 Ukazas, starting on 1 May. When asked by the FCC why Alps wanted to test so many, engineer John Cabigao replied: “The need for 3000 units is for testing and continuing development of the radar for vehicle use. As this type of automotive product does not yet exist in the market for public consumption, there are [sic] a lot of experimentation and testing that needs to be done on our side and with our customer to continue development.”
Although Cabigao did not identify that customer, he went on to explain that the radars would help it gather data to avoid collisions, test during rare traffic events, and demonstrate the technology. “Each vehicle will have 10 units so that [the] total number of vehicles tested is only 300,” he wrote.
Ten days later, on 27 March, GM’s Research Corporation filed a similar STA request with the FCC. It was also for experimental use of the Alps Ukaza radar, although GM requested that the quantity be redacted for confidentiality. “The number of units involved in the experiment can convey business sensitive information to competitors regarding future plans for services and technologies that have not yet been fully developed,” wrote engineer Robert Reagan III in GM’s filing.
Reagan has a LinkedIn page that describes his role as “[working] to ensure that our parts can do what they need to do and arrive at the assembly plant on time. For now, those parts are the radar sensors on the Chevrolet Bolt Autonomous Vehicle.”
GM unwittingly confirmed the quantity and destination of the Alps radar in a letter to the FCC this week. The 10 April letter described a meeting between Chris Murphy, GM’s lead counsel and chief privacy officer, and the new FCC Chairman, Ajit Pai. “Mr Murphy explained to the Chairman that GM is currently working with suppliers to test and optimize short-range vehicular radars for the development of self-driving vehicles,” read the letter. “Mr Murphy [also] discussed the Special Temporary Authority (STA) request made by GM’ s supplier Alps Electric North America filed in mid-March to temporarily authorize short-range radars in 76-81 GHz for testing purposes.”
However, it seems the Bolt is about to get more than just a bump to its short-range radar technology. On 28 March, a day after GM’s STA request, Bosch, another automotive supplier, made its own STA filing. This requested permission to test a mid-range radar for a Bosch customer’s “test fleet of highly automated driving (HAD) vehicles. They want to collect data with that fleet to develop their algorithms.”
The Bosch radars would be installed at each corner of a car, behind the fascia. Bosch sought permission to test 650 units, implying a fleet of up to 162 vehicles. It wanted to begin testing as soon as 14 April.
Like Alps, Bosch did not explicitly identify its customer. However, the company did specify the geographical locations where testing would occur. The latitudes and longitudes reveal that the radars are destined for Detroit, San Francisco, and Scottsdale, Arizona. The only company testing highly automated vehicles in all three locations is GM subsidiary Cruise Automation, with its self-driving Bolts.
In February Reuters reported that GM plans to deploy thousands of self-driving electric cars in test fleets in partnership with ride-sharing affiliate Lyft, beginning in 2018. A fleet of 300 vehicles could pave the way for a roll-out of such an autonomous taxi service, by allowing Cruise to collect huge quantities of data to train its systems and, ultimately, to validate the software with real-world driving experience.
But there are risks to growing a test fleet so quickly, says Karl Iagnemma, CEO of nuTonomy. “If you’re not ready to validate, there are significant diminishing returns,” he says. “Unless you’re really far down the road, you can collect enough data to keep your team busy with a far smaller fleet of cars.”
That matters because test fleets are expensive, requiring more than one engineer per car, according to Iagnemma: “The cost of keeping these cars on the road should not be underestimated.”
Bloomberg recently uncovered a public filing that shows GM will add more than a thousand employees to its autonomous research operation in San Francisco over the next four years.
A spokesperson for GM told Spectrum that “while we do not discuss potential future product plans, our plan is to launch our autonomous vehicle technology first in an on-demand ridesharing network.”
As of publication, the FCC had yet to grant any of the three STA requests for the new radars.
Contributing Editor Mark Harris is an investigative science and technology reporter based in Seattle, with a particular interest in robotics, transportation, green technologies, and medical devices. In 2012, he wrote an in-depth article for IEEE Spectrum on failures in AED defibrillators that won the Grand Neal Award from American Business Media. In 2014, he was Knight Science Journalism Fellow at MIT, and in 2015 he won the AAS Kavli Science Journalism Gold Award.