Cruise never planned make your own silicon. But in the quest to market robotaxis — and make money doing it — these never-planned pursuits can suddenly seem a whole lot more appealing.

Cruise realized that the vendor chips were priced too high, the parts were too big, and the reliability of third-party technology just wasn’t there, said Carl Jenkins, vice president of hardware at Cruise, at TechCrunch during a company tour. hardware lab last month.

Amid a hiring spree that began in 2019 and continued in 2020, Cruise dubbed his own hardware, including his own map and sensors. The investment helped the company develop smaller, lower-cost hardware for its vehicles. It also resulted in its first production map, the C5, which powers the current generation of self-driving Chevy Bolts.

When the company’s purpose-built Origin robotaxi begins hitting the streets in 2023, it will feature the C6 board. This card will eventually be replaced by the C7 which will have Cruise’s Dune chip. Dune will process all sensor data for the system, according to Cruise.

Typically, automakers use parts and sensors from Tier 1 suppliers to reduce R&D and manufacturing costs. Cruise couldn’t see a way to launch his autonomous homing without doing more work himself. The result is that the C7 board is 90% cheaper, has a 70% mass reduction, and uses 60% less power than vendor-provided chips.

It’s not just the chips that are supported by the company. While long-range lidars and ultrasonic sensors are always sourced from third parties, almost everything else, including cameras, short-range lidars and radars, is also developed in-house.

Cruise found that standard radars simply didn’t have the resolution they needed for their vehicles to work. Like the board, there’s a long-term cost reduction of about 90%, according to Jenkins.

“I’ve been told the price I need to hit with this hardware for 2025,” Jenkins said. “So I went to all the CTOs from Bosch, Continental and ZF in Germany. ‘What do you have in your research tanks that answers that? Nothing, not even started.’ Okay, so you start today, how long should I take?’ Seven years.”

By then, Jenkins was able to grow his team from 20 to 550.

Asked about the costs of building the Origin with in-house developed hardware versus parts sourced from suppliers, CEO Kyle Vogt told TechCrunch, “We couldn’t do it. It does not exist.

That’s not to say Cruise doesn’t want to be able to buy the gear he needs, though.

“What we found in the AV industry was that many components that had the ruggedness to operate in a harsh automotive environment did not have the capabilities needed for an AV. Components that had the capabilities (AV ) needed were not able to perform in these harsh environments,” Vogt said.

Made at Cruise, used at GM?

Automakers (not including Tesla) have taken a more cautious approach to self-driving vehicles that would be sold to consumers. Technology built and proven by Cruise could eventually end up in a GM product sold to a customer.

And there are reasons to believe that it will.

GM CEO and President Mary Barra has repeatedly said the automaker will make and sell personal autonomous vehicles by the middle of the decade.

“We use Cruise as a yardstick for us for autonomous vehicle technology and the stack and how it works,” GM President Mark Reuss told TechCrunch Editor-in-Chief Kirsten Korosec in a recent interview. As Cruise develops its AV technology, its parent company has focused its efforts on the Super Cruise and now Ultra Cruise advanced driver assistance systems.

“When we start researching and looking at personal autonomous vehicles, there are choices like does the car have pedals or does it have pedals that are deployable or does it not have pedals at all,” Reuss said. “And so we look at what people want and those are not easy questions to answer.”

A few years away from its mid-decade goal, GM still has a lot of work to do, including its strategy to bring these personal autonomous vehicles (or, as Reuss calls them, PAVs) to market. Feedback from its recent InnerSpace autonomous concept for Cadillac

GM has not decided whether these PAVs will be launched as a premium product or whether they will be attached to an existing vehicle model or a dedicated vehicle, Ruess added.

bumps in the road

Picture credits: Roberto Baldwin

Cruise currently runs a self-driving rideshare business in San Francisco, but only in the middle of the night (10 p.m. until 5:30 a.m.) and only within a 30% radius of the city. The company notes that this decision was based more on ensuring its vehicles work during less hectic traffic periods. He is currently working on expanding these area and time constraints.

It’s not just San Francisco that will see more driverless Chevy Bolts carrying passengers. Cruise plans to expand to Phoenix, Arizona and Austin, Texas within the next 90 days.

Scaling is Cruise’s next chapter. However, the hiccups keep coming. There have been several reports of robotaxis Cruise blocking intersections and other issues.

A vehicle was involved in a collision at an intersection, prompting the company to update the software on 80 of its vehicles. In April this year, a Bolt was pulled over for not having its headlights on and at one point pulled away from the policeman. And of course there’s the infamous band of over half a dozen Cruise Bolts that were assembled at an intersection and unable to figure out where to go next, causing traffic problems.

Asked about the grouping of vehicles, Vogt noted: “It’s part of the operation, part of the separation of the scaling. It’s a normal bump in the road. The CEO noted that this was an inconvenience and not a security issue. Vogt said AVs have a lot of back-end services, and one of them “tipped over” and didn’t come back online fast enough. How they all ended up in the same intersection is that at the time there was only one launch location for vehicles and they navigated along one of their main corridors near from this launch location. Since then, Cruise has incorporated resilience techniques into the AVs to make them more tolerant.

The company (and by extension, Vogt) is confident in its in-house built autonomous call system. Now he must convince skeptics that a ride in a self-driving vehicle is worth paying for in cities outside of tech-friendly San Francisco.

Our driverless journey

At the end of the tour, Cruise arranged a self-guided ride for us in a Bolt.

Our vehicle, nicknamed Ladybug, arrived and with a tap of the app we unlocked the doors and cruised (no pun intended) through the night into the city towards Japan Town.

Along the road, several vehicles were parked with their driver’s side doors open. The Bolt slowed slightly, flashed its turn signal, and slid briefly into the other lane before falling back into its own. At four-way stop intersections, he took on the persona of a cautious human, only retreating after determining that other vehicles would obey the rules of the road.

It was exciting at first, then boring, which is what driverless transportation should be all about. Yes, it’s a little weird being in a car driven by a robot, but after 20 minutes of being transported by a cautious robot, the last 10 minutes are spent wondering if you’re going to get stuck at a fair intersection to add some excitement to the ride.

Additional reporting by Transportation Editor Kirsten Korosec.