General Motors has announced that it will shelve its current plans for the custom robotaxi vehicle known as the “Origin” in a move that was criticized by now-departed Cruise founder Kyle Vogt but is probably the right path forward for the troubled division. GM will record a $583M charge on its books, and suffered a stock market hit.
Instead, Cruise will continue with a new version of the Chevy Bolt. The suspended Cruise pilot projects used a modified version of the original Bolt.
The Origin is/was a vehicle designed from the ground up to be a robotaxi. It’s somewhat large, and has face to face seating and sliding side entry doors. It has many similarities to the Zoox custom robotaxi (now owned by Amazon) though it is not back-to-front symmetrical. Notable features included a low entry easy for wheelchair adaptation, and a solid array of sensors with good visibility. There are not pedals or steering wheel, and it’s designed for a pleasant ride, not to drive.
For robotaxi boosters, it was initially disappointing to see such a forward looking project be suspended. The robocar business over the last few years has seen a strange mix of positive and negative news, including the dragging incident which led to Cruise being pulled from the streets of California by the DMV and the USA by the company. Cruise was in a scaled back mode, attempting to get back on its feet, and this highlighted whether now was the time to be producing the custom hardware.
The reality of robocars is that they are a very hard project. Easily one of the hardest software efforts in human history—still underway after 15 years of serious effort and tens of billions of investment by some of the most skilled teams in the world. While the hardware and sensors are important, the software is the bottleneck, and if you don’t have infinite money, the first step is to get your software project on a good path, and then make your hardware. That’s because you can make it work with existing cars—as Cruise and Waymo and several others are doing.
People in the computer industry have spent close to 60 years now under Moore’s law, which plots a path where computer chips get twice as good every year or two. You must plan with the expectation that the world will be different very soon, and then different again, in major ways. You delay decisions when you can, in both software and hardware, until you are ready.
If you have immense resources, you might do your car and software in parallel. If you pull that off, your new hardware will be ready when the software is. But you don’t know when the software will be ready, so there’s a big risk that you will design your vehicle and then as the software and world changes, want to change that design. Not all your work is wasted, but this is a costly way to do things. That’s why many people are critical of Tesla’s early decision to make their self-driving system work on their 2016 hardware platform. That promise (like almost all of Tesla’s predictions) was incorrect, but every other team decided to design for a car of the future, and spec out their future vehicle in the future, with the tools and knowledge of the future.
Cruise doesn’t know when their robotaxi service will go into production. So it’s too early to create the custom vehicle for it. However, the decision is not a slam-dunk. It takes at least 3 years, and usually more, for a car OEM to bring a new vehicle into production. So ideally you start designing it 3-4 years before you need it. At the time, it seemed to Cruise that they were in that position. Now, they are still working out their future.
What Cruise has learned on the Origin is not lost. When they are back on the roads, and ready to plan deployment again, they don’t have to start from scratch. Many of the concepts within it will endure and they’ll be able to make a new vehicle faster. It will have newer and better sensors, connectivity and compute. The size may change (the Origin is rather large, though a van-sized one also makes sense in completing a fleet.) They might decide Zoox’s symmetry so it can reverse direction is of value. They might decide something nobody has yet realized. These options will be open.
Zoox , as noted, was even bolder. That company was founded with the idea of only aiming for a custom “vehicle beyond the car.” They are right that you want that eventually, and that in a competitive market, the company with the better vehicle will have an edge. But in the early market, what matters is that it drives at all, not the comfort and convenience of a custom ride. Zoox ran out of money and had to sell itself—it definitely started custom design far earlier than needed.
Waymo is deploying a new robotaxi, built by Chinese firm Geely/Zeekr, but its design is not nearly as radical, and Waymo is the furthest along of the robotaxi companies.
It’s certainly big period for Robotaxi news, with Tesla pulling back the reveal of their robotaxi design (the software is still years from being ready) and Zoox expanding its plans, and Baidu/Apollo offering rides at prices much lower than human drivers to the point of getting protests from them.
