In a recent video from Comma, we learn how it arrived at the design for its latest hardware, the Comma 3X. For those unfamiliar, Comma takes a very different approach compared to other companies working toward autonomous vehicles. Instead of designing the software and the hardware, and keeping the whole system inside a proprietary black box for maximum future profits, Comma supports an open source effort. Anyone can contribute to the project, and people can even make their own custom versions of it (or “fork” it), and there are over 6,600 of these forks.
Here’s the video, where they explain how and why they designed the 3X the way they did, and how it fits into the wider effort (article continues below the video):
One of the first things Robbe reminds us is that Comma is not an AI company. At the core, it is a consumer electronics company first. The company doesn’t fund itself with sales of expensive software bundled with hardware — it only sells the hardware to run openpilot, which funds the company’s efforts.
To get where they are, and to get where they’re going, they follow a product lifecycle, which he gives a nice flowchart for. They sometimes have to go through several iterations or prototypes (sometimes 6-7 times) before arriving at a final design. Even after a final design, there will still be warranty returns after it reaches customers.
The first version of the company’s hardware was the EON, which wasn’t that different from a smartphone. It had a heatsink, a 3D printed case, and was setup for automotive use, but was still fundamentally a smartphone on the inside. The second version was the Comma 2, which was a refined version of the EON. Things were more complex, and the added complexity was meant to help support more vehicles than the original EON.
The Comma 3 (not the current 3X) was a big change from the “improved smartphone” approach earlier versions were. Instead of repurposing existing smartphone hardware, they instead designed everything from scratch based on what the company had learned in its early years. Robbe says that this was the hardest product the company has built, and may be the hardest thing they’ll ever do with hardware, but it sets the stage for the company’s future for the next several years.
Now, with the Comma 3X, he says they’re on the “downslope” of complexity. They plan to work with manufacturers to simplify what’s offered inside the device and lower costs, and they hope to arrive back somewhere closer to a smartphone, albeit one that’s tailored for driving cars an integrating into future cars, like the Aptera.
The concept of the 3X was pretty simple. It kept the same basic form factor as the Comma 3, but with iterative improvements and removing unneeded parts to get costs down. He quoted Elon Musk’s “the best part is no part” idea here, showing how they’ve managed to get the price down to $1250, or around half of what the Comma 3 cost.
What was left was also changed and upgraded. The circuit board remains complex, because it is a car-driving computer, but was made to be simpler and cheaper than the previous one. This process required a lot of manual work, as there were a lot of things to think about to keep it all working right and not conflicting, interfering, or making radio noise. Future versions should be easier, as they’ll be able to iterate on the current design more.
The 3X has a total of 136 unique parts, compared to the 203 parts the Comma 3 had. But, that’s still a lot to keep track of both in design and in terms of inventory, ordering, and substituting as needed. Shortages are still a problem, so they needed to be ready to change parts to stay in production. They use a number of software tools to not only track all this, but check for errors and omissions.
During the revision process, they use different colors for each new version to avoid confusion. This helps testers not run into problems and mixups. They do this with in-house manufacturing because they can iterate faster and make changes quicker. This shaves days off the process of each small revision. He goes through the whole manufacturing process, but I won’t recap that here.
The big advantage of doing every step in-house is that they are able to take advantage of vertical integration, and in a somewhat automated fashion. By being able to track and control the whole process, changes and adaptations for different parts suppliers are a lot easier. This both speeds up and increases quality of production.
After testing and compliance measurements in certified labs, field testing sometimes brings them back to the beginning of the process to start all of this over (and over and over). This doesn’t stop until they have something that passes all of the testing and allows them to move on to mass production. But, even then, they still track everything in case of failures so they can try to figure out what went wrong.
Next, they have to finish the assembly, provision the boards, test camera focus (they use fixed focus lenses), and make sure everything is ready to go for final assembly for the customer to turn a board into a full Comma 3X and final provisioning. This involves final testing, firmware upgrades as needed, and registration of the product in the computer. Sometimes, they add additional testing at the end of the process if they notice common failures.
Final devices all get put into a stress test before they ship out. They keep the computer running hard and hot for 24 hours, watching to avoid all common failures, followed by a final inspection before shipping out.
After all this, the chances of getting a “DOA” product or the wrong product from Comma should be extremely low. But, they still have an improved process where they can take a closer look at the product before they have the product shipped back. This has reduced ship-backs by “close to an order of magnitude,” and this allows them to focus harder on the products with actual problems and defects.
Between all of these process improvements, they’ve managed to cut the whole failure rate down to 3.9% (less than half of their all-time failure rate), but they’re aiming for 1-2%.
But, the development and manufacturing process is just one of many things the company does. You can learn more about other aspects of the company at its YouTube channel. Hardware and information about compatibility can be found at its main website.
Featured image from Comma.ai.
Have a tip for CleanTechnica? Want to advertise? Want to suggest a guest for our CleanTech Talk podcast? Contact us here.
EV Obsession Daily!
Tesla Sales in 2023, 2024, and 2030
CleanTechnica uses affiliate links. See our policy here.