The Future of Autonomous Vehicles is Open-Source
A device the size of a smartphone, combined with open-source software, can make your car semi-autonomous.
It wasn’t until a few years ago that I discovered the sheer joy of cruise control. I knew about it, knew what it did, but I never had a dire need for it as I didn’t do a lot of long-distance driving. When I finally tried it out on a long stretch of highway, my entire perspective changed.
I imagine that’s how people feel when they first experience even a low level of automation in their vehicle. It’s both terrifying and infinitely exciting. While Silicon Valley continues to make promises and collect venture capital, a new solution shows us that, behind all proprietary solutions and closely guarded secrets, open-source solutions could give the industry a way forward.
An Outspoken CEO, and a Startup That May Be onto Something
George Holtz is a hacker and entrepreneur best known for being the first person to jailbreak the original iPhone when he was 17. He also managed to reverse engineer the PS3 back in the day.
These days, he’s the CEO and founder of Comma.ai, a company that offers a combination of hardware and free open-source hardware that promises level 2 autonomous functionality in a variety of modern vehicles.
When we say “level 2” we’re referring to the five levels of automation:
- Level 0 is your everyday vehicle with no automation
- Level 1 offers driver assistance like cruise control and lane assist
- Level 2 is partial automation, which requires the driver to still be engaged, but handles things like speed, steering, and braking
- Level 3 is conditional automation. This essentially allows vehicles to drive fully automated in specific weather and driving conditions
- Level 4 is high automation where vehicles are fully autonomous in specific use-cases. This is the stage where regulations, and a lack of unified focus on the technology to use are holding progress back.
- Level 5 is full automation. In this scenario, you wouldn’t need wheels or pedals, and humans would never need to interfere with the driving process.
While many publications and people promised autonomous vehicles on the roads by 2020, they are still far from mainstream. Comma.ai seeks to cut through the red tape and what Holtz calls “Press demos,” to actually put level 2 automation in the hands of the average driver.
They do this by selling a piece of hardware called the Comma Two that acts like a modified smartphone. It’s powered via OBD-C instead of a battery like the previous iteration.
By purchasing a separate vehicle mount, you can set up the device below your rear-view mirror like a dash cam. Once it’s plugged into the vehicles OBD-II port, you can utilize the OpenPilot software via a download.
As previously mentioned, the software side is fully open-source and available on Github. The hardware features both front and back-facing cameras, the latter of which has two infrared sensors to monitor the driver via facial recognition. The system notifies the driver if they seem distracted or if their eyes aren’t on the road.
If the vehicle is in motion, it can also decelerate if it detects a distracted driver. By utilizing the built-in RADAR and drive-by-wire systems in cars manufactured after 2012, it can turn the steering wheel and use the gas or brakes when engaged.
Data from the road-facing camera, CAN, GPS, inertial measurement unit, magnetometer, thermal sensors, and operating system is all set to Comma.ai servers via a cellular connection for machine learning purposes, which allows the software to improve over the course of development, based on specific scenarios and data sets.
The current iteration of the hardware and software, which made an appearance for the first time at CES 2020, has had positive impressions from those who tested it on the roads. Lane centering on roads without clear markings, along with navigating curved trajectories and maintaining safe distance from other vehicles are all high points.
While it doesn’t work on all road types and in all conditions, it’s a strong contender for what a decentralized ecosystem for autonomous vehicles could look like. The accessibility, combined with mainstream compatibility and an ever-improving open source software solution is a convincing argument for the idea.
While the autonomous vehicle industry and Silicon Valley continue spinning their wheels, open-source offers a potential path forward for driverless cars as a whole.
