Without taking away anything from the substance or achievement of this release, I find phrases like "openpilot is an operating system for robotics." always quite fishy.
No, it's not an OS for robotics. You can't do actual robotics stuff with it, like drive actuators to control limbs or grippers, do motion control or SLAM or perception or any of the usual robotics stack.
Their website correctly says openpilot is an open source advanced driver assistance system that works on 275+ car models of Toyota, Hyundai, Honda, and many other brands. Should've stuck to that.
Thinking about it some more, it's probably just another engagement baiting strategy to get attention and I'm their gullible puppet. Well played.
I definitely thought it was a ROS clone based on that first line.
ROS doesn't need a clone, it needs a successor.
Took the bait as well.
I can’t wrap my head around the fact that 275 car models include all the actuators needed for self driving driving and there’s some kind of port third party software can hook into.
I don't know if there's a physical port in most cars, but it uses the CAN bus which has been around since the 1980s.
Also, most cars that have distance assist and lane keeping probably have the required hardware to control speed and steering to some extent.
Nevertheless, it's still impressive that so many cars are supported... and that it can be retrofitted like this at all!
They also lie about models that are supported and won't assist you when you run into that. Had to return one myself. Found no evidence anywhere that my model/year was ever actually supported and anyone was using it either.
Sounds like they refunded you?
I suspect it may be even more than that theoretically. A lot of VAG cars based on the same platform are missing.
I chose the Hyundai Ioniq 5 as my current car specifically because it’s compatible with OpenPilot. It’s been a total game-changer for my driving experience. Just like their tagline says, “make driving chill,” and for me, it truly delivers on that promise.
This reminds me of Waymo's approach to self-driving cars. Paraphrasing, but basically they found that progressively adding self-driving to help human drivers is bad, because it leads to the humans becoming complacent and not paying enough attention. Therefore they decided on an all or nothing approach, where their cars would be only and entirely self-driven.
Is it like an app you install on the car or is it a custom integration?
It's a dashcam that you put on the windshield with 2 cameras pointing forward and one inward (filming the driver).
It’s my understanding that in addition to the cameras it also uses the sensors already built in to the car which would include blind-spot detection, no?
It says my car is supported and my car doesn't have any blind-spot detection, nor does the requirements list that as needed, so maybe it's optional but not required?
What is filming the driver used for? Can you disable that?
To ensure the driver has eyes on the road all time. It detects if you're not paying attention and beeps at you. I don't think it can be disabled.
I'm confused:
"THIS IS ALPHA QUALITY SOFTWARE FOR RESEARCH PURPOSES ONLY. THIS IS NOT A PRODUCT. YOU ARE RESPONSIBLE FOR COMPLYING WITH LOCAL LAWS AND REGULATIONS. NO WARRANTY EXPRESSED OR IMPLIED."
Where can this be used? In a private parking lot?
The driver takes liability, of course, and this can be used wherever the driver deems it safe and useful.
> this can be used wherever the driver deems it safe and useful
With the disclaimer that this depends on the location of course. For example, I think in Spain (and probably EU wide?) modifications that affect steering and throttle control would need to undergo local homologation before you're legally allowed to drive with that on public roads at all.
Which, to be honest, makes a lot of sense. I don't think anyone would be happy if cars start using software MVPs automatically controlling throttle and steering while in real traffic.
They’re just trying to scare away people who thing they can chuck this on their car and suddenly have a self-driving robot that they don’t have to pay attention to.
This seems to be a mix of C++ and Python, including a script called "realtime.py" (oxymoron?). So am I now exposed to other people using Python on the roads to operate heavy machinery?
Well NIST says don't use C++ either: https://www.nist.gov/itl/ssd/software-quality-group/safer-la...
So what, you want everything written in RUST on a linux kernel with hard real-time patches? It uses machine vision anyway, which has no hard guarantees at all. The software it uses to detect lanes or cars is probabilistic by it's very nature.
Python does pretty good at soft real time if you manage your own event loop and disable the garbage collector, and you're a lot less likely to get "crash the entire stack" style memory allocation bugs. Sure, GO or RUST would be better, I think CPP could be worse if handled inexpertly.
Python has segfault issues, surprising exceptions and version incompatibility.
I've been using Linux/BSD for over a decade now. No C or C++ application has ever crashed, I cannot say the same about Python applications. Outright segfaults are rare but happen. Rogue exceptions are much more common and could basically have the same detrimental effect on a self-driving system as a segfault. And let's not talk about logic bugs due to version incompatibility and the obsessive rewriting of those who took control over CPython.
Ahh, you've been running some grad students first python project as if it was a serious project like curl with 20 years of history, and expecting it to have the same quality. But you've somehow avoided the tons of grad-student CPP programs with similar quality issues, or the broken code pushed by companies like crowdstrike or IBM.
Fair enough, your experience may vary. I'd suggest not judging the language by the standards of some hobbyist code that just so happened to end up on github. I've had tons of bugs in c/cpp programs over the years, some more critical than others.
I've seen a lot of shitty and unreliable python code, and a lot of good and mature C/CPP projects. I've also seen really bad security issues and crashes with bad C code, heartbleed, crowdstrike, etc.
For what it's worth I've never had youtube-dl hard crash on me, and I could argue that it's a more complicated problem to solve than what curl is solving. In an apples-to-apples comparison I think it does pretty well.
No matter what language you use for this you're going to be relying on an AI vision model with no hard guarantees.
Is this legal in the US?
I am not surprised that comma is still around.
Minimal VC funding, less than 100 employees, not outrageously increasing headcount each month, profitable and sells a product with good margins.
Not many startups do this anymore, they are just chasing funding every 3 months using OpenAI’s API, comma has their own models before the AI hype.
If I was in law enforcement, I’d be rubbing my hands in glee to get ahold of that saved video.
for a second, I though this was the ROS alternative I've been forever waiting for smh
[dead]
[dead]