Build Tesla-like touch screen car computer using a Raspberry Pi 3 | Hacker News Reader

Build Tesla-like touch screen car computer using a Raspberry Pi 3

autopi.io

28 pointsSpacemolte12 comments

HN users consider it spamming when people ask their friends to upvote and (especially!) post booster comments in the thread. So please don't do that.

https://news.ycombinator.com/newsfaq.html

My phone does 90% of this for any car I'm in even without the OBD-II interface. The few things I would like it to do that doesn't do well (in older cars) are:

   * Manage climate control
   * Provide good old fashion FM radio reception
   * Connect well to the amplifier via bluetooth

Beyond the basics like gas/brake turns signals and wipers, my IDEAL in car driver/passenger interface (built in or otherwise) does ONLY a few things

   * Provide a easily accessible tactile control for volume and climate control I can operate without looking down.
   * Show my current speed and efficiency (mpg) (RPM/Gear are ok if relevant)
   * Show me the mapping app of my choice on a screen I can see with good audio prompts
   * Provide a external antenna source for basic FM (GPS/GSM ok too) that is better then the phone can provide
   * Show a menu of all visible audio sources (bluetooth/wifi/aux-in) - Not something that takes 5 button presses to choose a different source.
   * Provide alerts for emergent issues with the car: Out of gas, Light failed, engine trouble.

Anything more then that is a distraction.

Edit: Don't get me wrong, Its early days for this project and if these things are in the roadmap I'm excited.

I picked up this for $18:

https://www.amazon.com/Foseal-Scanner-Adapter-Diagnostic-And...

There is code (not mine) for iOS here:

https://github.com/jeffmcfadden/CarDash

In its current state this doesn't appear to do more than a bluetooth OBD adapter paired with a cheap tablet can do - its really nothing like a Tesla touch screen at the moment, I guess the blog post title is just marketing.

Their use cases page however does a better job of showing what could be possible: https://www.autopi.io/use-cases/

That's exactly what I was thinking. I'm a competent mechanic (owned and fixed more cars than I can remember), and also electronically savvy; obviously modern cars have diagnostics via OBD2, but as you say, that's a world away from a Tesla-level control system.

Having had experience of various systems for more in-depth diagnostics (Ford, Vauxhall and PSA systems that I own, as well as the generic phone/bluetooth OBD code reader and diagnostic real-time display), I can't see that this will be easy to get sorted. Having seen the huge amount of work that a solution such as ForScan needs to get it running (and having spent some time feeding back issues to the creator), let alone something that is going to attempt to control a vehicle in everyday use, I think this will be a fair way away from being a real-world usable system - most cars just don't have the features documented, or even present.

The other thing is that I really don't think that a touchscreen is a good interface to use in a car. I'm a Tesla (and EV generally) fanboy, I guess, but this is one area that I don't get; you have to have your focus away from the road to do anything (which I guess won't matter when autopilot works fully all the time on any road) because you don't have tactile feedback and dedicated controls - you need to see where you're pressing to do something. There's a reason why cars' physical controls work so well, and I think a touchscreen (despite being really cool and futuristic) is the wrong way to go.

Somewhat OT but does anyone have experience hacking OBD-II? There are plenty of scanners that let you effectively read all the metrics, but actually issuing commands to get the ECU to do something isn't very documented. The protocol(s) itself seem fairly simple, it's about knowing the correct hex codes for your specific vehicle.

I've already gotten a computer in my car, so this is my next step. And yeah, the biggest challenge from my understanding is that there's some reporting standards required by law, and then once you have those, it's every man for himself on the rest of it. :P

George Hotz is specifically trying to solve this by crowdsourcing data from everyone's cars with his Panda project.

Haven't yet, but my starting point would be comma.ai's (and preceding projects) open source code. Pretty sure they interface with OBD

A must have !

This is actually nice! Have ordered one over kickstarter:)

"Ordered" or "invested in the idea"? It is definitely nice-looking, but I had only 1 in 4 items arriving from my Kickstarter "orders".

You're right - Ive invested in the idea. In any case it will be the right choice for my proof of concept, focusing on SW for CAN/OBD communication. reply

Reply on news.ycombinator.com