Need pointers for a computer specs for a car for possibly level 5 autonomy.

[bulgedition]

Budget (including currency):  $1500, maybe $2000, yes $2000

Country: Bulgaria, what does it have to do with the question?

Games, programs or workloads that it will be used for: mostly self driving, object detection etc.

Other details (existing parts lists, whether any peripherals are needed, what you're upgrading from, when you're going to buy, what resolution and refresh rate you want to play at, etc): I think I am an experienced software engineer, self thought, and decided to start a new hobby - build a self driving car from scratch, I mean the software is going to be from scratch, the car maybe second-hand one.

The goal is to achieve level 5 autonomy.

I need help for building the computer for the car. Can anyone give me some pointers, please?

 

[Eigenvektor]

17 minutes ago, bulgedition said:

I think I am an experienced software engineer, self thought, and decided to start a new hobby - build a self driving car from scratch, I mean the software is going to be from scratch, the car maybe second-hand one. […] The goal is to achieve level 5 autonomy.

You do understand that companies like Waymo (Google/Alphabet) and many others have put billions into this goal for several years already, right? And we're nowhere near full level 5 autonomy yet. Sorry, but this really is not something you can build in your garage on a shoestring budget.

[Letgomyleghoe]

honestly if you don't have the program written yet, we can't really give you advice. If you have no idea whether the program is cpu/gpu/ram bound there's no point in speccing out a build. supposing that you can build the car/ software in the first place.

24 minutes ago, bulgedition said:

Country: Bulgaria, what does it have to do with the question?

lets us check part availability and pricing in your region.

 

[PineyCreek]

Not to mention that regardless the computing power behind it, you'll likely be spending more on self-driving actuation (to move steering, accel, brake, signals), car health monitoring and a large amount on video cameras, scanning equipment (radar/lidar, stuff like that).

 

Why not start with some form of RaspberryPi-based connection via your computers diagnostic port to write your own code to watch and understand the car's health and plan for ideas to integrate that telemetry into the bigger picture of the self-driving car?

 

Or, go with a SBC-based car and learn about machine vision and pathing.  There's plenty of Python code around that.

 

Get the structure of your project figured on a smaller scale and implement on a smaller scale first...both for learning and cost...and potential liability.  A tiny battery-powered RC car with an SBC controlling it will still cost a fair chunk of change, but you'll learn plenty of information that could be applied at a larger scale...and also the result would have a far smaller likelihood to cause damage/death if something went wrong.