HN Explorer

Ask HN: Good resources for car hacking via CAN Bus?

1 points by ActorNightly ↗ HN
I was reading this about Mazda Kinematic Posture Control (https://www.passportmazda.com/blogs/1766/uncategorized/mazda-has-develops-new-technology-that-provides-the-mx-5-miata-with-better-stability-when-cornering/), which is basically a cornering aid where it uses brakes on the inside wheel to make the car turn better (basically like torque vectoring instead of just with brakes).

Considering most cars have had ABS for a while, I became interested whether or not its possible to develop this capability in a OBD plug in unit for any car. Basically, fool the ABS system in thinking that a wheel is spinning fast, causing it to apply brakes to that wheel.

The question that I have seems to be whether or not this is possible to do. For example, Comma AI units seems to be able to command brake and steering directly, however the supported cars all require some form of driver assistance, which means that they fake the signals from these driver assist systems. For cars that don't have this, it seems that Comma units will not work.

So is it such that a lot of cars without any driver aids simply do not have a way to reach the ABS unit via the CAN Bus?

Any good documentation or website on this stuff would be much appreciated as well.

2 comments

[ 6.3 ms ] story [ 20.2 ms ] thread
mdaniel
I saw that you mentioned Comma and that was going to be where I was going to point you because I doubt they have OOtB behavior for what you're describing but instead because I'd expect that kind of project has extensive CAN debugging tools, maybe even with onboarding docs for how to teach it about a new car or model

I guess that's a long way of saying, well, what have you already tried and what's the jam-up from your current approach?

It'll likely matter a great deal whether you're also trying to mod your daily driver (heh) or you have a play platform where "how about this?" doesn't cause fatalities if it segfaults some ECU

ActorNightly
I got a Comma Panda to start and have been playing around with gathering obd data. It works for my F150 truck to get raw packets, but the ODBC files don't seem to work (could be my fault though in using them). My truck is from mid 2010s so not supported by Comma, so I would have to do some reverse engineering to figure out the signals.

Im just wondering if this is even worth pursuing - if I can't for example even reach the ABS unit over OBD2 (for example, wheel speed sensors just tie directly into it without being exposed to the general network), then its not really possible. In the same way that Im guessing, if you don't have lane keeping assist in your car, the power steering system is a closed system, unreachable by OBD2.

I have a plan to do this safely (i.e run from laptop, only activate the program to test it out in empty parking lots).