EEVblog Electronics Community Forum
Electronics => Projects, Designs, and Technical Stuff => Topic started by: cleaningOut on May 10, 2017, 02:31:52 am
-
...
-
Why not just create yourself a gateway, can to ethernet. that your project then talks to the vehicle through,
I would say it would be nonviable to set about converting every ECU and can connected module in the vehicle to ethernet and not have everything go horribly wrong.
Being a nissan there is likely to only be 1 -3 canbuses inside the vehicle
-
It is technically infeasible to change from CAN to Ethernet. Doubly so if you don't control the code base.
Even if you could sort out the multicast nature of CAN frames (CAN IDs are not addresses), you wouldn't be able to reconcile the error model and wake-up behavior.
There is no general rule about self-driving cars, but most appear to use CAN bus for operation. Sensors and actuators need either massive amounts of I/O or trivial communication -- there isn't a middle ground.
-
Beyond the technical process of converting from CAN to something else, I'd investigate the ethernet versions meant for safety-cricital applications as a good starting place. Self-driving may not to very interesting if it can't stop/turn/accelerate because some sensor is flooding the network.
-
The idea of a DIY 4,000lb autonomous machine capable of 70+MPH full of all sorts of hardware and software hacks is frankly, terrifying. Please do not attempt to operate such a vehicle on public roads, or anywhere near me.
-
I'd be amazed if this is not far more complicated than you think.
My current car has loads of different busses
4 x separate K-CAN
2 x separate PT-CAN
MOST (digital audio amongst other stuff)
Multiple (approx half a dozen) FlexRay
Ethernet
D-CAN
Multiple LIN bus
Numerous individual CAN busses
I'd be amazed if you could take control of a modern car without many and various interfaces.