If cost of the final product is a concern, the "right" approach is to try and decipher the protocols (VPW, PWM, ISO, CAN) and do them on the duino. But it is a LOT of work, there are working examples of these in the public domain but they need to be tied together into a working model.
If you want results quick, then slap a 327 in there
I don't own any CAN vehicles, but that would be a good first protocol to get working, maybe a low cost atmel CAN based chip with an elm like interface to the duino would leave a clear path forward for the lower cost protocol specific elm chips?