Theoretically, it could be done using one of the adaptive filter techniques (RLS, LMS, etc) to generate an error signal (i.e the output from the RIO minus the input from the microphone). You then generate the correction signal to drive the error to 0 and add that to the output. It's actually pretty common: used on all phone lines for echo cancellation. Also, maybe someone used the BOSE headphones with active noise cancellation... same deal. Not sure how well it would work in the car given a much less controlled environment than headphones, but it's worth a try (if we can get to the DSP).

-T