Ideally we would have a throttle-to-speed control loop... that may be difficult to tune. We can try flooring then scaling with range....?

GPS headings seem to have less noise and match our estimates in the mean. We can do better with EKF, but why bother? Make it work before trying to optimize. TODO use the GPS values without offboard processing.

We see lots of outliers in GPS data. It seems to occur in all values read from the Venus - I think it's a serial parsing error, probably relating to the perpetual problems with timers. It may not be possible to fix this with the current set up, so we need to be able to reject these outliers. Store the start time and the last time to reject time. Similarly save GPS extrema values and reject any delta of more than a fraction of a degree...?

Possible solutions:
Put a cap in the signal circuit
Make sure voltage off Arduino is steady, maybe regulation needed
Use digitalWrite() with a delay, but this is not a favorite option