Kalman Filter Measurement Offset Error

I am attempting to use a Kalman Filter to predict the position of a moving vehicle.
One of the instruments I am using to do so is an accelerometer. However, after first testing the Accelerometer independently I noticed a feature of the accelerometer that I was unsure how to deal with.
After thorough experimentation, I determined that the accelerometer had a consistent mean error of about .25 m/s^2, along with a variance of around .3 m/s^2.
However, in Kalman Filter Equations the process noise error is typically measured by the variable wk, as being zero mean, with a variance(covariance) of Q.
My question is this, how do you model the process noise assuming it has a non-zero mean error? The most common approach I've seen in these situations is to include an offset noise bk, that is meant to represent this non-zero mean.
In this case I believe that setting b = -.2 to offset this mean error of .2m would be sufficient. However, it think simply modeling the mean process error of .2m would be sufficent? Does anyone have any feedback>


Best,

--Greg

 

Are you locked in to that particular accellerometer device? That particular device might be defective, or it may be a model not suitable for your application. How have you verified the error presence??