The inertial microsensors might be an interesting complement in automotive land navigation. However, they have low accuracy and the principle of inertial navigation make the sensor's errors to accumulate, leading to unacceptable results. In this work, we are interested in the design of a strong signal processing system for error correction. We work on a strapdown configuration with low-cost off-the-shelf sensors, for which we define the navigation process. The signal processing is based on Kalman filters which is well suited for navigation applications. We propose a design which allows to take account of the bias as well as sensitivity variations. Usually, these calibration parameters are assumed to be fixed. We discuss the benefit of this approach and present the improvements.