State estimation based on kinematic models considering characteristics of sensors

Abstract

The major benefit of the state estimation based on kinematic model such as the kinematic Kalman filter (KKF) is that it is immune to parameter variations and unknown disturbances and thus can provide an accurate and robust state estimation regardless of the operating condition. Since it suggests to use a combination of low cost sensors rather than a single costly sensor, the specific characteristics of each sensor may have a major effect on the performance of the state estimator. As an illustrative example, this paper considers the simplest form of the KKF, i.e., the velocity estimation combining the encoder with the accelerometer and addresses two major issues that arise in its implementation: the limited bandwidth of the accelerometer and the deterministic feature (non-whiteness) of the quantization noise of the encoder at slow speeds. It has been shown that each of these characteristics can degrade the performance of the state estimation at different regimes of the operation range. A simple method to use the variable Kalman filter gain has been suggested to alleviate these problems using the simplified parameterization of the Kalman filter gain matrix. Experimental results are presented to illustrate the main issues and also to validate the effectiveness of the proposed scheme.