A Unit Quaternion and Fuzzy Logic Approach to Attitude Estimation

摘要

With the ever increasing need to develop accurate, low cost systems to estimate attitude, MEMS IMUs have become popular. MEMS IMUs are inexpensive, but have poor drift and noise characteristics when compared to navigation grade IMUs. When uncompensated, the integrated attitude solution from MEMS IMUs tends to drift with time. MEMS IMUs need some form of active bias estimation to compensate for bias drift and reduce random walk in the attitude solution. Given fairly accurate sensor noise models, Kalman filter based algorithms have been used to derive bias estimates and an optimal attitude estimate. This paper introduces a new algorithm for attitude estimation with low cost inertial measurement units for aircraft applications. It is assumed that three axis angular rate and acceleration measurements are available to compute roll and pitch attitude. It is also assumed that magnetic heading is available from an independent magnetic sensor. The algorithm uses a unit quaternion to represent aircraft attitude. Estimates of the attitude and gyro biases are obtained through initial alignment under non rotating flight conditions. The angular rate data after initial alignment has a residual bias component that makes the attitude solution diverge. Reference unit quaternions for the attitude channels are constructed from accelerometer and magnetic sensor data. The reference unit quaternion is corrupted with linear acceleration and cross product terms, and cannot be relied upon, under all flight conditions. The initial estimates of the attitude quaternions for the roll and pitch channels and gyro biases are updated with an angular rate correction estimator. To counter the effect of the time-varying gyro bias and the corrupted reference attitude quaternion, fuzzy sets that adaptively adjust the gains of the estimator designed. Criteria for the fuzzy error and rate of change of error are defined, and the membership functions are designed.