This chapter presents a complete development of a complementary filter-based attitude estimation scheme suitable for implementation in a small, low-cost unmanned aerial vehicle. The filter is based on strapdown gyro integration using a matrix exponential closed form that preserves the rotation matrix orthonormality. Bias drift is removed using accelerometers and GPS course over ground (COG) as external aiding sensors. The accelerometers are used to estimate gravity in the body frame, removing both the longitudinal acceleration and the centripetal acceleration components from the measurement. The cross product between the measured and estimated values is used to generate an additional rotation rate fed into the integration. The GPS COG is compared to the vehicle heading in the inertial frame, and again a cross product is used to induce an additional rotation rate. The complementary filter is set up as a proportional integral filter, with the proportional gain correcting the current estimate, and the integral term serving to estimate the gyro biases. Calibration techniques for each of the sensors are discussed, and a block diagram of the entire filter is provided. The filter has been shown to be computationally efficient, simple to program, and robust in practice. It shows similar performance to a 15-state EKF but at less than 1/3 the computational load.
A Comparison of Multisensor Attitude Estimation Algorithms
G. De Maria;C. Natale;S. Pirozzi
2016
Abstract
This chapter presents a complete development of a complementary filter-based attitude estimation scheme suitable for implementation in a small, low-cost unmanned aerial vehicle. The filter is based on strapdown gyro integration using a matrix exponential closed form that preserves the rotation matrix orthonormality. Bias drift is removed using accelerometers and GPS course over ground (COG) as external aiding sensors. The accelerometers are used to estimate gravity in the body frame, removing both the longitudinal acceleration and the centripetal acceleration components from the measurement. The cross product between the measured and estimated values is used to generate an additional rotation rate fed into the integration. The GPS COG is compared to the vehicle heading in the inertial frame, and again a cross product is used to induce an additional rotation rate. The complementary filter is set up as a proportional integral filter, with the proportional gain correcting the current estimate, and the integral term serving to estimate the gyro biases. Calibration techniques for each of the sensors are discussed, and a block diagram of the entire filter is provided. The filter has been shown to be computationally efficient, simple to program, and robust in practice. It shows similar performance to a 15-state EKF but at less than 1/3 the computational load.I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.