Flight Attitude Testing Based on the Neural Network

In order to obtain attitude parameters, this paper proposes the new neural network model based on the theory of information fusion. After researching the features of the gyro/geomagnetic sensors, the model of BP network has been set up, and problems of bigger navigation error and lower angular speed accuracy have been solved. The output of gyro/geomagnetic sensors as input of the model, and real-time three axis angular speed as output. Through a large number of learning and training, the higher precision of angular speed has been gained, attitude angles have been obtained too by the approximation formula method, and systematic error has been reduced. The experimental results, coming from the no-magnetic turntable, show that the algorithm has quick convergence speed and high accuracy on forecasting the angular velocity.