Application of BP-Bagging model in temperature compensation for fiber optic gyroscope

In order to improve the precision of fiber optic gyroscope (FOG), BP neural networks are widely applied in identification and compensation of FOG bias drift caused by temperature variation. However, the single BP neural network model is poor in generalization ability, which can affect the stability of prediction results. According to the ideas of ensemble learning, a neural network ensemble is developed to effectively generate the individual learner with strong generalization ability and great diversity by using Bagging algorithm, which has higher stability and accuracy in prediction, compared with the single BP model. A BP-Bagging model is established to compensate the FOG temperature errors. The traditional modeling method of linear regression and single BP neural network are also investigated to provide a comparison with the novel proposed model. The simulation results show that the BP-Bagging approach has better performance compared with those traditional models in compensation of FOG temperature drift and improvement of FOG accuracy.