Compensation Method of Gyroscope Bias Hysteresis Error with Temperature and Rate of Temperature using Deep Neural Networks

In this paper, a new compensation method for hysteresis bias error of the ring laser gyroscope (RLG) is proposed. Deep neural networks using temperature and rate of temperature is applied to obtain the RLG bias. In the process of entering the deep neural networks, temperature and rate of temperature are split into several factors for higher accuracy. Through entering these factors to the deep neural networks, more accurate estimation performance is achieved than simply entering the temperature and rate of temperature. The RLG bias estimating performance of deep neural network is evaluated through comparing with various methods - 3rd order function, classic rate of temperature method, and radial basis function network (RBFN). The experimental results show that the proposed compensation method has more precise calibration performance than others.