Compensation for FOG temperature drift based on adaptive neuro-fuzzy inference

The temperature drift is one of main factors influencing on the precision of a fiber optic gyroscope (FOG). The characteristic of FOG temperature drift was analyzed through temperature experiments. The zero bias temperature drift was compensated using the polynomial fitting method. To deal with the problem of poor compensation precision caused by the fact that the traditional surface fitting method can not describe the relationship between the scale factor temperature drift and the temperature or the angular rate accurately, a novel compensation approach for FOG temperature drift was proposed based on the adaptive neuro-fuzzy inference method. Based on the fuzzy logic, the approach combines the least square method with the back-propagation hybrid optimization algorithm to design an adaptive neuro-fuzzy inference system, so the compensation precision was improved effectively. The experiment results show that the training error root mean square and the predicted error root mean square of the new compensation approach are less than 0.003()/s and 0.005()/s respectively in the temperature range from -30℃ to -60℃ and the angular rate range from -165()/s to 165()/s