Special Thermal Compensation Experiment and Algorithm Design for Inertial Navigation System

The inertial sensors in the inertial navigation system (INS) are susceptible to the ambient thermal field and these temperature-induced errors can be the principal factor that decreases the applied precision of the system. Software compensation is an effective method which can predict and eliminate the thermal errors directly from the sensor outputs. But it is a technical challenge to compensate the errors at high accuracy as well as make it adaptable to the complicated time-varying thermal environment. To overcome this, excellent temperature experiment and algorithm design are both indispensable and equally important.Among all the thermal errors, gyro bias drift tends to be the key problem due to its divergent influence on position error and nonlinear characteristic with temperature. To tackle this, two kinds of thermal experiments are specially designed to fully stimulate the drift behavior under different temperature conditions. One is called the constant gradient changing experiment, which inversely uses the thermal chamber-gyro transfer function to control the gyro temperature to change at the constant rate and is to reinforce the effect of temperature derivative in the experiment dataset. The other is called the thermal phase plane covering experiment, which contains a special temperature profile to manipulate the bias error to “walk around” the “thermal phase plane”. When the two experiments are put together, it demonstrates an obvious effect in enhancing the compensation's adaptability. In the modeling process, a nonlinear estimation algorithm is applied to precisely identify the bias error model. It is able to determine the weight of each kind of thermal factor and their influence mode on the total error in different temperature situations. For the accelerometer, we focus on all-parameter modeling, including the scale factor and bias of the I/F converter, and the scale factor, bias and misalignments of the accelerometer sensors. A system-level error model is deduced to separate the coupling of the I/F converter and accelerometer sensors. And thermal error terms of the accelerometer sensors are estimated and modeled by the continuous temperature calibration. Verification experiments on the INS with fiber-optic gyroscopes and quartz pendulous accelerometers show that the thermal environmental adaptability and navigation performance can be effectively improved.