A Real-time Thermal Compensation Method for a MEMS Accelerometer Based on SRU Neural Network

MEMS inertial sensors have become more and more popular in recent years due to their low-cost, small size and low power consumption, and have been widely used in various fields, but they have the disadvantage of being very sensitive to temperature changes (thermal drift). We propose a real-time thermal compensation method for a MEMS accelerometer to provide an online compensation strategy to mitigate temperature effects in this paper. First, this study uses an in-house-designed Guidance Navigation and Control Module (GNC Module) for temperature effect testing, and the output of the accelerometer is collected during the test. Second, the real-time compensation model based on the accelerometer output at the previous moment is proposed. And then, the false nearest neighbor algorithm is introduced to determine the number of neurons in the hidden layer of the neural network. At last, a recently popular Recurrent Neural Network (RNN) variant, the Simple Recurrent Unit (SRU-RNN), is used to process the MEMS accelerometer raw outputs. With temperature calibration, the thermal drift of the MEMS accelerometer is significantly improved, which undoubtedly proves the effectiveness of the employed method in this paper.