Calibration of a 3D MEMS Gyroscope Using Shallow Neural Networks Under Laboratory Conditions

inertial measurement units (IMUs) offer short-term stability and high sampling rates but are prone to errors such as bias and scale factor drift, necessitating precise calibration. Traditional calibration methods, like six-position static test (SPST) and gyro rate test (GRT), achieve high accuracy but are often costly and impractical for real-time applications. Recent advancements in artificial intelligence (AI) present promising alternatives for IMU calibration. This paper introduces a novel AI-based three-dimensional gyroscope calibration technique using a shallow feedforward neural network (FFN). The proposed method effectively corrects constant errors, as demonstrated through simulations, laboratory experiments, and real-world dynamic scenarios. Results show improvements of up to 88% in attitude estimation compared to manufacturer calibration, positioning the technique as a viable alternative to classical methods. However, a limitation of this approach is the FFN's inability to decompose the IMU error budget into individual components, such as bias and scale factor.