Study of grey model theory and neural network algorithm for improving dynamic measure precision in low cost IMU

The sensors' output data must be optimized because of the zero output varies along with time and temperature in the dynamic measuring accuracy of low cost inertial measurement unit (IMU). Two steps are done to achieve the designed precision. Firstly, the Grey model theory is proposed for the gyro's null drift output data process. Secondly, the RBF neural network is presented to compensate the gyro's null drift. Experiment proved that the mean variance of the zero drifting depresses from 0.0086°/ s to 0.0004°/ s and the deviation is only 30.8% of original sampled data, when the new error compensation algorithm is applied. The compensating algorithm raises the measure precision of IMU, whose static accuracy reaches to ±0.1° and dynamic accuracy is 1° (rms), and the cost is low.