Performance Improvement of a MEMS Gyroscope Using Filtering and Machine Learning Methods

Gyroscopes are sensors used to measure angular rate and find wide application in the aerospace industry. MEMS Gyroscopes have various advantages of lesser cost, mass, size, power but with a corresponding decrease in performance parameters. Noise is a limiting factor in MEMS sensors leading to poor Drift Stability, higher Angular and Rate random walks which will in turn affect the output of the sensors. Also, the thermal effect on MEMS sensors is well-documented as a limitation to their wide deployment as sensors in uncontrolled environments. A novel digital filtering method is introduced and carried out on the data, and by means of trimming the filter parameters, processed data of different bandwidth was obtained. Statistical measurements were done on the processed data and 90 percent improvement was obtained in the short-term stability, which highlights the efficiency of the filtering process. A temperature control algorithm is introduced for operation of sensor at stable temperature. Also, a Neural Network based learning approach is used to train the data with respect to temperature as the major parameter. When it was tested with original sensor data, a 100percent improvement in short term stability was obtained.