Errors in micro-electro-mechanical systems inertial measurement and a review on present practices of error modelling

Abstract

Micro-electro-mechanical systems (MEMS) technology-based accelerometers and gyroscopes are small size, mass produced, low cost inertial sensors, which are now being used in aerospace, underwater vehicles, automotive, robotics, mobiles, gaming consoles, prosthetic devices and many other applications. MEMS inertial sensors are available in many grades in market and selecting the appropriate grade sensor is very important. Owing to interaction of different types of energies, different noises are generated in MEMS devices; these noises cause significant change in output and the first section of this paper illustrates that. In application, where MEMS inertial sensors are used, the accuracy, repeatability and reproducibility of inertia measurement is probed primarily by complex testing, using extensive range of physical stimuli. Noises in inertial measurement are generally dealt by designing a unit measurement model. Noises are treated as additive error in linear unit model and are modelled using various techniques so that errors can be compensated to improve the accuracy. This paper reviews the theory, framework and methodology used in the error model of a MEMS inertial sensor and stochastic modelling of measurement. Experimental results from the most commonly used Allan variance techniques are discussed. Error modelling methodology, consisting of testing and calibration methods, designing thermal model, stochastic modelling and parameter estimation techniques, is illustrated. Figures and tables under each section summarize features, merits, limitation and future research scope. This paper should serve as a single reference for researchers and engineers working on application specific system design and instrumentation using MEMS inertial sensors. Conclusion from the study should help in selecting the appropriate grade of sensor as well as the best error modelling as per the trade-off existing between accuracy and development cost of error modelling.