Abstract
The affordability, small size, and minimal power requirements of microelectromechanical systems (MEMS) accelerometers makes them common in portable applications. The challenge with MEMS accelerometers is that they are subject to stochastic and deterministic errors. This work provides an examination of the benefits and limitations of average measurement-based homogeneous MEMS accelerometer fusion techniques. A gradient descent technique and two linear least squares approximation methods are also developed for accelerometer calibration. A new calibration method, called the moments technique, achieves low computational requirement and demonstrates similar accuracy to the gradient descent calibration method when accelerometer measurement biases are small. Predictive models of compound accelerometer measurement improvement are also developed. They exhibit good agreement with experimentally established trends and support the idea that simple compound accelerometer systems can be used to improve the quality of acceleration measurements in certain applications. However, this study also reveals an increasing trend in the relative error between the predicted and observed compound measurement improvement when more accelerometers are combined.