Tracking Using Fusion of Multiple Inertial Measurement Units

The objective of this thesis is to determine the effectiveness of fusing data from multiple inertial measurement units (IMU) to reduce bias and noise with an overall goal of achieving accurate tracking, which is the process of locating a body as it moves in a fixed environment. Every sensor is subject to noise, and each sensor has its own unique set of biases. The thesis presents a systematic overview of the sensors used in this research, which feature linear acceleration, angular acceleration, and a directional sensor, to form an inertial navigation system (INS). Sources of noise and bias that affect utility of the IMUs as well as data processing algorithms used for estimation and filtering are also presented. Related work in the topic area is summarized. Finally, seven experiments, which evaluate the accuracy of the acceleration measurements and overall displacement from both the fusion method and the raw single-sensor method, are presented. The accuracy of the acceleration measurements is evaluated by comparing the sensor measurements to the known theoretical acceleration using common statistical metrics. Tracking accuracy is evaluated by overall displacement accuracy and path displacement accuracy. It is found that multiple sensor fusion is not always capable of estimating the overall displacement more accurately than a single sensor. Additionally, fusion increases the signal-to-noise ratio of the accelerometer data. However, our results indicate that neither the fusion nor the single-sensor method are capable of accurately estimating the displacement path.