Magnetometer-Free Orientation Estimation Using Multiple Inertial Measurement Units

Abstract

Higher-grade Inertial Measurement Units (IMUs) provide a better estimation of the orientation but are more expensive than lower-grade IMUs. The increase in availability and affordability of lower-grade IMUs over the last few years has provided an alternate solution: combine multiple lower-grade IMUs to increase sensor accuracy at low cost. Existing literature has already shown the advantages of multiple IMUs. One advantage is a better estimation of linear and rotational motion. Another advantage is Fault Detection and Isolation from component redundancy. However, existing literature is confined to a set of configurations that are impractical (platonic solids) or small (arrays). In this thesis, we focus on combining FDI with estimation of linear and rotational motion and further improve FDI by changing the IMU configuration from an array to scattered placement of IMUs. First, we show in a Monte Carlo simulation that because of the array configuration, the angular velocity and angular acceleration covariance decrease proportionally with the number of accelerometers used rather than quadratically, which is theoretically possible. We also reveal that for human motion capture, the angular velocity and the specific covariance cannot practically compete with a higher-grade IMU. Second, we propose a novel FDI method, based on the parity space method, to reject disturbances which can be applied to bigger configurations and high rotational motion. As a proof of concept for real-life implementation, we tested FDI on a simulated human arm with three configurations. The first with lower-grade IMUs placed evenly around the arm. The second and third are placed in the middle of the arm where the second is a lower-grade array and the third is a single higher-grade IMU. A disturbance was applied, which resulted in a 4 degree, 21 degree and 13 degree maximum yaw angle error respectively, illustrating the superior disturbance rejection capability of the larger configuration. In conclusion, we show that combining multiple IMUs improves accuracy and when used in a larger configuration, may be beneficial over a higher-grade IMU. The practical limitations are discussed in this thesis.