Bayesian Estimation and Tracking: A Practical Guide

20.8 Future Work

The current measurement model assumes that the IMU is located at the center of the rigid body's coordinate system (CG). To make the sensor fusion estimator more practical, a lever-arm correction should be added to the measurement model that compensates for the centripetal and tangential linear accelerations caused by rotations of the accelerometers about the body's CG. We also intend to show that using sensor fusion to analyze photogrammetry and IMU data together improves fault tolerance, in that dropouts in either data source are mitigated by measurements from the other data source.

In addition to increased accuracy and fault tolerance, sensor fusion can potentially improve photogrammetry turnaround time. Collecting 2D tracking data is the most time-consuming step in the photogrammetric data reduction process. We intend to investigate restructuring that process by using sensor fusion to reduce the quantity of image measurements that are necessary to produce a 6 DOF solution of sufficient accuracy. The UKF uncertainty model provides a measure of the accuracy of a 6 DOF solution for a particular event. We can compare the uncertainties in the sensor fusion result with the photogrammetry-only result, and determine how much less photogrammetric data needs to be collected to obtain sufficient accuracy.

Although the rate gyroscopes in the current generation of IMUs appear to be very accurate, the accelerometers appear to exhibit a large enough calibration bias to induce ...