Bayesian Estimation and Tracking: A Practical Guide

20.4 The Generation of Synthetic Data

In order to evaluate the performance of the various tracking methods, in Chapter 19 we developed a synthetic rigid body and created a synthetic trajectory that is representative of a store release event. This gave us a synthetic “truth” trajectory that allowed us to create Monte Carlo sets of synthetic noisy observations as seen from a set of virtual cameras. The Monte Carlo synthetic measurement sets were then used in a variety of estimation (solver) methods and the root mean squared (RMS) track errors calculated from each estimation method allowed for comparative performance analysis.

The synthetic trajectories and Monte Carlo noisy camera observations sets from Chapter 19 will be reused in this chapter, and we will add noisy measurements as felt by a virtual IMU's accelerometers and rate gyroscopes attached to the synthetic rigid body. Thus, we will be able to perform comparative analysis of various estimation methods and also analyze any improvements in performance due to the addition of IMU measurements.

20.4.1 Synthetic Trajectory

We shall use the same synthetic trajectory developed in Section 19.5.2, except that we generate a sequence of times t that has a faster data rate

(20.21)

Although the IMU measurements are generated at the shorter period of 0.001 seconds, the image measurements are still generated at the slower period of 0.005 ...