Using a High-Fidelity Simulation Framework for Performance Singularity

A common way to evaluate the performance of a system is to compare the algorithmic outputs with ground truth to identify divergences in the system´s performance and discover the errors it is prone to. In the absence of such ground truth or as a follow-on to performance evaluation, performance analysis at the algorithmic level can provide developers insight into performance singularities. Such performance singularity identification and testing provides real-time meta-data that allows developers to understand the impact of singularities on the overall performance of the system. As an example of the concepts developed in this paper, we present a navigation solution based on image registration algorithms and the methodology used for the identification and testing of performance singularities of this algorithm.