Encounter-Based Simulation Architecture for Detect-And-Avoid Modeling

Abstract

This paper presents an encounter-based simulation architecture developed to facilitate flexible and efficient detect and avoid modeling in parametric or trade-space studies on large data sets. The basic premise of this tool is that large-scale input data can be reduced to a set of "canonical encounters" and that using the reduced data in simulations does not lead to loss of fidelity. A canonical encounter is specified as ownship and intruder flight portions potentially resulting in a loss of well clear along with a set of properties that characterize the encounter. The advantages of using canonical encounters include faster simulations, reduced memory footprint, ability to select encounters based on user-specified criteria, shared encounters across multiple teams, peer-reviewed encounters, and a better understanding of the input data set, to name a few. The performance of the encounter-based approach is compared to the approach used previously, which modeled flights from departure to destination using the Java Architecture for DAA Extensibility and Modeling (JADEM). The new architecture reduced the amount of data to be processed a hundredfold and the total computation time five-fold.