Analysis of deeply integrated and tightly coupled architectures

Abstract

This paper analyzes the impact that architectural features have on the performance of deeply integrated and tightly coupled algorithms. The effects of two specific architectural features are investigated. The first is the design of the Kalman filter used in the algorithms. The performance degradation caused by using a federated filtering architecture instead of a single, centralized filter is analyzed. The second feature is the usage of scalar and vector tracking loops. The advantage offered by vector tracking loops over scalar tracking loops is quantified. The effects of these two architectural features are determined by analyzing the comparative performance of three different algorithms. One algorithm uses a single Kalman filter to process the GPS signals and the inertial sensor data. The other two algorithms use a federated filtering architecture. One federated algorithm uses scalar tracking loops and the other uses vector tracking loops. Comparing the performance of the three algorithms allows the effects of filter design and tracking loop operation to be isolated. Covariance analysis and Monte Carlo simulations are used to study the performance of the algorithms with different inertial sensor grades and satellite constellations. The analysis reveals that the federated algorithm with vector tracking and the centralized filtering algorithm perform virtually identically. The federated algorithm with scalar tracking loops performs poorer. However, the performance of all three algorithms converge as the carrier to noise power density ratio declines. At low signal powers, all three algorithms provide identical performance. The results quantify how the architectural features of coupled systems affect their performance.