Development and Evaluation of Fusion Techniques

Abstract

The purpose of this effort was to investigate and develop promising innovative technologies that hold promise for improvements in the Air Force target tracking and fusion capabilities. The problem statement is that Nonlinear Non-Gaussian Processes (NNGP) present a major challenge in all types of military problems. This is because the real-world is nonlinear and non-Gaussian, despite assumptions made in most conventional fusion algorithms. This in-house program addressed this problem by researching and developing tracking filters that do not presume a linear Gaussian world. The most famous of these is the Particle Filter (PF). Progress made includes development of a MATLAB PF simulation capability for in-house analysis/testing and preliminary investigation of PF capabilities. Prior to this effort AFRL/IF had no capability in this area and PF were not being investigated. DEFT developed a two-dimensional PF and performed Monte Carlo simulation runs to test the performance of the PF as compared to the Extended Kalman Filter (EKF) for multiple bearings only sensors (ESM sensors). Results showed that in certain cases, the Monte Carlo averaged tracking error variance of the PF was much better than that of the EKF, but with ten times the computational cost. Under these same conditions (50 particles, 100 Monte Carlo runs, two bearings only sensors), the EKF lost track 28% of the time whereas the PF did not lose track at all. This effort established future directions for research. There is a need to do more nonlinear filtering analysis, development and enhancements under realistic scenarios with varying degrees of nonlinearity. Considerable study is needed to fully determine the conditions under which these techniques will lead to improved performance. It is possible that a system could be developed employing multiple nonlinear filters.

Open PDF

Document Details

Document Type
Technical Report
Publication Date
Oct 01, 2007
Accession Number
ADA474630

Entities

People

  • Adnan Bubalo
  • Eric C. Jones
  • Maria L. Scalzo
  • Mark Alford

Organizations

  • Air Force Research Laboratory

Tags

Communities of Interest

  • C4I
  • Cyber
  • Materials and Manufacturing Processes
  • Sensors

DTIC Thesaurus Topics

  • Aircrafts
  • Algorithms
  • Computational Complexity
  • Computational Science
  • Data Science
  • Filtration
  • Fokker Planck Equations
  • Information Science
  • Kalman Filters
  • Mathematical Filters
  • Monte Carlo Method
  • Sequential Monte Carlo Methods
  • Statistical Algorithms
  • Target Tracking
  • Test And Evaluation
  • Two Dimensional
  • Unmanned Aerial Vehicles

Fields of Study

  • Engineering

Readers

  • Adaptive Control and Estimation with Uncertainty in Dynamic Systems.
  • Geochemistry
  • Systems Analysis and Design