Smart Projectile State Estimation Using Evidence Theory

Abstract

Smart projectile state estimation is a challenging task due to highly nonlinear vehicle dynamic behavior and unreliable or noisy sensor feedback. Although Kalman-filter-based algorithms are currently the primary means of sensor fusion and state estimation for smart weapons applications, they are limited in estimation accuracy, their ability to combine data from a wide variety of sensors, and their ability to recognize and reject erroneous feedback. This paper explores the use of evidence theory (or Dempster-Shafer theory) for projectile state estimation purposes. Evidence theory offers a generalized framework capable of incorporating feedback from a wide variety of sensors with little a priori knowledge about sensor noise characteristics. The framework's main strengths are its ability to manage varying levels of dynamic uncertainty between sensors and its capability to recognize and report conflicting sensor feedback. The paper begins with an overview of evidence theory and a discussion of its strengths when applied to projectile state estimation. Then an evidence-theory-based filter is described, and an example roll-angle estimator is presented. Results are discussed and performance is compared with an extended Kalman filter. The ability to identify and eliminate malfunctioning sensors from the estimation process proves to be a key advantage of the proposed design over current methods.

Automatic Updates of Transition Potential Matrices in Dempster-Shafer Networks Based on Evidence Inputs

“…The following papers provide a mathematical background to the theory [ 4 , 9 ]. “Smart Projectile State Estimation Using Evidence Theory” provides a practical understanding of evidence theory using sensor fusion and state estimation as the backdrop [ 10 ]. Other practical explanations of DS theory are available [ 11 ].…”