Event Title

Orbit Estimation from Angles-Only Observations Using Nonlinear Filtering Schemes

Location

Room # EB302

Start Date

5-6-2019 9:40 AM

Description

This research assesses the performance of various filtering schemes for tracking uncooperative satellites through space-based measurements, and proposes new methodologies based on Gaussian Mixture Models and the Skewed Kalman Filter. Traditional filtering schemes, such as the Extended Kalman Filter (EKF) and Unscented Kalman Filter (UKF), both diverge while tracking under reasonable orbital conditions when there is a long duration between measurements. We analyze a simplified model, based on full position measurement of a MEO orbit, to isolate filter divergence due to nonlinear dynamics, and corrected the divergence by implementing a hybrid Monte-Carlo particle filter with Kalman-type updates. We next consider space-based optical measurements, which induce a separate source of divergence due to nonlinear measurement error, and observe that our hybrid particle filter cannot remedy this problem. We implement a Gaussian Mixture Model (GMM) to overcome this challenge, which is effective but comes at a high computational cost. Separately, we have analyzed a Skewed Kalman Filter (SKF) for incorporating skewness into Kalman-based methods, and outline a strategy for combining the GMM and SKF methodologies to provide a low-cost option for tracking uncooperative satellites using sparse space-based measurements.

Comments

Session 2

This document is currently not available here.

Share

COinS
 
May 6th, 9:40 AM

Orbit Estimation from Angles-Only Observations Using Nonlinear Filtering Schemes

Room # EB302

This research assesses the performance of various filtering schemes for tracking uncooperative satellites through space-based measurements, and proposes new methodologies based on Gaussian Mixture Models and the Skewed Kalman Filter. Traditional filtering schemes, such as the Extended Kalman Filter (EKF) and Unscented Kalman Filter (UKF), both diverge while tracking under reasonable orbital conditions when there is a long duration between measurements. We analyze a simplified model, based on full position measurement of a MEO orbit, to isolate filter divergence due to nonlinear dynamics, and corrected the divergence by implementing a hybrid Monte-Carlo particle filter with Kalman-type updates. We next consider space-based optical measurements, which induce a separate source of divergence due to nonlinear measurement error, and observe that our hybrid particle filter cannot remedy this problem. We implement a Gaussian Mixture Model (GMM) to overcome this challenge, which is effective but comes at a high computational cost. Separately, we have analyzed a Skewed Kalman Filter (SKF) for incorporating skewness into Kalman-based methods, and outline a strategy for combining the GMM and SKF methodologies to provide a low-cost option for tracking uncooperative satellites using sparse space-based measurements.