Multiple Hypothesis Tracking for Opportunistic Transmitter Identification
Sanderson, Jennifer Jean · 2025
Read Multiple Hypothesis Tracking for Opportunistic Transmitter Identification by Sanderson, Jennifer Jean (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/doe-osti-3005457-multiple-hypothesis-tracking-for-opportu.txt.
Introduction
Jennifer Jean Sanderson writes a course project paper, SAND2025-14606R, for a graduate electrical engineering class, ECE 8851. The topic is navigation by other people's signals. Low Earth orbit communication satellites flood the sky with radio transmissions never meant for navigation. Receivers can still extract timing from them, if they can tell which satellite sent which signal.
Sanderson applies multiple hypothesis tracking to that identification problem: keeping several candidate assignments alive until the evidence separates them. Public two-line element sets plus an orbit propagator supply rough satellite positions. The work builds on her earlier nearest-neighbor data association study with Kassas, presented in 2025. Megaconstellation navigation is attractive because it works where GPS is jammed or spoofed, a growing concern the paper speaks to directly.
Catalog record
| CREATOR | Sanderson, Jennifer Jean |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 9 KB |
| LICENSE | Public Domain |
| RIGHTS | PUBLIC DOMAIN |
| SOURCE | DOE OSTI |
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
SAND2025-14606R ECE 8851 Final Project Abstract Jennifer Sanderson Multiple Hypothesis Tracking for Opportunistic Transmitter Identification Motivation The exploitation of radio frequency (RF) signals from Low Earth Orbit (LEO) communication satellites for the purposes of navigation is an active field of research. In this alternative navigation system, navigation observables are passively extracted from existing RF signals. Because these signals are not designed for navigation purposes and often have proprietary signal structures that are unknown to the receiver, they lack much of the information required for navigation systems or such information is unavailable to the receiver. This information includes (1) the signal transmission time required for calculating the range measurement to the satellite, (2) ephemeris and clock corrections and (3) unique identifiers to distinguish between satellites. Recent work has been done to address these problems. Specialized signal processing techniques to passively extract navigation observables with limited knowledge of the signal characteristics have been developed [1]. Satellite ephemerides can be estimated using publicly available two line element sets (TLEs) coupled with an orbital propagator. Advanced estimation techniques such as SLAM- type tracking and differential positioning have been developed to mitigate the inherent errors of TLEs [2] [3] [4] [5]. The study of stability of the lower quality clocks common to communication systems has also been performed to develop dynamic models for estimation of relevant clock terms [6]. One metaphorical stone still left unturned however is that of associating received measurements with the predicted satellite ephemerides. Prior works utilized post-processed data and assumed the origin of a particular measurement was known. This is a reasonable concession to focus on other critical aspects of the technology. However, as opportunistic RF navigation technology matures and nears real-time implementation, a method for data association is needed. This effort seeks to address the need for passive data association – a design which does not require any decoding or demodulation of the received signal. Previous work has made use of a simple heuristic approach and demonstrated its efficacy on the OneWeb constellation. Associations were made based on a comparison of the normalized innovation squared (NIS) value between all potential satellites. The satellite which yielded the smallest NIS was selected [7]. It was shown in this work that geometric ambiguities between satellites can cause incorrect associations. The OneWeb constellation is forgiving in this regard – orbits are spaced far apart relative to anticipated receiver errors so incorrect associations only occurred when the receiver errors were large. Applying this approach to a larger constellation such as Starlink has been shown to be unusable even under near perfect receiver and satellite states. This method failed because greedy algorithms such as the minimum NIS approach, are rarely able to correct themselves if the initial association was wrong. In the Starlink constellation, there are often multiple satellites that yield similar NIS values. Selecting the smallest of these values at a single measurement epoch is not a reliable method. A Solution To combat this limitation, a deferred logic approach is proposed. A multiple hypothesis tracking (MHT) framework specialized for opportunistically received RF measurements will be developed. A high-level overview of this framework is described in the following section. The backbone of this algorithm is the use of a likelihood function to compare the probability of each potential association. As such, a specialized likelihood function for opportunistic transmitter identification will also be proposed. Lastly, the MHT algorithm will be evaluated using simulated trajectories with ambiguous transmitter geometries. Transmitter, receiver and measurement errors are modeled in these simulations and varied to determine their effects on the algorithm’s ability to correctly associate measurements. Geometries were designed to have at least one measurement epoch in which predicted measurements were close to each other. Examples of these transmitter geometries are shown in Fig. 1. The ability of the MHT algorithm to correctly associate transmitters with received measurements will be compared with the NIS method. Additionally, the time of convergence, or the amount of time required to correctly associate a transmitter with a received measurement, will be used to evaluate the feasibility of this proposed framework on a real-system. This work makes the following contributions: (1) the likelihood function for an opportunistic RF data association scenario will be derived, (2) a multiple hypothesis tracking framework will be developed for opportunistic RF data association and (3) the proposed framework will be evaluated on ambiguous transmitter geometries and will be compared to a heuristic approach. Figure 1- Example transmitters with ambiguous geometries relative to the receiver. Multiple Hypothesis Tracking Framework The primary idea surrounding the MHT is to delay making an association decision until the likelihood of a particular association exceeds a specified threshold. The basic framework of the MHT developed for this work is shown in Fig. 2. It consists of two categories of estimation filters, the primary and association filters. The primary filter is designed to localize the receiver. It is assumed for this work the primary filter does not track the transmitter’s states. An extended Kalman filter (EKF) will be used for the primary filter. The association filter is denoted by the dashed blue line box in Fig. 2. Its purpose is to estimate the receiver and all candidate transmitter states during the association process. When an association has been made, it will hand off its most recent estimates of the receiver and selected satellite’s states to the primary estimation filter. The focus of this work is in developing the association filter. Investigation of different primary filter designs is left for future work. The association filter is first initialized with the candidate transmitters’ and receiver’s states. If the primary filter is already in operation, the most recent state estimate will be used to initialize the receiver states in the association filter. Transmitters that fall outside of the receiver’s visibility region are removed from consideration. States are propagated using an appropriate dynamics model until a measurement is received. The likelihood function, also known as the track score, is then calculated for each candidate transmitter and received measurement combination. Associations with low tracks scores are removed from consideration to mitigate computational complexity. Track scores are combined with all previously calculated scores for each association and then converted to the association’s probability. A tunable threshold is used to determine the acceptance of associations. If none of the association probabilities exceed this threshold, the MHT will save all remaining track scores and continue propagating the receiver’s and transmitters’ states. If any of the association probabilities exceed this threshold, the constituent transmitter’s states are sent to the primary navigation filter for a measurement update and the MHT algorithm will terminate until a measurement from a new source is received. If unassociated measurements still remain after one association has been made, the MHT algorithm will continue, removing the associated transmitter(s) from consideration, until all remaining measurements are associated. Figure 2- High-level block diagram of the MHT framework References [1] S. Kozhaya and Z. M. Kassas, "On the fundamental tracking performance and design considerations of radio navigation," IEEE Journal on Selected Areas in Communications, pp. 2395-2409, 2024. [2] N. Khairallah and Z. M. Kassas, "Ephemeris tracking and error propagation analysis of LEO satellites with application to opportunistic navigation," IEEE Transactions on Aerospace and Electronic Systems, pp. 1242-1259,