Radar-Based Iterative JPDA Tracking and Improved Two-Arc Association for Rapid Cataloging of Non-Cooperative Targets in a Debris Cloud

Rapid cataloging of non-cooperative targets in newly formed space debris clouds requires robust radar-based multi-target tracking to form valid observation arcs and reliable association of arcs acquired during temporally separated observation windows. Dense target distributions and ambiguous measurement-to-track and arc-to-arc relationships often result in missed measurements and false associations. To address the challenge, this study proposes an iterative joint probabilistic data association (JPDA) procedure to perform radar tracking of multiple targets in a single observation window, in which a data mining operation is implemented after each JPDA iteration to maximize observation utilization rate. In the two-arc association stage, candidate arc pairs are first constructed using two-arc association algorithm based on the Lambert boundary-value problem, and are then further screened according to the residuals of two-arc orbit determination. This strategy suppresses false associations while maintaining high true positive associations. The algorithms were experimented with simulated radar observations of 100 targets in the Iridium 33-Cosmos 2251 debris cloud in the time period 2 h–4 h after the collision. Results show that the iterative JPDA tracking increased the number of successfully tracked targets in the first observation window to 96 from 88 using the single-run JPDA tracking, corresponding to an improvement of approximately 9%. The first-round data-mining operation increased the minimum observation utilization rate from 54.58% to 71.67%, demonstrating its effectiveness in recovering unused observations in the JPDA tracking. For the two-arc association, the Lambert-problem-based method suffered a false association rate of 95.17%, although it achieved an association recall rate of 96.88%. In contrast, the improved two-arc association method reduced the false association rate to 5.26% and maintained a high association recall rate of 93.75%. These results demonstrate that the proposed framework supports rapid and reliable tracking, association, orbit determination, and cataloging of non-cooperative space targets.