Abstract: | Abstract—Recognizing the pose of a satellite is an essential priority in docking missions and future space debris cleanup. In this work, we present a distributed framework for determining the pose estimate of a rigid object with a known 3D model through calibrated 2D images and consensus among multiple sensing agents. The introduced approach utilizes global image statistics without the need for local image features, while also leveraging a weighted consensus scheme to drive estimation in a robust fashion. By constructing an objective function that fuses segmentation energy and consensus cost, the approach can achieve marked improvement upon single-agent systems, which are shown to be prone to local minima. The method introduced is demonstrably resilient for space-based pose estimation where information transfer is hamstrung by signal interference and low-power computing. To rigorously evaluate our proposed method, we simulated a satellite inspection scenario where images cannot be transmitted between agents in real time and operate under the assumption of having limited bandwidth. Furthermore, image quality is subject to multiple sources of degradation to reproduce obstacles native to the space domain that alternative methods struggle with. The resulting pose estimation and consensus are not only competitive with alternative approaches but also offer a novel and pragmatic framework for future multi-agent space-based imaging problems. Index Terms—pose estimation, satellite, space, consensus, robust, region-based, networks |
Published in: |
2023 IEEE/ION Position, Location and Navigation Symposium (PLANS) April 24 - 27, 2023 Hyatt Regency Hotel Monterey, CA |
Pages: | 1322 - 1329 |
Cite this article: | Grange, Daniel, Sandhu, Romeil, Soderlund, Alexander A., Phillips, Sean, "Consensus on Region-Based Pose Estimation for Satellites," 2023 IEEE/ION Position, Location and Navigation Symposium (PLANS), Monterey, CA, April 2023, pp. 1322-1329. https://doi.org/10.1109/PLANS53410.2023.10140083 |
Full Paper: |
ION Members/Non-Members: 1 Download Credit
Sign In |