---
title: "Data-Centric Preprocessing for GNSS-Based Orbit Determination: Presented at SPAICE 2026"
description: At the 3rd Conference on AI in and for Space, Marta Guimarães presented new research on data-centric preprocessing for GNSS-based orbit determination. The work does not replace established OD algorithms. It adds a dedicated preprocessing stage that improves the robustness of the data they rely on.
---

[Neuraspace Stories](https://blog.neuraspace.com)

# [Data-Centric Preprocessing for GNSS-Based Orbit Determination: Presented at SPAICE 2026](https://blog.neuraspace.com/data-centric-preprocessing-for-gnss-based-orbit-determination-presented-at-spaice-2026)

 Written by [Neuraspace](https://blog.neuraspace.com/author/neuraspace) | Oct 26, 2026, 9:30:15 AM

 

At the 3rd Conference on AI in and for Space (SPAICE 2026), Neuraspace's AI Researcher Marta Guimarães presented new work on data-centric preprocessing for GNSS-based orbit determination. The work does not replace established OD algorithms. It adds a dedicated preprocessing stage that improves the robustness of the data they rely on.

**Paper:** Data-Centric Preprocessing for GNSS-Based Orbit Determination via AI, Signal Processing, and Orbital Dynamics  
**Authors:** Marta Guimarães, Robert Arthur, Cláudia Soares, Chiara Manfletti, Gerarda De Pasquale

**Challenge** GNSS-based orbit determination relies on input measurements whose quality can vary significantly across mission phases and operating conditions. In sensitive phases such as launch and early orbit, degraded or inconsistent data can significantly reduce estimation quality or lead to non-convergence. If misapplied, data conditioning can degrade convergence and numerical stability.

**Approach** The pipeline operates in three stages:

1. **Base filtering:** lightweight sanity checks remove measurements incompatible with physically meaningful orbital motion.
2. **Structure assessment via ML:** a convolutional neural network classifies segments as non-structured or structured.
3. **Assumption-controlled anomaly removal:** increasingly aggressive filters are applied only when supported by the data.

**Key Findings**

- The classifier reached an average F1-score of 0.947 over 10 independent trials, despite class imbalance
- The pipeline agrees with a baseline batch least-squares estimator on obvious outliers
- It additionally identifies dynamically inconsistent samples that remain invisible to position-only residual analysis
- The weighted root-mean-square of post-fit residuals was reduced by about 63%
- The number of solver iterations decreased from five to four
- Virtually all valid measurements were retained

**Practical Impact** Robust OD is not solely a modelling problem, but also a data-centric one. Structure-aware preprocessing offers a practical and operationally compatible pathway to improving stability and consistency in real-world GNSS-based OD workflows, without modifying the underlying dynamical model.

[**Read the paper.**](https://blog.neuraspace.com/hubfs/SPAICE___GNSS_Data_v3.pdf)

 

[View full post](https://blog.neuraspace.com/data-centric-preprocessing-for-gnss-based-orbit-determination-presented-at-spaice-2026)

```json
{
  "@context" : "http://schema.org",
  "@type" : "BlogPosting",
  "author" : {
    "@type" : "Person",
    "name" : "Neuraspace"
  },
  "dateModified" : "2026-10-26T09:43:05.547Z",
  "datePublished" : "2026-10-26T09:30:15Z",
  "headline" : "Data-Centric Preprocessing for GNSS-Based Orbit Determination: Presented at SPAICE 2026",
  "image" : {
    "@type" : "ImageObject",
    "height" : 639,
    "url" : "https://25553711.fs1.hubspotusercontent-eu1.net/hubfs/25553711/neuraspace_marta_guimaraes_spaice_2026.jpg",
    "width" : 960
  },
  "mainEntityOfPage" : "https://blog.neuraspace.com/data-centric-preprocessing-for-gnss-based-orbit-determination-presented-at-spaice-2026",
  "publisher" : {
    "@type" : "Organization",
    "logo" : {
      "@type" : "ImageObject",
      "height" : 60,
      "url" : "/hs/hsstatic/content_shared_assets/static-1.4092/img/default-amp-logo.png",
      "width" : 60
    },
    "name" : "Neuraspace Stories"
  }
}
```