juliensimon/constellation-tle-latest

---
license: cc-by-4.0
pretty_name: "Constellation TLEs — 18 Satellite Constellations"
language:
- en
description: >-
Daily TLE snapshots for 18 satellite constellations (1,783 satellites) from
CelesTrak. Covers GNSS, LEO broadband, GEO comms, and Earth observation fleets.
task_categories:
- tabular-regression
tags:
- space
- open-data
- tabular-data
- parquet
- satellite
- tle
- orbital-mechanics
- celestrak
- sgp4
- gnss
- gps
- galileo
- beidou
- glonass
- oneweb
- iridium
- constellation
size_categories:
- 1K<n<10K
configs:
- config_name: oneweb
data_files:
- split: train
path: data/oneweb.parquet
default: true
- config_name: kuiper
data_files:
- split: train
path: data/kuiper.parquet
- config_name: qianfan
data_files:
- split: train
path: data/qianfan.parquet
- config_name: hulianwang
data_files:
- split: train
path: data/hulianwang.parquet
- config_name: iridium
data_files:
- split: train
path: data/iridium.parquet
- config_name: globalstar
data_files:
- split: train
path: data/globalstar.parquet
- config_name: orbcomm
data_files:
- split: train
path: data/orbcomm.parquet
- config_name: planet
data_files:
- split: train
path: data/planet.parquet
- config_name: spire
data_files:
- split: train
path: data/spire.parquet
- config_name: gps
data_files:
- split: train
path: data/gps.parquet
- config_name: galileo
data_files:
- split: train
path: data/galileo.parquet
- config_name: beidou
data_files:
- split: train
path: data/beidou.parquet
- config_name: sbas
data_files:
- split: train
path: data/sbas.parquet
- config_name: ses
data_files:
- split: train
path: data/ses.parquet
- config_name: intelsat
data_files:
- split: train
path: data/intelsat.parquet
- config_name: eutelsat
data_files:
- split: train
path: data/eutelsat.parquet
- config_name: telesat
data_files:
- split: train
path: data/telesat.parquet
---

# Constellation TLEs — 18 Satellite Constellations

<div align="center">
<img src="banner.jpg" alt="An orbital sunrise illuminates the Earth's atmosphere, seen from the ISS" width="400">
<p><em>Credit: NASA</em></p>
</div>

*Part of the [Orbital Mechanics Datasets](https://huggingface.co/collections/juliensimon/orbital-mechanics-datasets-69c24caca4ab3934c9856994) collection on Hugging Face.*


Daily Two-Line Element (TLE) snapshots for **17** satellite constellations
totaling **1,783** satellites, sourced from [CelesTrak](https://celestrak.org/).
Each constellation is available as a separate config for independent loading.

## Constellations

| Constellation | Operator | Orbit | Satellites |
|---------------|----------|-------|----------:|
| OneWeb | Eutelsat OneWeb | LEO | 651 |
| Kuiper | Amazon | LEO | 239 |
| Qianfan (G60) | Shanghai Spacecom | LEO | 108 |
| Hulianwang (GuoWang) | China SatNet | LEO | 168 |
| Iridium NEXT | Iridium | LEO | 80 |
| Globalstar | Globalstar | LEO | 28 |
| ORBCOMM | ORBCOMM | LEO | 15 |
| Planet Labs | Planet Labs | LEO | 136 |
| Spire Global | Spire Global | LEO | 76 |
| GPS (NAVSTAR) | USSF | MEO | 39 |
| Galileo | EU/ESA | MEO | 50 |
| BeiDou | CNSA | MEO/GEO | 56 |
| SBAS | Various | GEO | 6 |
| SES | SES | GEO | 70 |
| Intelsat | Intelsat | GEO | 27 |
| Eutelsat | Eutelsat | GEO | 29 |
| Telesat | Telesat | GEO/LEO | 5 |

**Total: 1,783 satellites** — 1,501 LEO, 145 MEO, 137 GEO

## Dataset description

Two-Line Element sets (TLEs) are the standard format for representing satellite orbital
elements, developed by NORAD and used universally with the SGP4/SDP4 propagation model.
This dataset provides daily-fresh TLEs for every major non-Starlink constellation in orbit,
organized by operator for easy access.

The collection spans four categories:

**GNSS navigation** (GPS, Galileo, BeiDou, GLONASS, SBAS) — the backbone of global
positioning in MEO/GEO, critical for timing applications, geodesy, and as reference
orbits for propagation model validation.

**LEO broadband** (OneWeb, Kuiper, Qianfan, Hulianwang) — the next wave of
mega-constellations competing with Starlink. OneWeb is fully deployed at 1,200 km;
Amazon's Kuiper and China's Qianfan/Hulianwang are in early deployment phases.

**LEO communications and IoT** (Iridium NEXT, Globalstar, ORBCOMM) — established
mobile satellite service providers operating in low Earth orbit.

**Earth observation** (Planet Labs, Spire Global) — the largest commercial imaging
and weather monitoring fleets, operating in sun-synchronous orbits.

**GEO communications** (SES, Intelsat, Eutelsat, Telesat) — legacy geostationary
operators providing broadcast, broadband, and government services from 35,786 km.

Raw `.tle` files are provided alongside Parquet for maximum compatibility: orbit
propagation libraries like `python-sgp4`, `orekit`, and STK consume the standard
three-line TLE format directly. Because TLE accuracy degrades rapidly (especially
for LEO objects), daily updates are essential for operational applications.

Starlink TLEs are published separately in
[starlink-tle-latest](https://huggingface.co/datasets/juliensimon/starlink-tle-latest)
due to the constellation's size (7,000+ satellites).

## Raw TLE files

For applications that consume standard 3-line TLE format (e.g., SGP4 propagators):

- [`data/oneweb.tle`](https://huggingface.co/datasets/juliensimon/constellation-tle-latest/resolve/main/data/oneweb.tle) — OneWeb
- [`data/kuiper.tle`](https://huggingface.co/datasets/juliensimon/constellation-tle-latest/resolve/main/data/kuiper.tle) — Kuiper
- [`data/qianfan.tle`](https://huggingface.co/datasets/juliensimon/constellation-tle-latest/resolve/main/data/qianfan.tle) — Qianfan (G60)
- [`data/hulianwang.tle`](https://huggingface.co/datasets/juliensimon/constellation-tle-latest/resolve/main/data/hulianwang.tle) — Hulianwang (GuoWang)
- [`data/iridium.tle`](https://huggingface.co/datasets/juliensimon/constellation-tle-latest/resolve/main/data/iridium.tle) — Iridium NEXT
- [`data/globalstar.tle`](https://huggingface.co/datasets/juliensimon/constellation-tle-latest/resolve/main/data/globalstar.tle) — Globalstar
- [`data/orbcomm.tle`](https://huggingface.co/datasets/juliensimon/constellation-tle-latest/resolve/main/data/orbcomm.tle) — ORBCOMM
- [`data/planet.tle`](https://huggingface.co/datasets/juliensimon/constellation-tle-latest/resolve/main/data/planet.tle) — Planet Labs
- [`data/spire.tle`](https://huggingface.co/datasets/juliensimon/constellation-tle-latest/resolve/main/data/spire.tle) — Spire Global
- [`data/gps.tle`](https://huggingface.co/datasets/juliensimon/constellation-tle-latest/resolve/main/data/gps.tle) — GPS (NAVSTAR)
- [`data/galileo.tle`](https://huggingface.co/datasets/juliensimon/constellation-tle-latest/resolve/main/data/galileo.tle) — Galileo
- [`data/beidou.tle`](https://huggingface.co/datasets/juliensimon/constellation-tle-latest/resolve/main/data/beidou.tle) — BeiDou
- [`data/sbas.tle`](https://huggingface.co/datasets/juliensimon/constellation-tle-latest/resolve/main/data/sbas.tle) — SBAS
- [`data/ses.tle`](https://huggingface.co/datasets/juliensimon/constellation-tle-latest/resolve/main/data/ses.tle) — SES
- [`data/intelsat.tle`](https://huggingface.co/datasets/juliensimon/constellation-tle-latest/resolve/main/data/intelsat.tle) — Intelsat
- [`data/eutelsat.tle`](https://huggingface.co/datasets/juliensimon/constellation-tle-latest/resolve/main/data/eutelsat.tle) — Eutelsat
- [`data/telesat.tle`](https://huggingface.co/datasets/juliensimon/constellation-tle-latest/resolve/main/data/telesat.tle) — Telesat

## Schema (all configs)

| Column | Type | Description |
|--------|------|-------------|
| `name` | string | Satellite name (e.g., "ONEWEB-0012", "NAVSTAR 78", "GALILEO 27") |
| `line1` | string | TLE line 1: epoch, drag term, BSTAR, element set number |
| `line2` | string | TLE line 2: inclination, RAAN, eccentricity, arg of perigee, mean anomaly, mean motion |

## Quick stats

- **1,783** satellites across **17** constellations
- **1,501** LEO + **145** MEO + **137** GEO
- Snapshot: 2026-04-11 06:37 UTC

## Usage

```python
from datasets import load_dataset

# Load a specific constellation
gps = load_dataset("juliensimon/constellation-tle-latest", "gps", split="train")
oneweb = load_dataset("juliensimon/constellation-tle-latest", "oneweb", split="train")
galileo = load_dataset("juliensimon/constellation-tle-latest", "galileo", split="train")

# Use with sgp4 library for orbit propagation
from sgp4.api import Satrec
sat = Satrec.twoline2rv(gps[0]["line1"], gps[0]["line2"])

# Load all GNSS constellations
import pandas as pd
gnss = pd.concat([
load_dataset("juliensimon/constellation-tle-latest", c, split="train").to_pandas()
for c in ["gps", "galileo", "beidou", "glonass", "sbas"]
])
print(f"{len(gnss)} GNSS satellites")

# Compare constellation sizes
for config in ["oneweb", "iridium", "planet", "spire"]:
ds = load_dataset("juliensimon/constellation-tle-latest", config, split="train")
print(f"{config}: {len(ds)} satellites")
```

## Data source

[CelesTrak](https://celestrak.org/) (Dr. T.S. Kelso), mirroring NORAD/18th Space
Defense Squadron data. No authentication required.

## Update schedule

Daily at 05:30 UTC via [GitHub Actions](https://github.com/juliensimon/space-datasets).

## Related datasets

- [starlink-tle-latest](https://huggingface.co/datasets/juliensimon/starlink-tle-latest) — Daily Starlink + GPS TLEs (dedicated dataset for the largest constellation)
- [space-track-tle-history](https://huggingface.co/datasets/juliensimon/space-track-tle-history) — 238M+ historical TLEs for every cataloged object (1959–present)
- [constellation-census](https://huggingface.co/datasets/juliensimon/constellation-census) — Parsed orbital elements with status classification for all constellations
- [space-track-satcat](https://huggingface.co/datasets/juliensimon/space-track-satcat) — Complete NORAD satellite catalog (68K+ objects)
- [ucs-satellite-database](https://huggingface.co/datasets/juliensimon/ucs-satellite-database) — Active satellites with purpose, operator, and orbit metadata

## Pipeline

Source code: [juliensimon/space-datasets](https://github.com/juliensimon/space-datasets)

## Support

If you find this dataset useful, please give it a ❤️ on the [dataset page](https://huggingface.co/datasets/juliensimon/constellation-tle-latest) and share feedback in the Community tab! Also consider giving a ⭐️ to the [space-datasets](https://github.com/juliensimon/space-datasets) repo.

## Citation

```bibtex
@dataset{constellation_tle_latest,
author = {Simon, Julien},
title = {Constellation TLEs — 18 Satellite Constellations},
year = {2026},
publisher = {Hugging Face},
url = {https://huggingface.co/datasets/juliensimon/constellation-tle-latest},
note = {Based on NORAD data via CelesTrak (Dr. T.S. Kelso)}
}
```

## License

[CC-BY-4.0](https://creativecommons.org/licenses/by/4.0/)

搜集汇总

数据集介绍

构建方式

在航天动力学领域，精确的轨道数据对于卫星任务规划与空间态势感知至关重要。本数据集通过自动化流程，每日从CelesTrak平台获取18个主要卫星星座的TLE数据快照，涵盖全球导航卫星系统、低轨宽带通信、地球观测及地球静止轨道通信等多个类别。数据以原始TLE格式与Parquet结构化格式并行提供，确保了与SGP4等轨道传播库的即插即用兼容性，并通过持续更新机制维持了数据的时间敏感性。

特点

该数据集以其全面性与时效性脱颖而出，整合了1805颗在轨卫星的轨道参数，并细致划分为17个独立星座配置。其结构设计兼顾了研究便捷性，每个星座均以独立文件存储，便于针对性分析。数据内容不仅包含标准的双行轨道根数，还附有卫星名称、NORAD编号及历元时间等元数据，为轨道力学研究、星座性能评估以及空间交通管理提供了高颗粒度的信息基础。

使用方法

用户可通过Hugging Face的datasets库便捷加载特定星座数据，并直接与sgp4等轨道传播库集成，用于卫星位置与速度的精确计算。数据集支持以Pandas DataFrame形式进行灵活的数据操作与多星座联合分析，例如对比不同星座的规模或筛选特定轨道类型的卫星。对于需要原始格式的应用，可直接下载提供的.tle文件，无缝接入现有轨道分析工具链。

背景与挑战

背景概述

随着全球卫星星座规模的急剧扩张，对高时效性轨道数据的需求日益凸显。Constellation TLE-latest数据集由Julien Simon等人于2026年构建，依托CelesTrak（由T.S. Kelso博士维护）的权威数据源，系统整合了18个主要非星链星座的每日两行轨道根数（TLE）快照。该数据集覆盖了全球导航卫星系统、低轨宽带通信、地球观测及地球静止轨道通信等多个关键领域，共计1805颗卫星，为轨道力学、空间态势感知及卫星任务规划提供了标准化的数据基础。其核心研究问题在于解决多星座轨道数据的实时聚合与标准化访问，显著提升了轨道预测与碰撞规避研究的效率，对航天工程与空间科学研究产生了深远影响。

当前挑战

在轨道动力学领域，精确预测卫星位置面临严峻挑战，尤其低轨卫星受大气阻力等扰动影响，其TLE数据精度在数日内迅速衰减，这要求数据集必须实现每日高频更新以维持实用性。构建过程中，数据集成面临异构星座的轨道根数格式统一、大规模卫星数据的实时抓取与清洗，以及排除星链等超大规模星座以保持数据集可管理性等具体难题。此外，确保数据源（如NORAD/第18太空防御中队）的持续可靠访问，并在不同轨道传播库（如SGP4）间维持兼容性，亦是该数据集构建与维护的核心挑战。

常用场景

经典使用场景

在轨道力学与航天动力学领域，精确的卫星轨道预测是核心研究课题。Constellation TLE-latest数据集通过提供涵盖全球导航卫星系统、低轨宽带、地球观测等18个主要星座的每日两行轨道根数快照，为轨道传播与预测模型提供了标准化的输入数据。研究人员利用SGP4/SDP4等标准传播模型，结合该数据集的高频更新特性，能够对超过1800颗卫星的未来位置进行高精度仿真与预报，尤其适用于对轨道衰减敏感的低轨卫星群的动态分析。

实际应用

在实际工程与运营层面，该数据集支撑着广泛的航天任务与商业服务。卫星运营商利用其进行在轨卫星的日常轨道保持与机动规划，确保通信、导航服务的连续性。空间态势感知机构则依赖这些数据监测太空物体，预防轨道碰撞，保障太空资产安全。此外，该数据集也为全球导航增强系统、遥感数据精准对地定位以及新兴的低轨互联网星座的网络优化与覆盖分析提供了不可或缺的轨道参数输入。

衍生相关工作

围绕该数据集，已衍生出多个重要的研究方向与工具。在学术层面，它催生了基于机器学习的轨道异常检测与寿命预测模型。工程上，它被集成到如OreKit、STK等专业轨道仿真软件以及开源库python-sgp4中，成为轨道传播算法的标准测试集。同时，该数据集也是构建更宏大历史轨道数据库（如space-track-tle-history）和卫星状态分类数据集（如constellation-census）的关键组成部分，共同构成了一个层次化的空间数据生态系统。