Orbital Chaos

Deep learning meets orbital mechanics. We trained LSTM, Transformer, and multi-modal models on 3 years of NASA spacecraft data to predict orbits and measure the impact of solar wind on trajectory perturbations.

0
Data Points
0
Spacecraft Tracked
0
2023 - 2025
0
Resolution

Spacecraft

ISS

LEO ~408 km

International Space Station. Low Earth Orbit with significant atmospheric drag. 1.58M data points. Most affected by thermosphere expansion during geomagnetic storms.

DSCOVR

L1 ~1.5M km

Deep Space Climate Observatory at the Sun-Earth L1 Lagrange point. 131K data points. Directly measures solar wind before it reaches Earth.

MMS-1

HEO Magnetosphere

Magnetospheric Multiscale Mission. Highly elliptical orbit cutting through Earth's magnetosphere. 1.54M data points. Most sensitive to space weather conditions.

Models

Bidirectional LSTM

Encoder-decoder LSTM with autoregressive decoding. Maps 24h of position/velocity history to 6-24h predicted trajectories.

PyTorch Sequence-to-Sequence
ISS: 126 km DSCOVR: 12,797 km MMS1: 18,683 km

Transformer

Encoder-decoder Transformer with sinusoidal positional encoding and learned query tokens. Multi-head cross-attention.

PyTorch Attention
ISS: 295 km DSCOVR: 13,517 km MMS1: 19,237 km

Multi-Modal Fusion

Dual-encoder combining orbit positions with solar wind measurements. Cross-attention lets the model learn which solar wind conditions most affect the trajectory.

Cross-Attention Solar Wind
ISS: 175 km DSCOVR: 25,059 km MMS1: 19,457 km

SGP4 / Kepler Baseline

Physics-based orbit propagation using two-body Keplerian mechanics. The benchmark to beat. ML models target short-horizon improvement.

Baseline Physics

Results

6-hour prediction MAE (Mean Absolute Error) in km. Trained on dual RTX 5090 GPUs via RunPod.

Model ISS (LEO) DSCOVR (L1) MMS-1 (HEO)

MAE Comparison (log scale)

Live ISS Tracker

Real-time International Space Station position on a 3D globe. Updated every 5 seconds.

--
Latitude
--
Longitude
--
Altitude (km)
--
Velocity (km/h)
--
Last Update

Space Weather

Current solar wind conditions from NOAA SWPC. These are the signals our multi-modal model uses to predict orbit perturbations.

Kp Index
--
Solar Wind Speed
-- km/s
IMF Bz
-- nT
Storm Level
--

Try It

Run orbit predictions directly in your browser using our trained models on Hugging Face.

Data Pipeline

1

NASA SSC API + OMNI

Automated daily fetch of spacecraft positions (GSE/GEO) and solar wind parameters (IMF, flow speed, density, Kp, Dst) via cron.

2

Preprocessing

Velocity derivation, normalization, gap detection, sliding window creation. L1-to-Earth propagation delay correction for solar wind.

3

Training

AdamW + cosine LR schedule, early stopping, gradient clipping. Temporal train/val/test split (70/15/15). Both PyTorch and TensorFlow.

4

Evaluation

MAE/RMSE in km at 1h, 6h, 24h horizons. Separate analysis for quiet (Kp ≤ 3) vs storm (Kp > 5) conditions.

Research

We hypothesize that during geomagnetic storms, the thermosphere heats and expands, increasing atmospheric drag on LEO satellites in ways that SGP4 cannot predict. By incorporating real-time solar wind data measured at L1 (~45 min before arrival at Earth), our multi-modal model can anticipate these perturbations. The dataset captures Dst down to -406 nT and Kp up to 8+ — severe storm events ideal for testing this hypothesis.

Data pipeline active — fetching new observations daily at 02:00 UTC