dsgp4
Differentiable SGP4. Supports differentiability, ML integration & TLE parallel computations.
dSGP4 is a differentiable SGP4 program written leveraging the PyTorch machine learning framework: this enables features like automatic differentiation and batch propagation (across different TLEs) that were not previously available in the original implementation.
The project is a collaboration with the Oxford AI4Science Lab.
Notable features include:
- differentiable version of SGP4
- hybrid SGP4 and machine learning propagation: input/output/parameters corrections of SGP4 from accurate simulated or observed data are learned
- parallel TLE propagation (supports CPU, NVIDIA GPU, GPU for MacOS devices with Metal programming framework, etc.)
- tutorials on the use of differentiable SGP4 on several spaceflight mechanics problems (state transition matrix computation, covariance transformation, and propagation, orbit determination, ML hybrid orbit propagation, etc.)