dSGP4 is a differentiable SGP4 program, which also supports ML-enhanced orbital propagation.
copied from cf-staging / dsgp4dsgp4 is a differentiable SGP4 program. It is written using PyTorch, which enables automatic differentiation through SGP4 inputs and parameters, as well as batch propagation (across different TLEs), with the possibility of leveraging embarassingly parallel computations both on CPU and GPU. It also offers the possibility of enabling a machine learning module that can enhance SGP4 propagation learning from high-precision simulated or observed orbital data. It also features a TLE module for easily parsing, interfacing with and constructing Two-Line Elements data.