S.T.A.R.S

I built a router for LEO satellital constellations, it uses a DRL Graph neural network to learn the characteristics of a path that tends to fail in order to avoid it. Its supposed to replace the traditional dijkstra-based path finding providing a more stable and reliable way to transfer data, when the model choose a is run it chooses 3 paths and distributes the data to send amongst this three adjusting flow to each one individually based on congestion, bandwith and other factors

It also compresses data (telemetry mostly) and separate it into (n) quantity of files (determined by the compressed size) to ease up a little the traffic