Deep Learning Approach to Satellite Collision Avoidance Using Long ShortTerm Memory

Satellites are essential to our contemporary way of life in a variety of fields,

including communication, navigation, crisis management, and science. With an increasing

number of satellites, spacecraft, and other objects launched into space, collisions occur, and

the amount of debris in orbit increases, satellite collisions are a key problem in space

operations, and there are various challenges and issues, as a result, accurate and efficient

orbit forecasting is becoming more important for increased space situational awareness.

Deep learning (DL) has the potential to improve orbit prediction and collision avoidance in

space by improving accuracy, automating responses, and providing real-time decision

assistance. The proposed approach has the potential to significantly improve the safety and

sustainability of satellite operations in increasingly crowded orbital environments. This

paper introduces a satellite orbit forecasting model based on historical position and velocity

data that uses long short-term memory (LSTM) and benchmark data to avoid satellite

collisions by anticipating satellite position and velocity. According to the results, the model

of LSTM can increase the precision of orbit prediction, with position accuracy ranging in Rsquare (R2) from 95 to 99% and for mean absolute error (MAE) from 0.05 to 0.02.