Reentry Event Detection and Prediction using Deep Learning-based Satellite Constellation Analysis

Our Reentry Event Detection and Prediction project leverages deep learning algorithms to analyze satellite constellation data, identifying reentry events and predicting impact locations and times with an accuracy of 95% or higher. We employ a combination of convolutional neural networks (CNNs) and recurrent neural networks (RNNs) to handle edge cases such as multiple reentry events and uncertain satellite constellation data. The web-based API interface utilizes JSON data formats, responding within 1-second latency, and integrates seamlessly with USSF operators' existing toolchains through standardized API calls. By addressing the technical challenges associated with reentry event detection and prediction, our solution supports the Space Domain Awareness (SDA) community by providing a software capability that automatically detects reentry events and predicts impact locations and times, enabling timely decision-making and enhanced situational awareness.