WP2: Algorithm Selection and Development

---

Work Package 2 (WP2) focuses on the development and analysis of machine learning (ML) algorithms to enhance autonomous decision-making and data processing for space science missions. These algorithms will improve how spacecraft handle scientific data, reducing reliance on ground-based operations and optimizing data transmission.

Key areas of development include:

• Autonomous data management: Implementing ML techniques to selectively downlink plasma environment parameters and trigger special measurement modes based on real-time data analysis. Using deep learning, including Short-Time Fourier Transforms (STFT) and Convolutional Neural Networks (CNNs), the system can detect regions of interest and activate enhanced data collection.
   
• On-board particle data analysis: Developing AI-based methods for analyzing three-dimensional particle distribution functions (VDF). One approach reconstructs VDFs using image inpainting techniques, while another employs unsupervised learning to classify particle populations directly from raw data. These advancements enable real-time plasma characterization and efficient data transmission.
   
• Solar image processing: Leveraging Convolutional Autoencoders and Variational Autoencoders (beta-VAE, NVAE) to analyze multi-channel solar images. These models enhance feature extraction, supporting improved space weather monitoring and event detection.
   
• Space radiation prediction: Adapting the ESPERTA model for on-board forecasting of solar energetic particle (SEP) hazards, with real-time parameter adjustments based on the spacecraft’s position in the heliosphere. This predictive capability is crucial for protecting both spacecraft and future crewed missions.
   

WP2 works closely with other work packages to refine and validate these ML algorithms, integrating them with simulated and real mission data. The key outcomes of WP2 include the development of deployable ML techniques, scientific publications, and policy recommendations for the research community and space exploration stakeholders. These contributions will help pave the way for AI-driven advancements in future space missions.