ASAP 2024 Year in Review

 The WP2 team has defined and implemented a suite of machine learning (ML) algorithms running on consumer-grade computing systems to support on-board science operations for space missions. Laying the groundwork for a significant leap forward in the execution and scientific exploitation of exploratory and research-driven space missions, enhancing their capabilities and impact. In more detail, the WP2 team has studied, developed and optimized several algorithms that pertain to four main use cases: 

1. Autonomous Triggering Burst-Mode and Selective Data Downlink. The team investigated algorithms for the recognition of plasma regions, particularly in Earth’s magnetosphere. Two convolutional neural network (CNN) algorithms were studied to enable autonomous triggering of special measurement modes with increased data collection rate (Burst-Mode) and efficient selective downlink of plasma environment parameters. These advancements hold promise for mitigating resource constraints in plasma missions under consideration, allowing more efficient data transmission.
2. Advanced On-Board Analysis of 3D Particle Distribution Functions.  For velocity distribution functions (VDF) analysis, two clustering-based algorithms capable of distinguishing different populations in the proton distribution functions of the solar wind and in the electron distribution function in the magnetosphere were considered and investigated in depth.      These algorithms would allow us to significantly augment the capability to exploit the plasma mission scientific output, enabling advanced onboard data analysis and efficient selective downlink.
3. Onboard analysis of solar images. To improve solar image processing, four algorithms of mainly two types, Variational Autoencoder (VAE) and segmentation structure, have been studied. These can be used for both structure recognition and reduction of data downlink,      paving the way for more efficient and insightful on-board solar imaging analysis. 
4. Onboard prediction capability of SEP-related hazards. The ESPERTA model for predicting SEP-related hazards was upgraded by the WP2 team. This is the first algorithm for SEP prediction on-board human and science exploration missions, addressing critical safety and mission planning needs.

Moreover, the WP2 team carried out the fundamental activity that permitted the identification and definition of the algorithm’s requirements for the selected in-flight scenarios. These requirements are essential for the design of both the hardware and the software parts of the testbed. In particular, the implementation of custom AI/ML hardware accelerator and the definition and execution of the testbed validation plan must consider the requirements identified by the analysis performed within WP2. Through these advancements, the WP2 team has significantly bolstered the technological foundation for future plasma and space exploration missions, ensuring improved scientific output and resource efficiency. 

Looking ahead, WP2 will focus on supporting the ongoing efforts in WP4 and WP3, providing specific information on the ML algorithms for a refined definition of their requirements needed in the framework of the porting activity. The team will also continue optimizing the algorithms based on feedback from WP3 and WP4, addressing potential software and hardware resource limitations identified during subsequent phases of the project. This iterative process ensures seamless integration of WP2’s advancements into the broader project framework

 The focus of WP3 during 2024 has been on reviewing the state-of-the-art Neural Network (NN) inference for spacecraft using FPGAs. This effort culminated in a comprehensive report that summarizes key advancements in the field, current practices, and the unique requirements for deploying NN inference on FPGAs in space. The report also highlights the most common applications of this technology onboard spacecraft, serving as a valuable resource for future developments.

Building on this foundation, the team shifted attention toward analyzing and developing initial implementations of the NNs designed in WP2. Using High-Level Synthesis (HLS), these NNs are being synthesized into hardware to ensure efficiency and reliability. One notable aspect of this work includes leveraging the work from WP2 on CNN for plasma region classification tailored to data from NASA's Magnetospheric MultiScale (MMS) mission. Currently, efforts are centered on optimizing and refining the smallest version of this model to further enhance its performance.

WP3’s progress represents a significant step forward in the integration of advanced neural network capabilities into space missions, promising improved onboard data analysis and resource management.

Throughout 2024, WP4 focused on designing and developing a crucial software interface layer for the ASAP testbed, enabling seamless communication between high-level machine learning frameworks (such as TensorFlow, PyTorch, and ONNX) and the hardware used in space missions. This interface layer utilizes a soft-GPU, an application-independent FPGA design that accelerates model inference, enhancing the speed and efficiency of onboard computations for real-time analysis and decision-making during space missions.

The soft-GPU is being developed by IngeniArs using Hardware Description Language (HDL), ensuring compatibility with a variety of radiation-tolerant FPGAs, critical for space applications. This flexible approach allows the soft-GPU to be implemented across different hardware environments, making it adaptable for various mission scenarios. Ongoing discussions between IngeniArs and KTH throughout the year have refined the design and implementation strategy. 

As we move into 2025, work on finalizing and integrating this software interface layer into the ASAP testbed will continue, supporting the next steps in advancing space mission technology.

In 2024, the WP5 team developed a sophisticated software tool that enables virtual spacecraft and constellations to navigate and conduct measurements within 3D simulation outputs. At the core of this tool is a robust bilinear interpolation algorithm, which allows the virtual spacecraft to simulate real satellite behavior. This includes traversing complex trajectories and collecting data on evolving physical parameters such as electromagnetic fields, plasma density, and velocity.

The tool has been further enhanced to support constellations of satellites arranged in a tetrahedral formation, enabling the measurement of spatial gradients and higher-order physical quantities. Each satellite in the constellation can move independently, allowing us to simulate dynamic, real-world conditions and assess the geometric integrity of the formation under varying scenarios. To ensure reliability, we implemented control parameters to monitor and optimize key properties like elongation and planarity, which are critical for maintaining accurate measurements.

Currently, the WP5 team are working on the development of virtual instruments to mount onboard these virtual spacecraft. This includes designing a virtual top-hat analyzer to measure the three-dimensional particle velocity distribution and an energetic particle detector to study high-energy particle dynamics. These virtual instruments will significantly enhance the capabilities of the software, enabling it to simulate a wider range of in-situ measurements and further advancing its utility for space physics research and mission preparation.

This work underscores WP5's commitment to providing innovative tools that pave the way for more advanced and realistic space mission simulations.

 In 2024, WP6 concentrated on building the essential groundwork for the ASAP Project’s dissemination activities. The focus has been on creating a strong and interconnected communication infrastructure to ensure the project's ongoing visibility and engagement with our growing community.

• Website: Established as the central hub for information, the website is the go-to destination for everything related to the ASAP Project. Here, visitors can access project updates, resources, and key insights as the initiative progresses.
• LinkedIn Page: Designed to share updates and celebrate accomplishments, our LinkedIn presence fosters engagement with a broad audience. Through this platform, we aim to connect with industry professionals, partners, and the wider community to create a dynamic network of support and collaboration.
• YouTube Channel: While still in its initial stages, the YouTube channel is set to become a key repository for educational content and seminar presentations.      This platform will enhance our ability to disseminate knowledge, offering an accessible way to engage with the project’s insights and findings.

These initiatives collectively strengthen WP6's role in ensuring that the ASAP Project’s achievements and vision reach far beyond 2024, paving the way for impactful dissemination in the years to come.

The ASAP Project team actively participated in several conferences during 2024, sharing our progress, insights, and innovations with the broader scientific and space exploration community. These events provided valuable opportunities to engage with experts, exchange ideas, and highlight our contributions to advancing space mission technologies.

For a detailed overview of the conferences we attended and the topics we presented, visit our Events page here. Stay tuned for more updates as we continue to engage with the global research community! 

As we reflect on the milestones reached in 2024, we are inspired by the synergy, innovation, and ambition demonstrated across all work packages. The ASAP Project is not only paving the way for groundbreaking advancements in space mission technology, but also fostering a community dedicated to exploration, discovery, and collaboration.

Looking ahead to 2025, we remain committed to building on this solid foundation, with each work package striving to further expand the boundaries of what’s possible. Together, we are making significant strides toward realizing the transformative vision of the ASAP Project.

Warm regards,

The ASAP Project Team