Adaptive and Scalable Data Compression for Deep Space Data Transfer Applications using Deep Learning

Data transfer has emerged as a fundamental and interdisciplinary challenge for NASA and its partners in space-based mission for sustained Earth and Solar observation, exploration of our solar system, and investigations of the cosmos. Advances in sensor technology and an increasing desire for a deeper understanding of the geospace environment (Sun to Earth and beyond) have resulted in an explosion of data volume in recent years (unprecedented spatial and/or temporal resolution as well as multispectral data) and requires new innovative data compression approaches. For example, the Atmospheric Imaging Assembly (AIA) onboard NASA’s Solar Dynamics Observatory (SDO) launched in 2010 recently captured its 200 millionth image in 4K resolution (4096x4096 pixels), and its archive is over 10 PB of data products. While most missions prefer to have the cadence and resolution of the SDO imager (an image every 12 seconds at 4K resolution), the science goals of nearly all missions are limited by their data capabilities. The challenge is further emphasized by the use of data collected by deep-space spacecraft for real-time applications such as space weather forecasting, the subject of an Executive Order issued by former President Obama and supported by the current administration. For example, European Space Agency’s (ESA’s) plan to launch a dedicated spacecraft ‘Lagrange’ to the Sun-Earth L5 point for operational space weather forecasting will contain a suite of instruments that will collect a large volume of data that will all need to be transmitted to Earth in (near) real-time.

The design of science payloads involves trade-offs between sensor fidelity, onboard storage capacity, throughput and flexibility of the data handling systems, and the bandwidth and error characteristics of the communications channel. The basic requirements that heavily influence performance design trade-offs are size, power, weight, and complexity. The fifth requirement and always the decisive factor is mission risk. Data compression (to a manageable size before downlinking to Earth) serves as an enabling solution to avoiding data bottlenecks and fulfill the performance trade-offs and reliability requirements. Our research is motivated by the need for innovative and robust data compression techniques to mitigate the challenges of cost, volume, and latency for data transmission in space applications.

Our goal is to leverage new data science/machine learning extensions made possible by contemporary computational resources to develop new deep learning algorithms/technology for lossy compression of image data that has the promise of superior scientific performance while minimizing data loss. The proposed project is highly timely because of the contemporary computational resources (e.g., GPU’s) available for deep learning, the large solar imaging data sets available from multiple NASA missions, and the unprecedented interest in the understanding of our Sun for its ability to significantly impact human life.

This is a mature technology (TRL 7+) — the realistic path in is usually NASA's Technology Transfer Program: licensing an existing NASA patent, or a Space Act Agreement to use NASA facilities/expertise directly. NASA also runs a startup licensing program with no upfront fee for companies formed to commercialize a specific NASA technology.

None of these are guaranteed paths for this specific project — TechPort itself doesn't have an "apply" button. Reaching out to the contact(s) above with a specific question is usually the fastest way to find out what's actually open.