What is it about?

Future space missions to the moon, mars, and beyond will require spacecraft systems with more autonomy. Autonomous capabilities allow spacecraft to make important decisions to achieve their mission when they are far away from human operators. Artificial intelligence (AI) flight software can enable onboard autonomy but only if it can meet the requirements of the space environment and flight hardware, like limited size, weight, power, memory, and radiation tolerance. To meet these requirements we can implement modern AI algorithms, like neural networks, across different parts of spacecraft flight computers. We have shown one of the first cases of splitting a neural network across different parts of a flight computer onboard a satellite, using a custom deployment toolkit we designed.

Featured Image

Why is it important?

Examples of data-driven AI operating on spacecraft are limited and this work documents an important technology demonstration showing how neural networks on spacecraft can be reprogrammed in flight, a step that helps de-risk use of neural networks in future space exploration and earth observation missions. It also demonstrates how the onboard computing resources of a satellite can be better utilized for AI applications

Read the Original

This page is a summary of: Deploying Artificial Intelligence Capabilities by Hybridizing a Neural Network on a Satellite, October 2023, American Institute of Aeronautics and Astronautics (AIAA),
DOI: 10.2514/6.2023-4718.
You can read the full text:

Read

Contributors

The following have contributed to this page