Enabling Software-Defined Satellites with Edge AI

Software-Defined Satellites (SDSs) enable spacecraft to evolve on orbit through software, moving beyond fixed-function architectures toward adaptable, multi-mission platforms. Software-defined radios and reconfigurable sensors and onboard compute systems provide foundational flexibility, while the integration of artificial intelligence at the edge is critical to achieving true operational autonomy.

This presentation examines how AI-enabled onboard processing allows SDS architectures to dynamically optimize communications, sensing, and mission tasking. Drawing on Sidus Space’s experience designing and integrating flight-ready, software-defined spacecraft, the talk explores how performing inference and decision-making directly on orbit can reduce latency, manage limited downlink capacity, improve operational efficiency, and prioritize the transmission of high-value data. These capabilities are increasingly important for responsive Earth observation, flexible and persistent communications, and shared satellite platforms.

The presentation also addresses the practical challenges of deploying AI in resource-constrained space environments, including power, thermal, and radiation considerations, system design and test, and the verification of autonomous behaviors. Ultimately, tightly coupled edge AI is essential for SDS platforms to function as intelligent, cooperative nodes within distributed space infrastructures.