Space communication systems are operating in increasingly complex, congested, and contested environments. To provide effective communications despite these challenges, AI technologies – and a subset, Machine Learning (ML) technologies – are being developed and deployed in new ways. Small satellites and coordinated groups of satellites called constellations are of particular importance to the future development of space communications. These systems, when enhanced with AI/ML technologies, will be more capable, efficient, and robust than the older systems by autonomously sensing their environment, determining necessary actions on-board, and taking those actions to ensure communications are completed quickly and successfully.
In a communications system, the overall goal is always to meet the user’s communications requirements. Depending on the user, these can include maximizing data throughput, minimizing latency, maintaining uninterrupted or continuous communications, and/or ensuring data accuracy. The application of AI/ML technologies can be utilized throughout the communications process to address these or any other specific requirements to meet the user’s goals.
The AI evolution in small satellite communication systems is driven by the availability of more capable on-board processing hardware and by the need for autonomous operations, efficiency, and low end-to-end time latency. To meet these needs, AI and ML technologies can be applied at the radio, network, ground station, and/or end-to-end system levels and perform activities such as
- Autonomous operation
- Signal processing
- Transmission performance optimization
- End-to-end transmission routing
- Interference mitigation
- Efficient spectrum utilization
- On-board task scheduling
- On-demand scheduling with ground stations
With the associated complex satellite-to-user environment and operation demands, AI enhanced systems will allow for the timely transmission of information and reduce or eliminate direct human intervention in the communication process.
Comsat Architects has been applying AI and ML technologies to small satellite communication systems since 2015. Our first effort was a signal processing application for an auto-encoder neural network as a model of a node-to-node communication channel. Later efforts focused on the application of AI/ML technologies for autonomous operations, signal routing, and transmission link optimization. We continue to incorporate AI/ML technologies into small satellite communication systems for both commercial and defense applications. A key feature for these applications is the development of an integrated communication module that includes a Software Defined Radio and Single Board Computer, providing flexible communication systems with advanced processing power.
Reference:
M. McCaskey, E. Kukura, R. Corrigan, K. Bhasin and D. Chelmins, “Machine learning applied to an RF communication channel”, Proc. IEEE Nat. Aerosp. Electron. Conf. (NAECON), pp. 179-185, Jul. 2018.
