Cognitive Communications

The Cognitive Communications Project is developing, implementing, and infusing cognitive communication technology into the Space Communication and Navigation capability of NASA. The goal is to increase efficiency and resiliency across the entire network stack, from point-to-point links, to routing between multiple nodes, to integrated scheduling of space and ground relay services. A cognitive system can mitigate obstacles, respond to and learn from its environment, and achieve beneficial goals towards the completion of its primary mission. A cognitive engine (CE) is a decision-making algorithm that enables part of a cognitive system, and there can be multiple CEs interacting in a cognitive system.

Broadly speaking, a CE can be implemented in many ways utilizing different decision-making methods, including those based on machine learning, as long as the methods align to the goals of the overall cognitive system. The complexity of a CE is driven by the number of objectives that must be optimized simultaneously and the availability and reliability of input knowledge.

The Cognitive Communications project performs research in four distinct but intertwined cognitive areas:

• Links – concerning point-to-point connections between two devices
• Networks – concerning multiple devices routing information among multiple links
• System Management – concerning the interaction among devices (or networks) and supporting ground- and space-based infrastructure

CEs applied broadly across all levels of the protocol stack will determine link optimization, network routing, and system management. While each of these focus areas can mature independently, the end goal is to transition towards an overall cognitive system-of systems, optimized across all layers. The spacecraft itself and the communication provider networks must perform joint cross-layer, distributed decision-making that conforms to the mission objectives and network capabilities.

Examples of technology development within the prime focus areas include:

• Cognitive Links – Optimization strategies for point-to-point links, such as adaptive coding and modulation augmented with artificial intelligence to predict future link margins, rapidly adjust bandwidth and center frequency to avoid interference, and share spectrum with both cooperative and uncooperative entities. Lower Technology Readiness Level (TRL) research in this area includes the use of signal sensing to automatically configure a receiver to match the incoming signal, or the use of deep learning to completely model a communication link, which enables radios to learn to communicate optimally in an unpredictable environment.
• Cognitive Networks – Routing and data management strategies for high-delay and disruption-prone space environments. This work moves beyond traditional Delay Tolerant Networking (DTN) concepts to include the use of artificial intelligence algorithms to predict future routes based on past performance. Routing data between multiple DTNs provides opportunities for other cognitive augmentations, such as data processing en route and automatic tuning of DTN parameters. Another related concept under development in this area is dynamic reassembly of data fragments sent to multiple space relays and ground stations, or “Drop Data Anywhere.”
• System Management – Scheduling techniques for user missions that do not require operator intervention. User Initiated Service (UIS) allows mission spacecraft radios to schedule services directly with a communications relay or ground station using machine-to-machine communication protocols. An event manager software package handles all UIS requests, allowing optimal allocation of communication resources to the mission based on priority, quality of service, cost, capability, and other considerations. The UIS capability, when merged with artificial intelligence techniques, will converge toward near-optimal use of all network resources. As an added benefit, the event manager can serve as a predictive failure detection engine by detecting mission issues across multiple relays, or relay issues across multiple missions, identifying potential problems before the problems become failures.

The Cognitive Communications project demonstrated several cognitive capabilities using the SCaN Testbed on the International Space Station from 2015 to 2019. SCaN Testbed demonstrated cognitive link capabilities (including the first known space-to-ground link controlled entirely by an artificial intelligence algorithm) and implemented UIS using NASA's Space Network assets. SCaN Testbed was decommissioned in May 2019, and further development of cognitive capabilities transitioned to ground testbeds.

Currently, the Cognitive Communications project is building prototype cognitive communication systems for use in Earth orbiting spacecraft and lunar operations. Each prototype system is evaluated on a ground tested, allowing direct comparison of key performance indicators and an analysis of alternative technology approaches. Ground demonstrations will continue to advance the TRL of cognitive enabling technologies and result in a cohesive cognitive communication system that could be utilized in future technology demonstration missions.