The near future is full of missions to the Moon that will need a robust communication system. This is where ESA’s Moonlight project fits into place [1]. However, the question of how to establish multi-hop, near real-time communication services between various assets on and around Moon and Earth is still to be answered. The Delay/Disruption Tolerant Networking (DTN) Bundle Protocol (BP) provides an excellent starting point for this to be developed, since it provides end-to-end communications for highly stressed environments, that is, for environments with large delays and intermittent connectivity [2], [3]. However, the current version of the protocol does not provide the means to guarantee bounds for latency or data loss [4]. Consequently, an extension to the BP is required, providing the means for measuring, asserting and optimizing Quality of Service (QoS) parameters such as latency, data loss and jitter, hence guaranteeing a reliable and robust Moon-to-Earth connection. Within this extension, routing via alternative routes, available bandwidth, priorities, scheduled and opportunistic contacts must be taken into account for optimization of the required QoS [5]. Moreover, it is not only important for the real-time services to be developed, but also for them to successfully coexist with non-real-time services, such as store-and-forward scheduled data transmissions or messaging services. Furthermore, the future creation of a large scale communication network must be a persistent goal [6]. For the BP extension to be efficient and achievable at a larger scale, Pattern Recognition (PR) and Machine Learning (ML) will have to be used. This approach will help identify which combination of parameters is more reliable and efficient in each moment given the dynamically changing network topologies, and it will enable the optimization of these parameters through reinforcement learning.
