System of Systems for Distributed Disaggregated Communications via Reinforcement Learning and Backpressure (D2CRaB)

Abstract

This paper focuses on effectively communicating within distributed and disaggregated multi-domain battlespace operational system of systems (SoS) environments. To do so, the individual participating communications systems must be adaptive in response to ever-changing on-mission events such as, network congestion (e.g., degraded comms links and excessive network traffic), enemy interference (e.g., jamming, cyber-attacks), line of sight degradation (e.g., weather conditions). These challenges cause traditional communications systems networks to drop packets, which they presently do according to pre-specified Quality of Service (QoS) policies. However, these QoS policies alone do not actually mitigate the congestion problem because the impacted streams employ reliable communications protocols, such as TCP, that require dropped packets to be retransmitted over the same route. This approach can make the congestion problem even worse and waste the valuable bandwidth. The approach presented here introduces the novel System of Systems (SoS) for Distributed Disaggregated Communications via Reinforcement Learning (RL) and Backpressure (D2CRaB) to overcome the above stated limitations. D2CRaB does so in two ways: (1) by bridging route selection and congestion control via a new backpressure scheme and (2) by leveraging RL to enable the dynamic and continuous adaptation to changing network conditions. These new advancements, greatly extend the available bandwidth of the original congested communications system to that available of other communications systems within the multi-domain SoS. By so doing, congestion can be relieved, and packet drops can be greatly reduced.