MAXENTRIC TECHNOLOGIES LLC — Department of Defense SBIR Phase I: A18-051

Description

There is a definite need for tactical ad-hoc networks to understand to which degree communications can still occur in congested and contested environments where power sources are short-lived, the spectrum is limited and where jamming (both friendly and adversarial) can occur spontaneously and without warning. In this situation, network nodes will need to cooperate and coordinate negotiate ways for them to still pass information to each other despite being embedded in a congested and contested environments. Our solution will take results from machine learning and artificial intelligence and combine those with a software defined networking (SDN) concept designed for challenged tactical networks. In particular, Markov decision process (MDP) and Q-learning will be used in a reinforced learning (RL) approach to define data routing policies and media access policies in light of network conditions as input (congestion, jamming, channel fading, data delay, data bandwidth/throughput, packet loss, node mobility, node battery power level etc.). Our SDN design contains functions and optimized data storage to enable network nodes to operate in intermittently connected network conditions, which can be considered normal operating conditions in a tactical edge network.