COGNITIVE ANTI-JAMMING TDMA COMMUNICATIONS IN CONTESTED ENVIRONMENTS WITH REAL-TIME JAMMER IDENTIFICATION

To address the performance degradation of fixed scheduling policies used by TDMA-based tactical networks in contested environments, Bluecom Systems proposes to develop a cognitive anti-jamming and communications scheduling framework. Proposed cognitive protocols can be implemented in hardware that can be integrated in to TDMA-based tactical software-defined radios (SDRs), converting them to autonomous cognitive radios. Technical approach consists of: Designing a machine-learning based real-time hierarchical signal classification framework to distinguish between jammers and inadvertent interference, designing a reinforcement-learning (RL) aided cognitive anti-jamming and scheduling protocol for TDMA-based radios and the integration of these two modules. The most suitable classification engine and the feature inputs will depend on the type of jammers and interference assumed. During Phase I, artificial neural networks (ANN) and convolutional neural networks (CNN) will be explored as the classification engines. Assuming commonly encountered jammer signal models, various feature vectors will be used as inputs to these to determine the most suitable classifier and the feature sets. For cognitive anti-jamming policy learning, several RL algorithms will also be evaluated against these jammer models. A hardware-in-the-loop demo will be developed, utilizing commercial radios, to evaluate the developed cognitive anti-jamming and scheduling system in the presence of real jammer and interference.