eduzhai > Applied Sciences > Engineering >

Deep Reinforcement Learning Control for Radar Detection and Tracking in Congested Spectral Environments

  • Save

... pages left unread,continue reading

Document pages: 15 pages

Abstract: In this paper, dynamic non-cooperative coexistence between a cognitive pulsedradar and a nearby communications system is addressed by applying nonlinearvalue function approximation via deep reinforcement learning (Deep RL) todevelop a policy for optimal radar performance. The radar learns to vary thebandwidth and center frequency of its linear frequency modulated (LFM)waveforms to mitigate mutual interference with other systems and improve targetdetection performance while also maintaining sufficient utilization of theavailable frequency bands required for a fine range resolution. We demonstratethat our approach, based on the Deep Q-Learning (DQL) algorithm, enhancesimportant radar metrics, including SINR and bandwidth utilization, moreeffectively than policy iteration or sense-and-avoid (SAA) approaches in avariety of realistic coexistence environments. We also extend the DQL-basedapproach to incorporate Double Q-learning and a recurrent neural network toform a Double Deep Recurrent Q-Network (DDRQN). We demonstrate the DDRQNresults in favorable performance and stability compared to DQL and policyiteration. Finally, we demonstrate the practicality of our proposed approachthrough a discussion of experiments performed on a software defined radar(SDRadar) prototype system. Our experimental results indicate that the proposedDeep RL approach significantly improves radar detection performance incongested spectral environments when compared to policy iteration and SAA.

Please select stars to rate!

         

0 comments Sign in to leave a comment.

    Data loading, please wait...
×