Performance of AI-Empowered Anti-Jamming Communications under Hardware Impairments

The present contribution quantifies the effect of Hardware Imperfections (HWI) on Artificial Intelligence (AI)empowered anti-jamming wireless communication scenarios. To that end, we consider an AI-empowered wireless link that employs reinforcement learning (RL) to counteract reactive jamming attacks. It is shown that contrary to the common belief that HWI always degrade the performance of wireless systems, our findings reveal that imperfections in the jammer's spectrum sensing capability can actually enhance its effectiveness against AI-empowered anti-jamming wireless communications. Extensive simulations demonstrate that a non-ideal reactive jammer with detection errors (e.g., false alarms and miss-detections) can perform significantly better than an ideal reactive jammer while using the same jamming resources. This improvement is attributed to the randomness induced in jamming patterns due to sensing errors, which distorts the learning process of the antijamming AI agent resulting in a considerable reduction in the anti-jamming capability of the AI-empowered wireless link.