Comparison of Q-Learning and Genetic Algorithm for Narrow-Band Cognitive Radio Networks.

Narrow-band communication is widely used in military operations. Due to the limited bandwidth, efficient usage of the available spectrum with limited overhead is required. In this work, we present an efficient Dynamic Spectrum Management model for distributed multi-hop networks using narrow-band waveform. The system design is made robust using a collaborative feedback technique. Each node maintains a channel availability matrix and historical channel performance information based on this feedback. The channel allocation and radio parameters are optimized using Genetic Algorithm and Q-Learning algorithm. We observe that the robust system design, along with Genetic Algorithm and Q-Learning, efficiently mitigates the interference impact on a transmission link and consequently improves the overall transmission success rate and throughput for an end-to-end transmission.