Machine Learning-Based Multi-Layer Multi-Hop Transmission Scheme for Dense Networks

Multi-hop communication has attracted a lot of attention recently due to its ability to extend the coverage range and to overcome blockage. In this letter, we propose a machine learning-based selection approach that adaptively chooses the best forwarding scheme in hybrid multi-hop dense networks. The proposed transmission scheme employs a multi-layer selection where each layer represents one possible relaying case, namely amplify-and-forward, half-detection, or full-detection of the transmitted symbol, or even no-relaying. Moreover, the proposed system dynamically learns the proper forwarding scheme out of these layers for each involved relay to minimize the transmission error rate based on the relay location, and its residual energy. A heuristic approach is proposed for the forwarding scheme selection and transmission power control where a minimum threshold value for the transmission power of each relay node is derived in order to satisfy a target QoS requirement. The results are used for training a decision trees-based prediction model that achieves a remarkable accuracy beyond the 99 for both training and testing patterns.