Deep learning solution for central-node integration challenge in clustered routing protocols during fire emergencies

Abstract Many clustered routing protocols integrate mediator nodes within the network structure, such as relays, gateways, and forwarders. These nodes collect and aggregate data from the clusters and then submit the data to the Base Station (BS), facilitating communication between clusters. It eliminates the redundancy in data and minimizes the delay by reducing the transmission distance to the BS. However, integrating such nodes poses significant challenges in dynamic environments, such as fire incidents. Based on their location and responsibility, their loss often leads to increased packet loss and suboptimal routing decisions under changing network conditions. This paper proposes a solution to this challenge by introducing a novel adaptive routing protocol. A hybrid deep learning model is employed to predict the loss of such nodes and select a suitable alternative node, combining dual-layer convolutional neural networks (CNNs) and bidirectional long short-term memory (BiLSTMs). The proposed protocol was validated against the non-adaptive approach during three scenarios and two fire directions. Key evaluation metrics used are the throughput, packet delivery ratio (PDR), packet loss ratio (PLR), delay optimization, and packet prioritization ratio. Simulation results demonstrate the effectiveness of the proposed approach in enhancing the central node’s performance during fire incidents and ensuring reliable delivery of monitored data to its destination. For the first fire scenario, the proposed approach records 8.08% PLR and 91.91% PDR, while for the second fire scenario, it records 8.27% PLR and 91.27% PDR. This study provides a robust framework for central node-based and clustered protocols, ensuring reliable communication in critical scenarios.