Energy Aware Congestion Control Mechanism for AIoT-Enabled WSNs

In todays era, the Artificial Intelligence of Things (AIoT) plays a crucial role in monitoring various physical parameters within industrial environments. AIoT is deployed to monitor toxic gas exposure, overheating machinery, and fires. It helps to prevent hazardous conditions within the industry. These AIoTbased systems use consumer Wireless Sensor Networks (WSNs) to collect real-time data from the monitoring environment. However, this network suffers from congestion issues due to a large number of multidimensional data packets. It results in packet loss, high transmission delays, and reduced network lifetime. The primary reason for network congestion is the bad clustering of Sensor Nodes (SNs) and sub-optimal route determination. SNs are unable to buffer large amounts of incoming data packets due to bad clustering mechanisms and sub-optimal data routing route determination. It causes packet loss and frequent retransmissions. Existing approaches are ineffective in minimizing packet loss and retransmissions in AIoT environments due to high traffic volumes. They suffer from poor network performance. Therefore, this paper proposes an energyaware congestion control mechanism for AIoT-enabled WSNs. An Improved Bat Algorithm (IBA)-based unequal cluster formation scheme is proposed to balance load among the SNs. The proposed scheme divides the entire network into an optimal number of unequal clusters. As a result, energy consumption is reduced, and energy usage is more evenly distributed among the deployed SNs. It also helps to prevent node overloading and minimizes data packet losses. Furthermore, a Q-learning-based congestion control and optimal data routing scheme is proposed to mitigate congestion within the network. In the proposed scheme, Q-learning agent dynamically improves routing decisions for delay-sensitive data packets. It significantly reduces transmission delays and improves the overall performance of the network. Extensive simulations and testbed experiments are carried out to evaluate the performance of the proposed scheme. The simulations and experimental results demonstrate that the proposed scheme achieves up to 37.85% average improvement in throughput, 38% improvement in network lifetime, 37% reduction in average energy consumption, 41.10% reduction in delay, and a 39.54% average reduction in packet loss rate as compared with existing state-of-the-art algorithms.