An ACO-LSTM Approach for Energy-Efficient Cluster Head Selection in Wireless Sensor Networks

Energy-efficient cluster-head (CH) selection is essential in wireless sensor networks because CHs carry the additional cost of reception, aggregation, and long-range forwarding. This paper presents a hybrid ant colony optimization and long short-term memory (ACO-LSTM) approach that combines multi-criteria search with temporal prediction. ACO explores candidate CH sets using residual energy, communication distance, coverage, and load-related information, while LSTM estimates each node's future suitability from recent energy and role histories. The two outputs are fused to rank CH candidates at every round. The method is evaluated against LEACH in networks of ۱۰۰-۳۰۰ nodes, five packet-generation levels, and two fixed sinks. Evaluation uses packet delivery ratio (PDR), first-node-death (FND) time, network energy, throughput, delay, and topology evolution. The results show that the proposed method maintains an average PDR of about ۰.۶۵۵ for ۱۵۰-۳۰۰ nodes, whereas LEACH levels off near ۰.۵۲۵. ACO-LSTM also achieves a slightly later FND for most tested network sizes while keeping delay stable in the temporal experiment. These findings indicate that temporal awareness improves reliability without sacrificing network stability, making the method suitable for energy-constrained sensing and Internet-of-Things applications.

کلیدواژه ها:

wireless sensor networks ، cluster-head selection ، ant colony optimization ، LSTM ، LEACH ، energy efficiency