Optimised route selection for multipath routing in MANET using latent encoder coupled generative adversarial network

Mobile ad hoc networks (MANETs) enable decentralised communication in environments without traditional infrastructure, such as disaster zones and military settings. However, they face challenges like high latency and energy inefficiency in route selection, affecting overall network performance. To address this, an optimised route selection for multipath routing in MANET using latent encoder coupled generative adversarial network (LECGAN-RS-MR-MANET) is proposed. Initially, the tyrannosaurus optimisation algorithm (TOA) is employed for efficient route exploration, followed by LECGAN for optimal node selection through latent feature learning. The improved sooty tern optimisation algorithm (ISTOA) ensures effective route maintenance, while the multitask multi-attention residual shrinkage convolutional neural network (MMRSCNN) handles route recovery. The proposed LECGAN-RS-MR-MANET is evaluated and demonstrates 23.92% lower energy consumption, 24.92% lower end-to-end delay, and 22.42% lower routing overhead compared to existing techniques. These findings confirm that the proposed model significantly improves routing performance and reliability in MANETs.