Abstract

The recent accelerated growth of Internet of Things (IoT) implementations in smart environments has amplified the situation regarding the energy efficiency and computational scalability coupled with dynamic resource management in heterogeneous and timevarying conditions. Old-fashioned centralised and non-adaptive based optimization mechanisms are growing to be ineffective mainly because they are very rigid, and they consume a lot of overhead. In order to overcome these limitations, the given paper will introduce an AI-assisted hybrid optimization strategy which incorporates the principles of neural computing with those of genetic algorithm (GA)-run evolutionary optimization to facilitate resource-optimal and intelligent performance of IoT-driven smart surroundings. Under the suggested methodology, lightweight neural computing frameworks implemented at the edge tier offer real-time local awareness and forecast estimation of the workload, energy requirements and network states of affair. A multi-objective GA then uses these predictions to produce dynamic optimization of the critical system parameters such as task scheduling, duty cycling, transmission power and edge-cloud offloading decisions. Closed-loop feedback allows the hybrid neural to act proactively and to implement global optimization and dynamically adapt itself through a hybrid neural structure. Comprehensive simulations of the framework on the basis of different node densities and traffic conditions prove that the proposed framework outperforms the traditional heuristic, neural-only, and GA-only strategies through optimization significantly. Experiments have shown significant energy savings and latency at the end, and significant increases in network lifetime and Quality of service (QoS) sustainability. The solution proposed is scalable, computationally efficient and is well adapted to the deployment in next-generation smart cities, smart buildings and industrial IoT systems that need smart, autonomous and resource-aware operation.