HYBRID QUANTUM-INSPIRED GRAPH NEURAL NETWORK MODEL FOR ADAPTIVE RESOURCE ALLOCATION IN INTELLIGENT IOT-CLOUD ENVIRONMENTS

The proliferation and scale of IoT deployments continues to increase, and finding the right amount of cloud resources to match the workload is becoming more challenging than ever. The effect of fixed allocation rules is eroded by changing demand, while the machine-learning schedulers that calculate the score of each task individually fail to capture the dependency that governs IoT-cloud traffic. In this paper, a hybrid Quantum-Inspired Graph Neural Network (QI-GNN) is introduced for the resource allocation problem in an IoT-cloud environment. The model builds a task-dependency graph directly from incoming IoT workload traces, enriches each node with quantum-inspired superposition and entanglement encodings, then runs GraphSAGE message passing paired with an actor-critic reinforcement learning policy to allocate resources in real time. We tested QI-GNN on the Google Cluster Trace v3 dataset against four baselines, DQL, DDPG-D3QN, GNN-RL, and FedRL. At 1000 concurrent tasks it beat the best of these, FedRL, by 11.4% in resource utilization, 25.7% in average completion time, 21.0% in energy consumption, 11.2% in throughput, and 42.7% in SLA violations.