Integrating Deep Reinforcement Learning (DRL) with GAACO for Resource Scheduling in Cloud Computing

Abstract

Efficient scheduling of resources is a major concern in cloud computing, which is mainly due to dynamic workloads, scarcity of resources and requirement for quality of service (QoS). This paper develops a hybrid dispatching model, which combines Genetic Algorithm-Ant Colony Optimization (GAACO) with Deep Reinforcement Learning (DRL). The hybrid algorithm was implemented and evaluated using CloudSim 3.0.3 to improve the adaptivity and efficiency. Experimental results demonstrate that GAACO+DRL consistently outperforms GAACO alone, reducing makespan by up to 30%, lowering average waiting time by 20–35%, improving throughput, balancing workload distribution, reducing energy consumption, and completely eliminating SLA violations. These findings highlight the effectiveness of combining met heuristic optimization with reinforcement learning to achieve stable, efficient, and scalable resource scheduling in cloud computing.