A Game-Theoretic Model for Enterprises Cloud Computational Resource Allocation Based on Predefined-Time Distributed Algorithm

In the era of smart manufacturing, the way enterprises handle tasks has been revolutionized by cloud computing, shifting from on-premises IT environments to the cloud. However, the competition among multiple enterprises for limited computational resources available at cloud service providers presents a significant challenge in fairly allocating resources to minimize response delays. To offer a reasonable allocation strategy for cloud service providers in a multi-enterprise cloud environment, this paper introduces a novel aggregative game model within a three-tier computational offloading architecture across local manufacturing devices, resource-constrained small base stations, and high-capacity cloud data centers. Due to the aggregation of enterprises' strategies, traditional evolutionary algorithms are infeasible to solve this game problem. Stemming from recurrent neural networks, a predefined-time distributed generalized Nash equilibrium seeking algorithm with event-triggered communication is proposed. From a communication perspective, the event-triggered distributed setting ensures that enterprises engage in discrete-time local communication, aligning with the practical, competitive and information-preserving nature of the enterprise interactions. From a computational perspective, the algorithm allows cloud service providers to autonomously balance computational efficiency and decision accuracy, with user-defined convergence times for decision updates. This enables the rapid determination of equilibrium strategies, providing cloud service providers with a scalable solution that was unattainable in previous methods.