Dual Decision-Driven Task Offloading and Resource Allocation for Multi-User Artificial Intelligence

With the development of 6G future networks and artificial intelligence (AI), there is an urgent need to integrate the ultra-high transmission rate and ultra-low latency characteristics of 6G into intelligent applications and carry out a series of optimizations to provide a better AI service experience. This article aims to reduce the average service completion time of multi-access edge computing (MEC) intelligent applications in 6G networks to improve service quality for user-intensive artificial intelligence image recognition scenarios. To simulate the real multi-user intelligent image recognition scene as realistically as possible, the image recognition AI task processing request is divided into two decisions through the dual decision queue architecture (DDQA), and an efficient hierarchical applicable heuristic algorithm (EHAHA) is proposed. After sending the task processing request, it is reasonably decided whether to unload and the appropriate server to which it should be unloaded. Finally, experimental analysis shows that although the considered AI task processing process is relatively comprehensive, the EHAHA proposed in this paper is better than the centralized processing benchmark method, bidding algorithm, and greedy algorithm, and can provide agile, stable, and efficient intelligent services.