Optimizing Resource Allocation and Request Routing for AI-Generated Content (AIGC) Services in Mobile Edge Networks With Cell Coupling
Tao Deng, Dongyu Chen, Juncheng Jia, Mianxiong Dong, Kaoru Ota, Zhanwei Yu, Di Yuan · IEEE Transactions on Vehicular Technology · 2024
In this paper, we investigate the deployment and service of pre-trained foundation models (PFMs) in mobile edge networks with cell coupling. We formulate a joint resource allocation and request routing optimization problem (RARP) to achieve a trade-off between the accuracy loss and cost of artificial intelligence-generated content (AIGC). For problem solving, we propose an alternating optimization algorithm (AOA) that decomposes RARP into two sub-problems and iteratively optimizes them. Specifically, for the first sub-problem, we reformulate it as a linear programming problem and use the off-the-shelf optimization solver to solve it. For the other sub-problem, we propose a deep reinforcement learning based algorithm to optimize the deployment to PFMs. Performance evaluations validate the efficiency of AOA.