Online Resource Allocation for QoE Optimization in CoMP-assisted Embms System
Jian Chen,Kaili Zhai,Jie Jia,An Du,Yi Xue,Xingwei Wang
DOI: https://doi.org/10.1007/s12083-023-01555-0
IF: 3.488
2023-01-01
Peer-to-Peer Networking and Applications
Abstract:Coordinated multi-point (CoMP) transmission allows multiple base stations (BSs) to transmit the same signal to users simultaneously at the same resource block for higher throughput. It can be applied to the evolved Multimedia Broadcast/Multicast Service (eMBMS) framework to provide additional gains for users in the same multicast group. However, optimal resource allocation is required for the system with multiple cells and multicast groups with various quality of experience (QoE) requirements. This paper investigates the online QoE maximization for a CoMP-enhanced eMBMS system by jointly optimizing the BS clustering, sub-channel assignment, and power allocation. Moreover, we propose a hybrid learning framework combining deep reinforcement learning (DRL) and model-based optimization techniques. More specifically, for the DRL scheme, we propose a multi-agent deep Q network (DQN) to find the online BS clustering and sub-channel assignment for each multicast group with fast convergence. For the model-based scheme, we propose a fractional programming-based method to find the optimal power allocation and evaluate the joint action reward in the proposed DRL framework. With the reward obtained by optimal power allocation design, the action space for the DRL algorithm can thus be reduced, and the joint optimization algorithmic framework can be constructed. Numerical results demonstrate the effectiveness of the proposed algorithm, and the performance gain can be improved by increasing the number of BSs and sub-channels.