TY - JOUR TI - Graph attention network-based multi-agent reinforcement learning for slicing resource management in dense cellular network AU - Shao, Yan AU - Li, Rongpeng AU - Hu, Bing AU - Wu, Yingxiao AU - Zhao, Zhifeng AU - Zhang, Honggang T2 - IEEE Transactions on Vehicular Technology AB - Network slicing (NS) management devotes to provi- sioning various services with distinct requirements over the same physical communication infrastructure and allocating resources on demands. Considering a dense cellular network scenario that contains several NS over multiple base stations (BSs), it remains challenging to design a proper resource management strategy in real time, so as to cope with frequent BS handover and meet distinct service requirements. In this paper, we propose to formulate this challenge as a multi-agent reinforcement learning (MARL) problem and leverage graph attention network (GAT) to strengthen the spatial cooperation between agents. Furthermore, we incorporate GAT into deep reinforcement learning (DRL) and correspondingly design an intelligent resource management strategy for NS. More specially, we have also testified the universal effectiveness of GAT for advancing DRL in multi-agent system, by applying GAT on the top of deep Q-network for the value-based method and advantage actor-critic for the policy- based method. Finally, we verify the superiority of the GAT-based DRL algorithms through extensive simulations. DA - 2021/10// PY - 2021 VL - 70 IS - 10 SP - 10792 EP - 10803 J2 - IEEE Trans. Veh. Tech. UR - https://www.rongpeng.info/files/Paper_2021TVT.pdf ER -