【摘要】 由于具有大容量、高效灵活的多地址访问、高频谱效率、动态带宽分配等优点,正交频分复用接入无源光网络(OFDMA-PON)成为了下一代光接入网络的最有潜力的选择之一.在OFDMAPON中,不同的光网络单元(ONU)可以共享子载波资源来支持网络资源管理和有效带宽分配.在上行传输中,多个ONU可以在整个传输周期内的不同时隙(TS)内共享正交低比特率子载波(SC)来传输上行数据.本文提出了一种基于深度强化学习(DRL)的动态子载波分配(DSA)策略.该策略以动态灵活的方式联合分配OFDMA-PON中时隙、子载波和调制格式等三维资源,通过采用合适的调制格式,同时优化业务延迟和ONU发射功率.将本文提出的基于DRL的DSA算法与传统的二维DSA算法进行仿真比较,结果表明,本文提出的DSA算法不仅大大降低了业务延迟,还可以节省ONU发射功率.更多还原

【Abstract】 Due to the advantages of large capacity,flexible multiple address access,high spectrum efficiency and dynamic bandwidth allocation,orthogonal frequency division multiplexing access passive optical network(OFDMA-PON)has become one of the most potential choices for the next generation optical access network.In OFDMA-PON,it allows different optical network units(ONUs)to share subcarriers(SCs)to support network resource management and effective bandwidth allocation.In uplink transmission,multiple ONUs can share orthogonal low bit rate SCs to transmit data in different time slots(TSs)during the entire transmission cycle.In this paper,a dynamic subcarrier allocation(DSA)strategy based on deep reinforcement learning(DRL)is proposed.The strategy jointly allocates time slots,subcarriers and modulation formats in a dynamic and flexible manner.By using the optimal modulation format,the delay service quality is ensured and the ONU transmit power can be reduced.The DSA algorithm using DRL is compared with the traditional two-dimensional DSA algorithm.The simulation results show that the proposed DSA algorithm using DRL reaches lower traffic latency with energy saving.更多还原