CR-NOMA中基于深度确定策略梯度的能效优化策略

Deep deterministic policy gradient-based energy efficiency optimization algorithm for CR-NOMA

利用认知无线电非正交多址接入(cognitive radio non-orthogonal multiple access,CR-NOMA)技术可缓解频谱资源短缺问题,提升传感设备的吞吐量。传感设备的能效问题一直制约着传感设备的应用。为此,针对CR-NOMA中的传感设备,提出基于深度确定策略梯度的能效优化(deep deterministic policy gradientbased energy efficiency optimization,DPEE)算法。DPEE算法通过联合优化传感设备的传输功率和时隙分裂系数,提升传感设备的能效。将能效优化问题建模成马尔可夫决策过程,再利用深度确定策略梯度法求解。最后,通过仿真分析了电路功耗、时隙时长和主设备数对传感能效的影响。仿真结果表明,能效随传感设备电路功耗的增加而下降。此外,相比于基准算法,提出的DPEE算法提升了能效。

Cognitive radio non-orthogonal multiple access(CR-NOMA)technology was used to alleviate the shortage of spectrum resource,and improve the throughput of sensor devices.But the energy efficiency problem had been restricting the application of sensor devices.Therefore,for CR-NOMA,deep deterministic policy gradient-based energy efficiency optimization(DPEE)algorithm was proposed.By jointly optimizing the transmission power and time slot splitting coefficient,the energy efficiency of sensor devices was improved.The energy efficiency optimization problem was modeled as a Markov decision process,and it was solved by the deep deterministic policy gradient(DDPG)method.Finally,the influence of circuit power consumption,time slot durations and number of main devices on energy efficiency were analyzed.The simulation results show that the energy efficiency decreases as the circuit power consumption of sensor device increases.In addition,compared with other algorithms,the proposed algorithm improves energy efficiency.