星地异构认知无线电智能频谱决策与能效优先资源分配技术研究

The available spectrum resource of satellite communication is decreasing while the competition of spectrum resource between satellite communication network and terrestrial wireless mobile network is increasing. The shortage of available spectrum resource has been a bottleneck factor which restricts the development of satellite communications. Satellite cognitive radio has been proposed by space communication researchers as one of the most effective methods to solve the problems of spectrum scarcity and frequency interference. Based on the development demand of satellite-terrestrial heterogeneous converged network by space-ground integrated networking according to the 13th five-year plan for national scientific and technological innovation, in order to improve the utilization ratio of space spectrum resources, this project focuses on the key technologies of cognitive radio applied in satellite-terrestrial heterogeneous network. In allusion to the problems as hard spectrum detection, poor timeliness of spectrum decision and low efficiency of resource allocation faced by traditional cognitive radio technologies when applied in satellite-terrestrial heterogeneous cognitive communication network, the technologies of intelligent spectrum decision and energy efficient resource allocation are explored systematically in this project. Firstly, in order to solve the high resource overhead of spectrum data collection caused by multi-band and wideband spectrum environment of satellite networks, the wideband spectrum data collection technology based on compressed sensing is researched; then the mechanisms of intelligent spectrum data mining based on deep learning and spectrum state prediction based on chaos time series are respectively proposed by the introduction of machine learning theory, with the consideration of the influence of long time delay and heavy channel fading of satellite links on the availability of spectrum sensing results. And on this basis, in allusion to the characteristics as limited resources of satellite nodes and dynamic variation of space heterogeneous spectrum environment, the opportunistic spectrum access and robust power allocation algorithms for satellite-terrestrial hybrid networks based on the joint optimization of energy efficient and spectrum efficient are designed referred to green cognitive radio concept. Finally, the simulation experiment platform is established to test and verify the performances of proposed algorithms. The anticipated research fruits of this project will help promote the perfection and development of space cognitive communication theory and technologies for our country.

卫星认知无线电是解决空间频谱资源竞争和干扰日益加剧问题的有效途径。本项目立足于我国“十三五”天地一体化信息网络科技创新规划对星地异构网络融合的发展需求，探索星地异构认知无线电智能频谱决策与能效优先资源分配技术。针对卫星网络多频段宽带频谱应用环境导致频谱数据采集资源开销过大的问题，提出基于压缩感知的低采样宽带频谱数据获取技术；进而考虑卫星链路长时延、大衰落等特点对频谱判决结果时效性的影响，构建基于机器学习理论的新型频谱检测与决策模型，分别设计基于深度学习的频谱数据特征挖掘和基于混沌时间序列的频谱状态预测机制；在此基础上，针对卫星资源受限和空间异构频谱环境动态变化等特点，聚焦绿色认知无线电理念，深入分析星地异构认知无线电干扰机理，提出能效与谱效联合优化的星地异构动态频谱接入和低复杂度高效鲁棒功率分配算法。预期成果有助于促进我国空间认知通信理论与技术的完善和发展。

本项目以卫星认知无线电应用研究为牵引，开展星地异构认知无线电智能频谱决策与能效优先资源分配技术研究,主要研究工作和取得的研究成果包括：.1、开展了盲稀疏度下基于压缩感知的空间宽带频谱数据获取技术研究。一是针对卫星认知通信宽带频谱和平台资源受限的特点，提出了基于信号重构模型参数分析的宽带频谱非稀疏保护策略；二是针对宽带频谱信息的动态变化和噪声干扰严重的特点，提出了盲稀疏度下基于粒子滤波的稀疏信号重构算法。.2、开展了基于机器学习的卫星频谱数据特征挖掘与状态预测机制研究。一是针对卫星认知网络环境中待检测信号强度不断变化的特点，提出了基于信道历史状态信息的频谱状态判决策略；二是针对基于卫星认知通信中传统频谱感知算法感知性能低、受时延影响大的问题，提出了基于LSTM神经网络的多门限值优化感知预测算法；三是针对星地认知通信多授权用户频谱接入问题，提出了基于图论的动态频谱划分算法。.3、开展了基于能量效率与频谱效率联合优化的星地异构动态频谱接入策略研究。一是针对基于星地认知网络频谱共享问题，提出了基于博弈论的动态频谱接入算法；二是为了降低卫星链路时延对频谱接入的影响，提出了基于动态频谱分配的随机频谱接入策略。.4、开展了感知不确定下基于能效优先的星地异构认知无线电鲁棒功率分配算法研究。一是针对Underlay模式下卫星网络覆盖范围大导致用频冲突干扰问题，分别提出了能效谱效联合博弈功率控制算法和能效最大化的信道分配与功率优化算法；二是针对信道感知误差导致功率控制算法性能下降的问题，提出了基于位置信息误差的能效优先功率控制算法；三是针对链路传输时延长造成的基于实时信道感知的认知用户中断概率高的问题，提出了基于中断概率的鲁棒功率能效优化算法；四是针对现有星地认知网络采用正交多址接入技术限制了大规模用户接入和频谱利用率提升问题，提出了基于Q-Learning的星地认知网络NOMA功率分配算法。