项目名称: 基于受约束部分可观马氏决策模型的稀疏宽带信号感知机制的建模和控制研究
项目编号: No.61503358
项目类型: 青年科学基金项目
立项/批准年度: 2016
项目学科: 自动化技术、计算机技术
项目作者: 姜晓枫
作者单位: 中国科学技术大学
项目金额: 19万元
中文摘要: 时域频谱感知问题是当前认知无线通信网络研究的热点问题。该问题从时间片角度发现当今无线频谱资源使用过程中存在大量空闲时间片,为了提高频谱利用率,很多无线设备都试图利用这些空闲时间片上的频谱资源。然而,基于时间片结构的频段状态特性变化极快,无线设备只能获得少量频谱认知信息,基于获得的认知信息很难实时准确估计当前频段状态,经常会访问被占用的时间片,发生严重的通信干扰。本项目采用受约束的部分可观Markov决策过程控制理论对宽带频谱感知问题进行建模,通过非线性规划方法将多目标受约束的部分可观Markov决策优化问题等价转化成单目标极小极大优化问题,并建立基于不完美状态信息的灵敏度分析方法求解其最优控制策略。结合宽带频谱系统可分解的特点,将原高维优化问题分解成多个相互关联耦合的低维问题,实现时域宽带频谱感知问题的低复杂度优化控制方法。
中文关键词: 部分可观马尔科夫决策过程;约束马氏决策过程;性能势方法;基于灵敏度的优化
英文摘要: The spectrum sensing in the time domain is an important problem in the research on wireless communication networks. This problem discovers many idle time slots in the spectrum utilization procedure from the view of time slot. Many wireless devices try to utilize these idle time slots to increase the spectrum utilization rate. However, the spectrum state characteristics based on the time slot structure keep changing fast. The wireless device can only observe a little information about the spectrum, and access the occupied time slots. These events may result in serious radio interference. This project models the wideband spectrum sensing problem based on the theory of constrained partially observable Markov decision processes, and uses the nonlinear programming method to transform the multi-objective constrained partially observable Markov decision optimization problem into the single-objective minimax optimization problem, and develops the sensitivity analysis method to seek the optimal control policy. Consider that the wideband spectrum can be divided, the original high dimensional problem can be divided into multiple interconnected low dimensional problems, and the low-complexity optimization approach of the wideband spectrum sensing in the time domain can be developed.
英文关键词: Partially observable Markov decision process;Constrained Markov decision process;Performance potential method;Optimization based on sensitivity analysis