【作者】 冯贵年；

【导师】 樊平毅；

【作者基本信息】 清华大学 ， 信息与通信工程， 2010， 博士

【摘要】 无线Ad Hoc网络以其无中心、自组织、多跳路由的特点在众多无线网络中独树一帜,并且在军事和民用等领域得到了广泛的应用。然而,事物总是具有两面性。无线Ad Hoc网络研究中,面临着网络拓扑动态变化、节点能量有限、信道带宽有限等诸多挑战。面对这些限制,本文从改善网络的整体传输性能出发,选取两方面问题进行研究:其一,整体上如何构建干扰优化的网络拓扑;其二,局部上如何提高结构单元的传输性能。针对整体的研究,论文从网络干扰分析入手,提出了一种新的网络干扰量化模型,并基于此给出了干扰优化的拓扑控制算法。首先,我们从理论上描述了双向链路间发生干扰的充要条件,并据此证明了在双向拓扑网络中,在双链路干扰情况下,节点使用相等功率分配策略和最小功率分配策略,对网络干扰的影响是一致的结论。接着,引入了一个新的干扰量化指标——平均部分干扰系数。理论研究与仿真分析表明,平均部分干扰系数与网络节点密度无关,可以近似表示为链路长度的线性函数。基于前述分析,后续提出了两种干扰优化拓扑控制算法——分块干扰最小算法BIMA和分布干扰最小算法LIMA。仿真结果表明,算法BIMA和LIMA不仅具有优良的抑制干扰和降低能耗的性能,而且还具备较好的spanner特性、较低的节点度以及对网络节点移动的鲁棒性。针对局部的分析,考虑了基于三节点模型的中继网络作为研究的基本结构单元,采用渐近最优喷泉码下的协作通信技术提高其传输性能。首先,以端到端的恒定传输速率作为传输的前提,分析了二进制删除信道下,中继网络采用多种协作方式下的信道容量。结果表明,维持各信道尽可能繁忙将带来较好的中继网络特性。基于此,引入渐近最优喷泉码的概念,给出了基于中继流控制的贪婪解码转发协作方案。此后,考虑中继缓存不溢出的要求,我们利用排队理论给出了一种优化功率分配策略,并分别在加性高斯白噪声信道和瑞利衰落信道下给出了功率分配系数的数值解。仿真结果表明,该功率分配策略下的贪婪解码转发协作方案不仅满足中继缓存不溢出的要求,而且还改善了中继网络在可达速率和信噪比方面的性能。更多还原

【Abstract】 Wireless ad hoc network is a kind of new network architecture which has no center, no organizer, and multi-hop routing. Such characteristics make it distinguish from other wireless networks and have extensive applications in both military and civilian fields. However, every coin has a flip side. Due to variable network topology, finite node energy and limited frequency bandwidth, how to improve the whole transmission performance of wireless ad hoc networks is a complicated and open problem. Facing the problem, many studies are done internally and externally, and two fundamental questions are investigated in the dissertation: the first one is how to construct a network topology with minimum interference globally; the other one is how to increase the transmission performance in a structural unit locally.To external research, we first present a quantitative model of network interference starting from the analysis to network interference. We describe the pair-wise interference condition theoretically. By the condition, it has been proved that the interference condition is identical for two power allocation strategies: the equal-power allocation strategy and the minimum-power allocation strategy. Based on the analysis above, a new metric, the average partial interference co-efficient, is given. Theoretical research and simulation analysis indicate that the average partial interference coefficient has no relationship with node density in the network, and it can be approximated as a linear function of link distance. Then, we present two particular interference optimization topology control algorithms, BIMA (Blocked Interference Minimum Algorithm) and LIMA (Local Interference Minimum Algorithm). Simulation results show that both BIMA and LIMA can not only provide a good performance in controlling network interference and conserving node energy, but also can maintain good spanner property and keep low node degree. Besides, both BIMA and LIMA show a good robustness to node mobility.To internal requirement, the relay network based on three nodes model is given as the basic structural unit. And we adopt cooperative communication using fountain codes to improve the transmission performance in the basic structural unit. Firstly, under the precondition of end-to-end invariable transmission rate, a simplified model is analyzed. Over binary erasure channels, the capacities of the relay network under several cooperative modes are analyzed. The results show that it will bring better transmission performance than others in the relay network if the cooperative mode can keep the status of each channel in busy mode as far as possible. According to such property, we present a greedy decode-and-forward cooperative scheme based on the flow control at the relay node after inducing the concept of the asymptotically optimal fountain codes. In the scheme, two requirements are considered. One is to prevent the relay from accumulating too many blocks because of the limited buffer size; the other one is to keep all the three channels as busy as possible in order to improve the overall throughput. To the requirements, the optimal power allocation strategy is studied by queue theory. Simulation results indicate that the greedy decode-and-forward cooperative scheme under the optimal power allocation strategy can not only keep the relay buffer from overflowing, but also improve the performances of achievable rate and bitwise signal-noise-rate in the relay network based on three nodes model.更多还原