海杂波的奇异谱分析不仅能从理论上揭示海洋表面的动力学机理,同时也是对海探测雷达的关键技术之一.本文提出基于小波leaders的海杂波时变奇异谱分析方法,将时间信息引入海杂波的奇异谱分析之中,从而实现动态的解析描述海杂波随时间变化的奇异谱特性.在理论上,通过信号自身加窗,将时间信息引入传统的奇异谱(或称多重分形谱),实现了对海杂波时变奇异谱分布分析;在算法上,充分利用了小波leaders技术对于多种奇异性的提取能力(包括chirp奇异性和cusp奇异性),通过对时变奇异性指数和时变尺度函数的Legendre变换,实现对海杂波时变奇异谱分布的计算;在应用部分,采用经典的多重分形模型——随机小波序列(RWC)以及三级海态条件下连续波多普勒体制雷达海杂波进行仿真分析,实验结果表明:1)基于小波leaders的奇异谱分布能跟踪海杂波的时变尺度特性,有效展示其时变奇异性谱分布;2)算法具有较好的负矩特性和统计收敛性.该方法能为复杂非线性系统及随机多重分形信号分析提供参考.

Singularity spectrum analysis of sea clutter is the key technology of detecting radar for sea target, which can discover the dynamic mechanism of the sea surface theoretically. In this paper, based on wavelet leaders the time-varying singularity spectrum distribution of sea clutters is proposed, which introduces time information to the traditional singularity spectrum, and displays the time-varying characteristic of singularity spectrum analytically. In theory, by way of self-windowed fractal signal, we introduce the time information to the traditional singularity spectrum, and realize multifractal spectrum distribution of sea clutters. In algorithm, based on the wavelet leaders, we adapt the process of embodying chirp-type and cusp-type singularities, and obtain the time-varying singularity spectrum distribution of sea clutters by the Legendre transform of the time-varying scaling function. In practice, we analyze the classical multifractal model-random wavelet series and the real sea

研究了leader 有控制输入且followers 未知该输入条件下的线性多智能体一致性跟踪问题。提出两种一致性跟踪算法，证明两种算法在leader 到followers 存在一棵有向生成树且follower 间拓扑是有向条件下，网络就能跟踪leader 的状态。对于第一种算法，节点根据相邻节点或leader 的状态来求解其控制输入，并基于代数Riccati 不等式给出连续情形下算法稳定性条件。第二种算法直接利用相邻节点或leader的状态，使followers 在上述网络条件下跟踪leader 的状态，同样基于代数Riccati 不等式给出算法稳定性条件。仿真结果验证了算法的有效性。

This paper considers consensus tracking of multi-agent systems with general linear dynamics and a leader whose control input is nonzero and unavailable to any follower. Two tracking algorithms are designed for each follower. We prove that the followers can track the leader using both algorithms if the interaction graph among them is directed and there ex-ists a directed path from the leader to each follower. For the first algorithm, each follower calculates the neighbors0 control inputs based on their states, then the stability condition is given based on the algebra Riccati inequality. The second algorithm utilizes neighbors0 states directly and the followers can track the leader with the aforementioned topology based on the algebra Riccati inequality. Simulations indicate the capabilities of the algorithms.

主要研究了 leader-following多智能体网络的分布式编队问题。假设多智能体网络中的每个 follo-wer不能实时得到leader的信息，针对每个 follower 设计相应的分布式观测器。考虑了无时延的切换拓扑多智能体网络和存在时变时延的固定网络两种情况。利用 Lyapunov 稳定定理，给出了系统稳定和控制协议有效的充分条件。最后，给出了数字仿真结果以验证所设计协议的有效性，即 followers 形成所要求的编队并且与 leader达到速度一致。

This paper investigates the distributed formation control of leader-following multi-agent network. Supposing each follower can not get the real-time information of the virtual leader,the distributed observer is designed for each follower.The paper considers both the condition of switching topologies without time delay and fixed topologies with time-varying communication delay.By using the Lyapunov stability theory, the sufficient conditions of system’ s stability and the designed protocols’ effectiveness are given.Finally, the numerical simulation results are presented to illustrate the effectiveness of the designed protocols, namely,the followers form the designed formation and reach velocity consensus with the virtual leader.

社交网络社区Leader选举，即识别社区内影响力最大用户，是社交网络结构分析重要任务之一，在识别意见领袖、增进社区融合等方面具有重要的理论和应用研究价值。传统的社区 Leader 选举技术如 UserRank和PeopleRank，主要基于社交网络链接分析实现，忽略了用户本身属性的相似度度量。因此得到的社区Leader不能有效保证其社区代表性。本文提出了一种新的用户关系建模方法，将传统的链接分析和用户属性相似度度量融合，有效识别具有代表性的高影响力用户。实验结果表明，提出的方法不仅可以选举出社区内部具有代表性的高影响力Leader用户，还可以通过社区Leader选举使得社区内其他用户的查询效率得到有效提高。

The election of community leader, which is used to select the high-impact user in the social community, is one of the important structural analysis based tasks in social network. It has important theoretical and practical value in both identifying opinion leaders and promoting community integration. However, most of the proposed approaches such as UserRank and PageRank, neg-lecting the important attributes similarity between different users, are implemented based on the linking analysis in social network. In this paper, a new user relationship model is proposed by efficiently combining the traditional linking analysis based on graph structure and similarity metric based on user attributes. Experimental results demonstrate that the high-impact user in the social community can be selected and in turn be used to improve other users'' query efficiency.

针对分布式应用中资源共享的一致性问题,分析了Paxos算法中3种角色的特点,提出了一种基于该算法的外部分布式锁服务系统,该系统定义了8种锁原语基本操作,利用Fast-Leader-Election算法选举Leader节点,设计了数据初始化、数据写同步算法。实验结果表明该系统具有分布式锁服务的容错性,在高并发下具有高可用性及数据一致性。

Aiming at the consensus problem of resources sharing in distributed applications,this paper analyzes the characteristics of three kinds of roles in Paxos algorithm,and proposes an external distributed lock service system based on this algorithm. This system defines eight kinds of locking primitive basic op-erations,elects Leader nodes by Fast - Leader - Election algorithm,and designs data initialization,data- write synchronization algorithm. Test results show that this proposed system has fault tolerance of dis-tributed lock service,and has high availability and data consistency under high concurrency.