聚类分析是数据挖掘中经常用到的一种分析数据之间关系的方法.它把数据对象集合划分成多个不同的组或簇,每个簇内的数据对象之间的相似性要高于与其他簇内的对象的相似性.密度中心聚类算法是一个最近发表在《Science》上的新型聚类算法...聚类分析是数据挖掘中经常用到的一种分析数据之间关系的方法.它把数据对象集合划分成多个不同的组或簇,每个簇内的数据对象之间的相似性要高于与其他簇内的对象的相似性.密度中心聚类算法是一个最近发表在《Science》上的新型聚类算法,它通过评估每个数据对象的2个属性值(密度值ρ和斥群值δ)来进行聚类.相对于其他传统聚类算法,它的优越性体现在交互性、无迭代性、无数据分布依赖性等方面.但是密度中心聚类算法在计算每个数据对象的密度值和斥群值时,需要O(N^2)复杂度的距离计算,当处理海量高维数据时,该算法的效率会受到很大的影响.为了提高该算法的效率和扩展性,提出一种高效的分布式密度中心聚类算法EDDPC(efficient distributed density peaks clustering),它利用Voronoi分割与合理的数据复制及过滤,避免了大量无用的距离计算开销和数据传输开销.实验结果显示:与简单的MapReduce分布式实现比较,EDDPC可以达到40倍左右的性能提升.展开更多
This paper deals with the formation control problem of multiple unmanned aerial vehicles(UAVs) with collision avoidance. A distributed formation control and collision avoidance method is proposed based on Voronoi part...This paper deals with the formation control problem of multiple unmanned aerial vehicles(UAVs) with collision avoidance. A distributed formation control and collision avoidance method is proposed based on Voronoi partition and conventional artificial potential field. The collision avoidance is achieved by partitioning the whole space into non-overlapping regions based on Voronoi partition theory, which is taken as the task region to confine the movement of each UAV. The general motion control law is designed based on the conventional artificial potential field. As this often leads to local optimum when two UAVs are going to collide with each other and they may stay still where the repulsive force is adversely equivalent to the attractive force. To address this problem,the destination switch scheme is further proposed to let UAVs switch destinations when they reach the local equilibrium. Finally,the effectiveness of proposed formation control algorithm is validated by simulations and experiments.展开更多
Voronoi diagram is founded by using computational geometry based on originaldistribution of the waypoints, and then the elements from Voronoi diagram are metamorphosed by usingthe rule for airsppce partition, and the ...Voronoi diagram is founded by using computational geometry based on originaldistribution of the waypoints, and then the elements from Voronoi diagram are metamorphosed by usingthe rule for airsppce partition, and the controller's workload is accounted in each element that ismade up of Metamorphic Voronoi polygon. Then in accordance with the rule about balance ofcontroller's workload, Simulated Annealing algorithm (SA) is used to achieve the optimization ofcombination of those elements , and the new resolution has satisfied the restriction of two rulesfor airspace partition. Therefore, the boundaries of the aggregates of these elements are theoptimal borderlines of sectors. The result of actual airspace design example validates therationality of the sector optimization method presented in this paper.展开更多
针对移动传感器网络(Mobile sensor networks,MSNs)中动态目标(事件源)的监测优化问题,为提高网络覆盖质量,建立基于Voronoi剖分的监测性能(Quality of monitoring,QoM)评价函数,提出基于群集控制的传感器节点部署分布式控制算法.每个...针对移动传感器网络(Mobile sensor networks,MSNs)中动态目标(事件源)的监测优化问题,为提高网络覆盖质量,建立基于Voronoi剖分的监测性能(Quality of monitoring,QoM)评价函数,提出基于群集控制的传感器节点部署分布式控制算法.每个节点在本地结合最小二乘法和一致性算法来估计目标相对位置.相比传统算法,本文算法只需本地和单跳通信(可观测)邻居的信息,从而减小通信时长和能耗.算法在提高以目标为中心的一定区域监测性能的同时,使全体传感器速度趋于一致,从而在尽量保持网络拓扑结构的同时减少了整体移动能耗.在目标匀速或目标加速度信息全网可知的情况下,全体传感器速度渐近收敛至目标速度,且监测性能收敛至局部最优.所采用的目标位置估计滤波算法计算简单、切实可行.展开更多
A new method for sector optimum partition of airspace is proposed by dividing the fright altitude into several layers according to the distribution characteristics of the controller＇s workloads in an airspace. On the...A new method for sector optimum partition of airspace is proposed by dividing the fright altitude into several layers according to the distribution characteristics of the controller＇s workloads in an airspace. On the basis of the original distribution of the waypoints at each level of altitude, the sweel5 line algorithm of Voronoi diagram is used to divide them into certain polygons （ elements）, and the controller＇s workloads are calculated in each Voronoi polygon. Then by the rule about balance of controller＇s workload and by adding conditions of control handover or coordination for the sector, a mathematical model for the controller＇s workload based sector optimization is built. By the model, the Voronoi polygons are optimally partitioned. As a result, a 3D sector optimum partition of the whole airspace is formed by combining the sector optimum partitions at every layer. The actual airspace partition for Xiamen Airport has proved the reasonability and effectiveness of the 3D sector optimum partition of airspace proposed.展开更多
Let E be a Moran set on R1 associated with a bounded closed interval J and two sequences(nk)k =1 ^∞ and(Ck =(ck,j)j=1)^nk)k≥1. Let μ be the Moran measure on E associated with a sequence(Pk)k≥1 of positive probabil...Let E be a Moran set on R1 associated with a bounded closed interval J and two sequences(nk)k =1 ^∞ and(Ck =(ck,j)j=1)^nk)k≥1. Let μ be the Moran measure on E associated with a sequence(Pk)k≥1 of positive probability vectors with Pk =(pk,j)j =1,^nk, k ≥ 1. We assume that k≥1 1≤j≤nk^inf min Ck,j>0,k≥1 1≤j≤nk^inf minCk,j>0,k≥1 1≤j≤nk ^inf min pk,j>0. For every n ≥ 1, let αn be αn n optimal set in the quantization for μ of order r ∈(0,∞) and{Pa(αn)}a∈α∈an an arbitrary Voronoi partition with respect to αn. We write Iα(α,μ):=∫Pα(αn)^d(x,αn)^τ dμ(x),α∈αn;J(αn,μ):=α∈αn^minⅠα(α,μ),J(αn,μ):=α∈αn^max Ⅰα(α,μ).We show that J(αn,μ),J(αn,μ) and en^r,r(μ)-en^r +1,r(μ) are of the same order as 1/n en^r ,r(μ), where en^r ,r(μ):=∫d(x,αn)^r dμ(x) is the nth quantization error for μ of order r. In particular, for the class of Moran measures on R1, our result shows that a weaker version of Gersho’s conjecture holds.展开更多
文摘聚类分析是数据挖掘中经常用到的一种分析数据之间关系的方法.它把数据对象集合划分成多个不同的组或簇,每个簇内的数据对象之间的相似性要高于与其他簇内的对象的相似性.密度中心聚类算法是一个最近发表在《Science》上的新型聚类算法,它通过评估每个数据对象的2个属性值(密度值ρ和斥群值δ)来进行聚类.相对于其他传统聚类算法,它的优越性体现在交互性、无迭代性、无数据分布依赖性等方面.但是密度中心聚类算法在计算每个数据对象的密度值和斥群值时,需要O(N^2)复杂度的距离计算,当处理海量高维数据时,该算法的效率会受到很大的影响.为了提高该算法的效率和扩展性,提出一种高效的分布式密度中心聚类算法EDDPC(efficient distributed density peaks clustering),它利用Voronoi分割与合理的数据复制及过滤,避免了大量无用的距离计算开销和数据传输开销.实验结果显示:与简单的MapReduce分布式实现比较,EDDPC可以达到40倍左右的性能提升.
基金supported by the National Natural Science Foundation of China(Grant Nos.61603303 and 61473230)the Natural Science Foundation of Shaanxi Province(Grant Nos.2017JM6027 and 2017JQ6005)+2 种基金the China Postdoctoral Science Foundation(Grant No.2017M610650)the Innovation Development Foundation of Aisheng(Grant No.ASN-IF2015-1502)the Fundamental Research Funds for the Central Universities(Grant No.3102017JG02011)
文摘This paper deals with the formation control problem of multiple unmanned aerial vehicles(UAVs) with collision avoidance. A distributed formation control and collision avoidance method is proposed based on Voronoi partition and conventional artificial potential field. The collision avoidance is achieved by partitioning the whole space into non-overlapping regions based on Voronoi partition theory, which is taken as the task region to confine the movement of each UAV. The general motion control law is designed based on the conventional artificial potential field. As this often leads to local optimum when two UAVs are going to collide with each other and they may stay still where the repulsive force is adversely equivalent to the attractive force. To address this problem,the destination switch scheme is further proposed to let UAVs switch destinations when they reach the local equilibrium. Finally,the effectiveness of proposed formation control algorithm is validated by simulations and experiments.
文摘Voronoi diagram is founded by using computational geometry based on originaldistribution of the waypoints, and then the elements from Voronoi diagram are metamorphosed by usingthe rule for airsppce partition, and the controller's workload is accounted in each element that ismade up of Metamorphic Voronoi polygon. Then in accordance with the rule about balance ofcontroller's workload, Simulated Annealing algorithm (SA) is used to achieve the optimization ofcombination of those elements , and the new resolution has satisfied the restriction of two rulesfor airspace partition. Therefore, the boundaries of the aggregates of these elements are theoptimal borderlines of sectors. The result of actual airspace design example validates therationality of the sector optimization method presented in this paper.
文摘针对移动传感器网络(Mobile sensor networks,MSNs)中动态目标(事件源)的监测优化问题,为提高网络覆盖质量,建立基于Voronoi剖分的监测性能(Quality of monitoring,QoM)评价函数,提出基于群集控制的传感器节点部署分布式控制算法.每个节点在本地结合最小二乘法和一致性算法来估计目标相对位置.相比传统算法,本文算法只需本地和单跳通信(可观测)邻居的信息,从而减小通信时长和能耗.算法在提高以目标为中心的一定区域监测性能的同时,使全体传感器速度趋于一致,从而在尽量保持网络拓扑结构的同时减少了整体移动能耗.在目标匀速或目标加速度信息全网可知的情况下,全体传感器速度渐近收敛至目标速度,且监测性能收敛至局部最优.所采用的目标位置估计滤波算法计算简单、切实可行.
基金The National Natural Science Foundation of China (No.60472117)
文摘A new method for sector optimum partition of airspace is proposed by dividing the fright altitude into several layers according to the distribution characteristics of the controller＇s workloads in an airspace. On the basis of the original distribution of the waypoints at each level of altitude, the sweel5 line algorithm of Voronoi diagram is used to divide them into certain polygons （ elements）, and the controller＇s workloads are calculated in each Voronoi polygon. Then by the rule about balance of controller＇s workload and by adding conditions of control handover or coordination for the sector, a mathematical model for the controller＇s workload based sector optimization is built. By the model, the Voronoi polygons are optimally partitioned. As a result, a 3D sector optimum partition of the whole airspace is formed by combining the sector optimum partitions at every layer. The actual airspace partition for Xiamen Airport has proved the reasonability and effectiveness of the 3D sector optimum partition of airspace proposed.
基金supported by National Natural Science Foundation of China(Grant No.11571144)
文摘Let E be a Moran set on R1 associated with a bounded closed interval J and two sequences(nk)k =1 ^∞ and(Ck =(ck,j)j=1)^nk)k≥1. Let μ be the Moran measure on E associated with a sequence(Pk)k≥1 of positive probability vectors with Pk =(pk,j)j =1,^nk, k ≥ 1. We assume that k≥1 1≤j≤nk^inf min Ck,j>0,k≥1 1≤j≤nk^inf minCk,j>0,k≥1 1≤j≤nk ^inf min pk,j>0. For every n ≥ 1, let αn be αn n optimal set in the quantization for μ of order r ∈(0,∞) and{Pa(αn)}a∈α∈an an arbitrary Voronoi partition with respect to αn. We write Iα(α,μ):=∫Pα(αn)^d(x,αn)^τ dμ(x),α∈αn;J(αn,μ):=α∈αn^minⅠα(α,μ),J(αn,μ):=α∈αn^max Ⅰα(α,μ).We show that J(αn,μ),J(αn,μ) and en^r,r(μ)-en^r +1,r(μ) are of the same order as 1/n en^r ,r(μ), where en^r ,r(μ):=∫d(x,αn)^r dμ(x) is the nth quantization error for μ of order r. In particular, for the class of Moran measures on R1, our result shows that a weaker version of Gersho’s conjecture holds.
