WO2020082768A1 - 一种用于无人设备集群控制的多体队形重构方法 - Google Patents

一种用于无人设备集群控制的多体队形重构方法 Download PDF

Info

Publication number
WO2020082768A1
WO2020082768A1 PCT/CN2019/092627 CN2019092627W WO2020082768A1 WO 2020082768 A1 WO2020082768 A1 WO 2020082768A1 CN 2019092627 W CN2019092627 W CN 2019092627W WO 2020082768 A1 WO2020082768 A1 WO 2020082768A1
Authority
WO
WIPO (PCT)
Prior art keywords
formation
device cluster
unmanned device
target
probability density
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Ceased
Application number
PCT/CN2019/092627
Other languages
English (en)
French (fr)
Inventor
魏小辉
廖玮
李旭波
高天驰
李龙
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Nanjing University of Aeronautics and Astronautics
Original Assignee
Nanjing University of Aeronautics and Astronautics
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Nanjing University of Aeronautics and Astronautics filed Critical Nanjing University of Aeronautics and Astronautics
Priority to US17/287,527 priority Critical patent/US11977396B2/en
Publication of WO2020082768A1 publication Critical patent/WO2020082768A1/zh
Anticipated expiration legal-status Critical
Ceased legal-status Critical Current

Links

Images

Classifications

    • GPHYSICS
    • G05CONTROLLING; REGULATING
    • G05DSYSTEMS FOR CONTROLLING OR REGULATING NON-ELECTRIC VARIABLES
    • G05D1/00Control of position, course, altitude or attitude of land, water, air or space vehicles, e.g. using automatic pilots
    • G05D1/10Simultaneous control of position or course in three dimensions
    • G05D1/101Simultaneous control of position or course in three dimensions specially adapted for aircraft
    • G05D1/104Simultaneous control of position or course in three dimensions specially adapted for aircraft involving a plurality of aircrafts, e.g. formation flying
    • GPHYSICS
    • G06COMPUTING OR CALCULATING; COUNTING
    • G06FELECTRIC DIGITAL DATA PROCESSING
    • G06F30/00Computer-aided design [CAD]
    • G06F30/20Design optimisation, verification or simulation
    • GPHYSICS
    • G05CONTROLLING; REGULATING
    • G05DSYSTEMS FOR CONTROLLING OR REGULATING NON-ELECTRIC VARIABLES
    • G05D1/00Control of position, course, altitude or attitude of land, water, air or space vehicles, e.g. using automatic pilots
    • G05D1/60Intended control result
    • G05D1/69Coordinated control of the position or course of two or more vehicles
    • G05D1/695Coordinated control of the position or course of two or more vehicles for maintaining a fixed relative position of the vehicles, e.g. for convoy travelling or formation flight
    • GPHYSICS
    • G06COMPUTING OR CALCULATING; COUNTING
    • G06FELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00Pattern recognition
    • G06F18/20Analysing
    • G06F18/23Clustering techniques
    • G06F18/232Non-hierarchical techniques
    • G06F18/2321Non-hierarchical techniques using statistics or function optimisation, e.g. modelling of probability density functions
    • GPHYSICS
    • G06COMPUTING OR CALCULATING; COUNTING
    • G06NCOMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N20/00Machine learning
    • GPHYSICS
    • G06COMPUTING OR CALCULATING; COUNTING
    • G06NCOMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N7/00Computing arrangements based on specific mathematical models
    • G06N7/01Probabilistic graphical models, e.g. probabilistic networks
    • YGENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
    • Y02TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
    • Y02PCLIMATE CHANGE MITIGATION TECHNOLOGIES IN THE PRODUCTION OR PROCESSING OF GOODS
    • Y02P90/00Enabling technologies with a potential contribution to greenhouse gas [GHG] emissions mitigation
    • Y02P90/02Total factory control, e.g. smart factories, flexible manufacturing systems [FMS] or integrated manufacturing systems [IMS]

Definitions

  • the invention relates to the technical field of unmanned equipment cluster control, and in particular to a multi-body formation reconstruction method for unmanned equipment cluster control.
  • unmanned equipment cluster technology has been tried to be applied in various fields, such as unmanned aerial vehicle clusters, unmanned vehicle clusters have become more and more widely used, especially in performance, transportation, reconnaissance and other tasks Use less.
  • unmanned vehicle clusters have become more and more widely used, especially in performance, transportation, reconnaissance and other tasks Use less.
  • arranging the individuals in the cluster into a specific team is one of the key issues.
  • the problem of formation reconstruction also becomes more and more complicated, mainly reflected in path planning and collision prevention.
  • Formation reconstruction can generally be divided into three sub-tasks: first, control the formation of the controlled object to change to the target formation; second, keep the formation stable and be able to recover itself under disturbance; third, during the movement avoid collision.
  • most of the current formation reconstruction algorithms are based on the Leader-Follower mode. It is necessary to determine a leader among all the controlled objects, and other individuals as followers follow the leader in some way. Some algorithms are improved based on the leader-follower model. For example, the formation structure is regarded as the model of the Henneberg structure. These algorithms need to map the position in the formation to the controlled object in advance, and then plan the path. In addition, it is necessary to set up unmanned devices that act as followers and leaders. If the current formation is disrupted due to external interference during the task execution process, the path needs to be re-planned.
  • Embodiments of the present invention provide a multi-body formation reconstruction method for unmanned device cluster control, which can mitigate the problem of the current leader-follower mode control algorithm that is difficult to cope with disturbed working conditions.
  • Step 1 Read a preset target formation, the target formation is a set of coordinates of a specified number of points, the specified number matches the number of unmanned devices in the unmanned device cluster;
  • Step 2 Establish a probability density function P (x) associated with the target formation
  • Step 3 Obtain the current formation of all controlled objects in the unmanned device cluster, and establish a probability density function Q (x) associated with the current formation;
  • Step 4 Reduce the cross entropy of P (x) and Q (x) by adjusting the position of the controlled object in the unmanned device cluster;
  • Step 3 and Step 4 Repeat Step 3 and Step 4 until the formation of the controlled objects in the unmanned device cluster is consistent with the target formation.
  • the multi-body formation reconstruction algorithm achieveds real-time calculation of flight and adjusts the flight formation of the cluster, without the need to establish a one-to-one correspondence between the positions in the formation and the controlled objects in advance, and without the need for prior path Planning, each controlled object does not have a leader and a follower, and each individual will not collide when moving according to the instructions generated by the algorithm, thereby mitigating the current leader-follower mode control algorithm, which is difficult to deal with interference.
  • FIG. 1 is a schematic diagram of a target formation provided by an embodiment of the present invention
  • FIG. 2 is a contour diagram of a probability density function related to a target formation provided by an embodiment of the present invention
  • FIG. 3 is a schematic diagram of the current formation of a controlled object provided by an embodiment of the present invention.
  • FIG. 4 is a contour diagram of a probability density function related to the current formation provided by an embodiment of the present invention.
  • FIG. 5 is a schematic diagram of a running track of a controlled object provided by an embodiment of the present invention.
  • FIG. 6 is a schematic flowchart of a method provided by an embodiment of the present invention.
  • An embodiment of the present invention provides a multi-body formation reconstruction method for unmanned device cluster control, as shown in FIG. 6, including:
  • Step 1 Read the preset target formation.
  • the target formation is a set composed of coordinates of a specified number of points, and the specified number matches the number of unmanned devices in the unmanned device cluster.
  • Step 2 Establish a probability density function P (x) associated with the target formation.
  • Step 3 Obtain the current formation of all controlled objects in the unmanned device cluster, and establish a probability density function Q (x) associated with the current formation.
  • Step 4 By adjusting the position of the controlled object in the unmanned device cluster, the cross entropy of P (x) and Q (x) is reduced.
  • Step 3 and Step 4 Repeat Step 3 and Step 4 until the formation of the controlled objects in the unmanned device cluster is consistent with the target formation.
  • the specific method for reading the preset target formation in step 1 includes:
  • the specific manner of establishing the probability density function P (x) associated with the target formation in step 2 includes:
  • the final probability density function P (x) can be expressed as:
  • the standard deviation is ⁇ , where x refers to the element in the set (x 1 , ..., x n ), that is, P i (x) is expressed as P i (x 1 , ..., x n ).
  • the specific manner of establishing the probability density function Q (x) associated with the current formation in step 3 includes:
  • the final Q (x) can be expressed as:
  • Q i (x) is a mean value of Yi to, [sigma] is the standard deviation of the normal probability density function of n, with n from variable x component, x j represents the j-th component of x. Represents the j-th component of the coordinate of the i-th controlled object position in the current formation.
  • step 4 reduces the cross entropy of P (x) and Q (x) by adjusting the position of the controlled object in the unmanned device cluster, including:
  • the gradient descent model is:
  • the ⁇ is attenuated. Since the algorithm used in this embodiment is similar to optimization, in order to avoid falling into a local optimal solution, the initial value of the parameter ⁇ should be large enough. As the task progresses, in order to speed up the convergence rate, ⁇ can be decayed with time.
  • Event-driven approach divide the space around the leader into several areas, and pass the partition of the follower through the area as an event to the controller, and then the controller generates control instructions based on these events so that the follower reaches and stays in the designated area .
  • the existing algorithm does not take into account the fact that for a cluster composed of the same individuals, any two individuals are interchanged, and their formation remains unchanged, thereby increasing the complexity of the calculation.
  • the path needs to be re-planned.
  • a UAV cluster composed of 20 controlled objects is restructured into an S-shaped queue from any initial position.
  • step 1 Perform step 1 to obtain the set of coordinates ⁇ X i ⁇ of all points in the S-shaped queue:
  • the calculation result of P (x) can refer to the contour plot of the probability density function related to the target formation as shown in FIG. 2.
  • step 3 Perform step 3 to obtain the current formation of all controlled objects represented by the set ⁇ Y i ⁇ , where And establish a probability density function Q (x) related to the current formation of all controlled objects, Q i (x) is a multivariate normal distribution with Y i as the mean and ⁇ as the standard deviation, namely:
  • the calculation result can refer to the contour map of the probability density function related to the current formation as shown in FIG. 4, and the current formation of the controlled object can refer to FIG. 3 at this time.
  • step 4 Perform step 4 to adjust the positions of all controlled objects so that the cross entropy of P and Q becomes smaller; the method of calculating the cross entropy is:
  • Monte Carlo method is adopted to calculate 2048 times in a random variable with P (x) as the probability density function.
  • the multi-body formation reconstruction algorithm provided in this embodiment achieves real-time calculation of flight and adjusts the flight formation of the cluster, without the need to establish a one-to-one correspondence between the positions in the formation and the controlled objects in advance, and without the need for prior path Planning, there is no leader and follower between the controlled objects, and each individual will not collide when moving according to the instructions generated by the algorithm.

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • Data Mining & Analysis (AREA)
  • General Engineering & Computer Science (AREA)
  • Evolutionary Computation (AREA)
  • Artificial Intelligence (AREA)
  • Software Systems (AREA)
  • Automation & Control Theory (AREA)
  • Aviation & Aerospace Engineering (AREA)
  • Radar, Positioning & Navigation (AREA)
  • Remote Sensing (AREA)
  • Probability & Statistics with Applications (AREA)
  • Mathematical Physics (AREA)
  • Computing Systems (AREA)
  • Mathematical Optimization (AREA)
  • Pure & Applied Mathematics (AREA)
  • Algebra (AREA)
  • Computational Mathematics (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Mathematical Analysis (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Evolutionary Biology (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Computer Hardware Design (AREA)
  • Geometry (AREA)
  • Medical Informatics (AREA)
  • Feedback Control In General (AREA)
  • Control Of Position, Course, Altitude, Or Attitude Of Moving Bodies (AREA)

Abstract

本发明实施例公开了一种用于无人设备集群控制的多体队形重构方法,涉及无人设备集群控制技术领域,能够缓减目前难以应对受到干扰的工况的问题。本发明包括:步骤1、读取预设的目标队形;步骤2、建立与所述目标队形关联的概率密度函数;步骤3、获取所述无人设备集群中的所有被控物体的当前队形,并建立与所述当前队形关联的概率密度函数;步骤4、通过调整所述无人设备集群中的被控物体的位置,减小交叉熵;重复执行步骤3和步骤4,直至所述无人设备集群中的被控物体组成的队形与所述目标队形一致。本发明适用于无人设备集群控制的多体队形重构。

Description

一种用于无人设备集群控制的多体队形重构方法 技术领域
本发明涉及无人设备集群控制技术领域,尤其涉及一种用于无人设备集群控制的多体队形重构方法。
背景技术
近年来,无人设备集群技术开始被尝试应用在各个领域,例如无人机集群,无人车集群的应用就越来越广泛,尤其是在表演,运输,侦察等任务中均有或多或少的运用。在这些任务中,将集群中的个体排列成特定的队形成为了关键问题之一。随着集群中个体数量的增加,队形重构问题也变得越来越复杂,主要体现在路径规划,防止碰撞等方面。
队形重构一般可以分为三个子任务:第一,控制被控物体的队形变换到目标队形;第二,保持队形稳定,在扰动下能够自行恢复;第三,在移动过程中避免碰撞。目前现有的队形重构算法大多基于Leader-Follower模式,需要在所有的被控物体中确定一个leader,其它的个体作为follower按照某种方式跟随leader。还有一些算法是根据leader-follower模式改进而来的,例如将队形结构视作Henneberg结构的模式。这些算法都需要事先将队形中的位置与被控物体一一对应,然后进行路径规划。并且需要设定好充当follower、leader的无人设备,在任务执行过程中如果遭遇外界干扰导致当前队形被扰乱则需要对路径重新规划。
但在实际应用中,一旦集群的规模太大则几乎难以实时得通过路径重新规划,来解决外界干扰导致当前队形被扰乱的问题,就比方说:于2018年5月1日在西安城墙永宁门举行的1374架无人机编队表演,在实际演出过程中出现了重 大失误,结果是集群中的无人机大量掉落,事故原因被初步认定为,由于干扰或者设备故障的,无人机无法接收到GPS信号的情况,从而导致作为leader的无人机失控。这就是由于现有的控制算法难以应对受到干扰的工况,导致了所有直接或者间接跟随的无人机出现了混乱。
发明内容
本发明的实施例提供一种用于无人设备集群控制的多体队形重构方法,能够缓减目前leader-follower模式的控制算法中,难以应对受到干扰的工况的问题。
为达到上述目的,本发明的实施例采用如下技术方案:
步骤1、读取预设的目标队形,所述目标队形是由指定数量的点的坐标构成的集合,所述指定数量匹配所述无人设备集群中的无人设备的数量;
步骤2、建立与所述目标队形关联的概率密度函数P(x);
步骤3、获取所述无人设备集群中的所有被控物体的当前队形,并建立与所述当前队形关联的概率密度函数Q(x);
步骤4、通过调整所述无人设备集群中的被控物体的位置,减小P(x)与Q(x)的交叉熵;
重复执行步骤3和步骤4,直至所述无人设备集群中的被控物体组成的队形与所述目标队形一致。
本实施例提供的多体队形重构的算法,做到了实时计算飞行并调整集群的飞行队形,无需事先将队形中的位置与被控物体建立一一对应的关系,无需事先进行路径规划,各被控物体不存在leader和follower之分,每个个体按照算法产生的指令移动时不会发生碰撞,从而也就缓减了目前leader-follower模式的控制算法中,难以应对受到干扰的工况的问题。
附图说明
为了更清楚地说明本发明实施例中的技术方案,下面将对实施例中所需要使用的附图作简单地介绍,显而易见地,下面描述中的附图仅仅是本发明的一些实施例,对于本领域普通技术人员来讲,在不付出创造性劳动的前提下,还可以根据这些附图获得其它的附图。
图1为本发明实施例提供的目标队形示意图;
图2为本发明实施例提供的与目标队形有关的概率密度函数的等高线图;
图3为本发明实施例提供的被控物体的当前队形示意图;
图4为本发明实施例提供的与当前队形有关的概率密度函数的等高线图;
图5为本发明实施例提供的被控物体运行轨迹示意图;
图6为本发明实施例提供的方法流程示意图。
具体实施方式
为使本领域技术人员更好地理解本发明的技术方案,下面结合附图和具体实施方式对本发明作进一步详细描述。下文中将详细描述本发明的实施方式,所述实施方式的示例在附图中示出,其中自始至终相同或类似的标号表示相同或类似的元件或具有相同或类似功能的元件。下面通过参考附图描述的实施方式是示例性的,仅用于解释本发明,而不能解释为对本发明的限制。本技术领域技术人员可以理解,除非特意声明,这里使用的单数形式“一”、“一个”、“所述”和“该”也可包括复数形式。应该进一步理解的是,本发明的说明书中使用的措辞“包括”是指存在所述特征、整数、步骤、操作、元件和/或组件,但是并不排除存在或添加一个或多个其他特征、整数、步骤、操作、元件、组件和/或它们的组。应该理解,当我们称元件被“连接”或“耦接”到另一元件时,它可以直接连接或耦接到其他元件,或者也可以存在中间元件。此外,这 里使用的“连接”或“耦接”可以包括无线连接或耦接。这里使用的措辞“和/或”包括一个或更多个相关联的列出项的任一单元和全部组合。本技术领域技术人员可以理解,除非另外定义,这里使用的所有术语(包括技术术语和科学术语)具有与本发明所属领域中的普通技术人员的一般理解相同的意义。还应该理解的是,诸如通用字典中定义的那些术语应该被理解为具有与现有技术的上下文中的意义一致的意义,并且除非像这里一样定义,不会用理想化或过于正式的含义来解释。
本发明实施例提供一种用于无人设备集群控制的多体队形重构方法,如图6所示,包括:
步骤1、读取预设的目标队形。
其中,所述目标队形是由指定数量的点的坐标构成的集合,所述指定数量匹配所述无人设备集群中的无人设备的数量。
步骤2、建立与所述目标队形关联的概率密度函数P(x)。
步骤3、获取所述无人设备集群中的所有被控物体的当前队形,并建立与所述当前队形关联的概率密度函数Q(x)。
步骤4、通过调整所述无人设备集群中的被控物体的位置,减小P(x)与Q(x)的交叉熵。
重复执行步骤3和步骤4,直至所述无人设备集群中的被控物体组成的队形与所述目标队形一致。
在本实施例中,步骤1中读取预设的目标队形的具体方式,包括:
获取所述目标队形中各个点的坐标的集合{X i}。
其中,{X i}中的元素表示所述目标队形中各个点的目标位置,i=1,2,...,N,所述目标队形中的点表示为
Figure PCTCN2019092627-appb-000001
表示第i个目标位置的坐标第j个分 量,j=1,...,n,n表示根据目标队形所在空间的维数,n取值为2或3。
在本实施例中,步骤2中建立与所述目标队形关联的概率密度函数P(x)的具体方式,包括:
确定以X i为均值的多元正态分布P i(x),其中:
Figure PCTCN2019092627-appb-000002
建立与所述目标队形关联的概率密度函数P(x),其中,
Figure PCTCN2019092627-appb-000003
x={x j},xj表示其第j个分量,j=1,…,n,N表示被控物体的数量,n表示根据目标队形所在空间的维数,n的取值可以为2或3,σ表示正态分布的标准差。
由于P i(x)是建立在n维空间里的一个函数,所以自变量x有n个分量,x j表示x的第j个分量,x就是一个形式参数,好比常用f(x)=x+1表示一个x的函数,此处的x就是一个形式参数,那么f(y)=y+1,不论形式参数换成什么字母,函数f所表示的意思是不变的。
最终的概率密度函数P(x)可以表示为:
Figure PCTCN2019092627-appb-000004
其中,P i(x)是一个多元正态分布的概率密度函数,这个多元正态分布的均值是
Figure PCTCN2019092627-appb-000005
标准差是σ,其中的x指代集合应(x 1,…,x n)中的元素,即P i(x)表示成P i(x 1,…,x n)。
在本实施例中,步骤3中建立与所述当前队形关联的概率密度函数Q(x)的具体方式,包括:
获取所述当前队形中的所有被控物体的坐标的集合{Y i},其中,
Figure PCTCN2019092627-appb-000006
表示所述当前队形中的第i个被控物体位置的坐标第j个分量。
确定以
Figure PCTCN2019092627-appb-000007
为均值的多元正态分布Q i(x),其中:
Figure PCTCN2019092627-appb-000008
建立与所述当前队形关联的概率密度函数Q(x),其中,
Figure PCTCN2019092627-appb-000009
x={x j},j=1,…,n,N表示被控物体的数量,σ表示正态分布的标准差。
最终的Q(x)可以表示为:
Figure PCTCN2019092627-appb-000010
其中,Q i(x)是一个以Yi为均值,σ为标准差的n元正态分布概率密度函数,自变量x有n个分量,x j表示x的第j个分量。
Figure PCTCN2019092627-appb-000011
表示所述当前队形中的第i个被控物体位置的坐标第j个分量。Q i(x)也是建立在n维空间里的一个函数,所以自变量x有n个分量,x j表示x的第j个分量,x就是一个形式参数,好比常用f(x)=x+1表示一个x的函数,此处的x就是一个形式参数,那么f(y)=y+1,不论形式参数换成什么字母,函数f所表示的意思是不变的。
在本实施例中,步骤4中通过调整所述无人设备集群中的被控物体的位置,减小P(x)与Q(x)的交叉熵,包括:
获取梯度下降模型,所述梯度下降模型为:
Figure PCTCN2019092627-appb-000012
其中,
Figure PCTCN2019092627-appb-000013
由于这个期望的解析解较难求出,因此采取蒙特卡洛方法计算,在以P(x)为概率密度函数的随机变量中采样M次,获得样本{X m},m=1,2,...,M,因此
Figure PCTCN2019092627-appb-000014
所述交叉熵为:E=-∫ ΩP(x)ln[Q(x)]dx,Ω={(x 1,...,x n)|x 1,...,x n∈ (-∞,+∞)}。
利用所述梯度下降模型,更新所述被控物体的位置:
Figure PCTCN2019092627-appb-000015
其中,
Figure PCTCN2019092627-appb-000016
表示计算期望值,且每次更新所述被控物体的位置时,将位于Y i处的被控物体移动至
Figure PCTCN2019092627-appb-000017
γ用于表示移动的速率,γ是一个大于0的常数。
进一步的,在每一次重复执行步骤3之前,衰减σ。由于本实施例采用的算法类似于优化,为了避免陷入局部最优解,应当将参数σ的初始值取的足够大,随着任务的进行,为了加快收敛速度,可以让σ随时间衰减。
在优选方案中,初始状态下σ=10,即在第一次执行步骤3时,σ=10。
目前现有的队形重构算法大多基于Leader-Follower模式,需要在所有的被控物体中确定一个leader,其它的个体作为follower按照某种方式跟随leader,其中包括:
(1)按照距离的方式:使得每一个follower与leader及它们的邻近个体保持一定的距离从而达到队形重构和保持的目的;
(2)按照相对位置的方式:使得每一个follower相对leader的位置达到控制目标从而达到队形重构和保持的目的;
(3)事件驱动的方式:将leader周围的空间划分为若干个区域,将follower穿越区域的分隔面作为事件传给控制器,然后控制器根据这些事件产生控制指令使得follower到达并保持在指定区域。
然而现有的算法并没有考虑对于由相同的个体构成的集群,任意两个个体互换,其队形不变的特性,从而增加了计算的复杂性。而且在任务执行过程中如果遭遇外界干扰导致当前队形被扰乱则需要对路径重新规划。
本实施例提供的方案,能够较为有效的解决上述问题,举例来说:
如图1所示的目标队形示意图,将一个由20个被控物体构成的无人机集群由任意初始位置重构成一个S形队列。
执行步骤1,获取S形队列中所有点的坐标的集合{X i}:
Figure PCTCN2019092627-appb-000018
Figure PCTCN2019092627-appb-000019
执行步骤2,令
Figure PCTCN2019092627-appb-000020
其中
Figure PCTCN2019092627-appb-000021
P(x)的计算结果可参照如图2所示的与目标队形有关的概率密度函数的等高线图。
执行步骤3,获取所有被控物体的当前队形用集合{Y i}表示,其中
Figure PCTCN2019092627-appb-000022
并建立一个与所有被控物体当前队形有关的概率密度函数Q(x),
Figure PCTCN2019092627-appb-000023
Figure PCTCN2019092627-appb-000024
其中Q i(x)是以Y i为均值,σ为标准差的多元正态分布即:
Figure PCTCN2019092627-appb-000025
其中σ=10。
算结果可参照如图4所示的与当前队形有关的概率密度函数的等高线图,此时被控物体的当前队形可参照如图3。
执行步骤4,调整所有被控物体的位置使得P和Q的交叉熵变小;计算交叉熵的方法是:
E=-∫ ΩP(x)ln[Q(x)]dx
通过梯度下降法调整被控物体的位置,即用下式更新被控物体的位置:
Figure PCTCN2019092627-appb-000026
其中:
Figure PCTCN2019092627-appb-000027
由于这个期望的解析解较难求出,因此采取蒙特卡洛方法计算,在以P(x)为概率密度函数的随机变量中采样2048次,每次采样的方法为:首先在集合{1,2,...,20}中随机选择一个数,记作i,然后在以P i(x)为概率密度函数的随机变量中采样一次。获得样本{X m},m=1,2,...,2048,因此
Figure PCTCN2019092627-appb-000028
重复执行步骤3-4,,然后令σ=0.995σ,直到交叉熵达到最小值。被控物体运行轨迹可以参照图5。
当σ的取值足够大时,在任务进行过程中,被控物体之间不会发生碰撞,原因如下:
设Y i,Y j为两个不同的被控物体的位置,则Y i,Y j之间距离的平方d 2=(Y i-Y j)(Y i-Y j) T,则
Figure PCTCN2019092627-appb-000029
由于:
Figure PCTCN2019092627-appb-000030
Figure PCTCN2019092627-appb-000031
当σ→+∞时,
Figure PCTCN2019092627-appb-000032
由于(Y i-X m)≠ (Y j-X m),所以
Figure PCTCN2019092627-appb-000033
又因为
Figure PCTCN2019092627-appb-000034
关于σ是连续的,因此存在一个s∈(0,+∞)使得当σ>s时
Figure PCTCN2019092627-appb-000035
又因为
Figure PCTCN2019092627-appb-000036
关于时间t是连续的,因此在任务执行过程中
Figure PCTCN2019092627-appb-000037
即在任务执行过程中
Figure PCTCN2019092627-appb-000038
始终大于0或始终小于0,所以d 2关于时间t是单调的,所以d 2的最小值出现在任务开始时或任务结束时,在任务执行过程中不会发生碰撞。
本实施例提供的多体队形重构的算法,做到了实时计算飞行并调整集群的飞行队形,无需事先将队形中的位置与被控物体建立一一对应的关系,无需事先进行路径规划,各被控物体不存在leader和follower之分,每个个体按照算法产生的指令移动时不会发生碰撞。
本说明书中的各个实施例均采用递进的方式描述,各个实施例之间相同相似的部分互相参见即可,每个实施例重点说明的都是与其他实施例的不同之处。尤其,对于设备实施例而言,由于其基本相似于方法实施例,所以描述得比较简单,相关之处参见方法实施例的部分说明即可。以上所述,仅为本发明的具体实施方式,但本发明的保护范围并不局限于此,任何熟悉本技术领域的技术人员在本发明揭露的技术范围内,可轻易想到的变化或替换,都应涵盖在本发明的保护范围之内。因此,本发明的保护范围应该以权利要求的保护范围为准。

Claims (7)

  1. 一种用于无人设备集群控制的多体队形重构方法,其特征在于,包括:
    步骤1、读取预设的目标队形,所述目标队形是由指定数量的点的坐标构成的集合,所述指定数量匹配所述无人设备集群中的无人设备的数量;
    步骤2、建立与所述目标队形关联的概率密度函数P(x);
    步骤3、获取所述无人设备集群中的所有被控物体的当前队形,并建立与所述当前队形关联的概率密度函数Q(x);
    步骤4、通过调整所述无人设备集群中的被控物体的位置,减小P(x)与Q(x)的交叉熵;
    重复执行步骤3和步骤4,直至所述无人设备集群中的被控物体组成的队形与所述目标队形一致。
  2. 根据权利要求1所述的方法,其特征在于,所述读取预设的目标队形,包括:
    获取所述目标队形中各个点的坐标的集合{X i};
    其中,{X i}中的元素表示所述目标队形中各个点的目标位置,i=1,2,...,N,所述目标队形中的点表示为
    Figure PCTCN2019092627-appb-100001
    表示第i个目标位置的坐标第j个分量,j=1,...,n,n表示根据目标队形所在空间的维数,n取值为2或3。
  3. 根据权利要求2所述的方法,其特征在于,所述建立与所述目标队形关联的概率密度函数P(x),包括:
    确定以X i为均值的多元正态分布P i(x),其中:
    Figure PCTCN2019092627-appb-100002
    建立与所述目标队形关联的概率密度函数P(x),其中,
    Figure PCTCN2019092627-appb-100003
    x={x j},j=1,…,n,N表示被控物体的数量,σ表示正态分布的标准差。
  4. 根据权利要求2所述的方法,其特征在于,所述建立与所述当前队形关联的概率密度函数Q(x),包括:
    获取所述当前队形中的所有被控物体的坐标的集合{Y i},其中,
    Figure PCTCN2019092627-appb-100004
    表示所述当前队形中的第i个被控物体位置的坐标第j个分量;
    确定以
    Figure PCTCN2019092627-appb-100005
    为均值的多元正态分布Q i(x),其中:
    Figure PCTCN2019092627-appb-100006
    建立与所述当前队形关联的概率密度函数Q(x),其中,
    Figure PCTCN2019092627-appb-100007
    x={x j},j=1,…,n,N表示被控物体的数量,σ表示正态分布的标准差。
  5. 根据权利要求4所述的方法,其特征在于,所述通过调整所述无人设备集群中的被控物体的位置,减小P(x)与Q(x)的交叉熵,包括:
    获取梯度下降模型,所述梯度下降模型为:
    Figure PCTCN2019092627-appb-100008
    其中,
    Figure PCTCN2019092627-appb-100009
    所述交叉熵为:E=-∫ ΩP(x)ln[Q(x)]dx,Ω={(x 1,...,x n)|x 1,...,x n∈(-∞,+∞)};
    利用所述梯度下降模型,更新所述被控物体的位置:
    Figure PCTCN2019092627-appb-100010
    其中,
    Figure PCTCN2019092627-appb-100011
    表示计算期望值,且每次更新所述被控物体的位置时,将位于Y i处的被控物体移动至
    Figure PCTCN2019092627-appb-100012
    γ用于表示移动的速率,γ是一个大于0的常数。
  6. 根据权利要求5所述的方法,其特征在于,在每一次重复执行步骤3之前,衰减σ。
  7. 根据权利要求6所述的方法,其特征在于,在第一次执行步骤3时,σ=10。
PCT/CN2019/092627 2018-10-22 2019-06-25 一种用于无人设备集群控制的多体队形重构方法 Ceased WO2020082768A1 (zh)

Priority Applications (1)

Application Number Priority Date Filing Date Title
US17/287,527 US11977396B2 (en) 2018-10-22 2019-06-25 Multi-body formation reconstruction method for unmanned device cluster control

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN201811229466.3 2018-10-22
CN201811229466.3A CN109472064B (zh) 2018-10-22 2018-10-22 一种用于无人设备集群控制的多体队形重构方法

Publications (1)

Publication Number Publication Date
WO2020082768A1 true WO2020082768A1 (zh) 2020-04-30

Family

ID=65665711

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2019/092627 Ceased WO2020082768A1 (zh) 2018-10-22 2019-06-25 一种用于无人设备集群控制的多体队形重构方法

Country Status (3)

Country Link
US (1) US11977396B2 (zh)
CN (1) CN109472064B (zh)
WO (1) WO2020082768A1 (zh)

Cited By (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN112099525A (zh) * 2020-08-31 2020-12-18 北京航空航天大学 一种航天器编队飞行低通讯连通保持协同控制方法
CN113220034A (zh) * 2021-05-18 2021-08-06 北京航空航天大学 自主重构和人工干预重构相结合的无人机集群重构系统
CN113359860A (zh) * 2021-07-20 2021-09-07 北京航空航天大学 一种基于通信状态的无人机集群重构方法
CN115113530A (zh) * 2022-07-19 2022-09-27 西北工业大学 一种闭合光滑曲线上的集群自组织均匀分散控制方法
CN120993940A (zh) * 2025-10-17 2025-11-21 北京航空航天大学 一种复杂环境下空地协同的集群时变编队控制方法

Families Citing this family (10)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN109472064B (zh) * 2018-10-22 2023-05-26 南京航空航天大学 一种用于无人设备集群控制的多体队形重构方法
CN109901584B (zh) * 2019-03-21 2020-04-24 南京大学 一种基于自组织的机器人编队方法、可读存储介质及机器人
CN113504794B (zh) * 2021-07-23 2023-04-21 中国科学院地理科学与资源研究所 一种无人机集群重构方法、系统和电子设备
CN113485426B (zh) * 2021-07-23 2023-04-14 中国科学院地理科学与资源研究所 一种无人机集群重构方法、系统、存储介质和电子设备
CN114397907B (zh) * 2021-12-20 2024-03-26 北京航空航天大学 一种基于多智能体的集群自适应重构行为建模方法
CN116466729B (zh) * 2022-01-12 2026-01-09 中国科学院声学研究所 一种基于交叉熵方法的uuv航路规划方法
CN115639841B (zh) * 2022-12-15 2023-03-17 北京航空航天大学 一种基于鲁棒牵制的无人机集群编队控制系统及控制方法
CN115755984B (zh) * 2022-12-23 2026-02-03 北京理工大学 一种飞行器编队位置重构解析优化方法
CN116859736B (zh) * 2023-07-07 2024-08-16 哈尔滨工程大学 面向无人艇编队护航任务的编队重构控制方法
CN117434847B (zh) * 2023-12-22 2024-04-02 西北工业大学 一种微纳卫星集群宏观队形重构与保持控制器设计方法

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN102591358A (zh) * 2012-03-12 2012-07-18 北京航空航天大学 一种多无人机的动态编队控制方法
CN103941728A (zh) * 2014-04-24 2014-07-23 北京航空航天大学 一种无人机密集自主编队的队形变换方法
US20180074520A1 (en) * 2016-09-13 2018-03-15 Arrowonics Technologies Ltd. Formation flight path coordination of unmanned aerial vehicles
CN108594853A (zh) * 2018-04-27 2018-09-28 中国人民解放军陆军工程大学 无人机队形控制方法
CN109472064A (zh) * 2018-10-22 2019-03-15 南京航空航天大学 一种用于无人设备集群控制的多体队形重构方法

Family Cites Families (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN103440541B (zh) * 2013-09-18 2016-03-30 山东大学 多风电场短期输出功率的联合概率密度预测方法
CN105044122B (zh) * 2015-08-25 2019-01-22 安徽工业大学 一种基于半监督学习模型的铜件表面缺陷视觉检测方法

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN102591358A (zh) * 2012-03-12 2012-07-18 北京航空航天大学 一种多无人机的动态编队控制方法
CN103941728A (zh) * 2014-04-24 2014-07-23 北京航空航天大学 一种无人机密集自主编队的队形变换方法
US20180074520A1 (en) * 2016-09-13 2018-03-15 Arrowonics Technologies Ltd. Formation flight path coordination of unmanned aerial vehicles
CN108594853A (zh) * 2018-04-27 2018-09-28 中国人民解放军陆军工程大学 无人机队形控制方法
CN109472064A (zh) * 2018-10-22 2019-03-15 南京航空航天大学 一种用于无人设备集群控制的多体队形重构方法

Cited By (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN112099525A (zh) * 2020-08-31 2020-12-18 北京航空航天大学 一种航天器编队飞行低通讯连通保持协同控制方法
CN112099525B (zh) * 2020-08-31 2021-10-15 北京航空航天大学 一种航天器编队飞行低通讯连通保持协同控制方法
CN113220034A (zh) * 2021-05-18 2021-08-06 北京航空航天大学 自主重构和人工干预重构相结合的无人机集群重构系统
CN113359860A (zh) * 2021-07-20 2021-09-07 北京航空航天大学 一种基于通信状态的无人机集群重构方法
CN113359860B (zh) * 2021-07-20 2022-07-22 北京航空航天大学 一种基于通信状态的无人机集群重构方法
CN115113530A (zh) * 2022-07-19 2022-09-27 西北工业大学 一种闭合光滑曲线上的集群自组织均匀分散控制方法
CN120993940A (zh) * 2025-10-17 2025-11-21 北京航空航天大学 一种复杂环境下空地协同的集群时变编队控制方法

Also Published As

Publication number Publication date
CN109472064B (zh) 2023-05-26
CN109472064A (zh) 2019-03-15
US20210349477A1 (en) 2021-11-11
US11977396B2 (en) 2024-05-07

Similar Documents

Publication Publication Date Title
WO2020082768A1 (zh) 一种用于无人设备集群控制的多体队形重构方法
Duan et al. Imperialist competitive algorithm optimized artificial neural networks for UCAV global path planning
US20210287556A1 (en) Method and apparatus for generating optimal path for an unmanned aerial vehicle
CN110456821B (zh) 基于动态触发机制的飞行器轨迹最优控制方法和系统
CN108681787A (zh) 基于改进双向快速扩展随机树算法的无人机路径优化方法
DE112015007209T5 (de) Universalsteuerung für robuste Trajektorienfolge bei unbemannten Multirotorluftfahrzeugen (UAVs)
CN105867421A (zh) 一种基于ph曲线的无人机路径规划方法
Liu et al. Leader–follower UAVs formation control based on a deep Q-network collaborative framework
CN104216289A (zh) 基于分布式进化算法的多飞行器编队的控制方法和装置
CN110488875B (zh) 基于动态逆的无人机跟踪目标初始段航向误差修正方法
CN114355976A (zh) 一种基于学习的控制无人机在风扰下完成轨迹跟踪的方法
WO2021077769A1 (zh) 一种基于流式计算的地图创建方法及其系统
CN110632940B (zh) 一种带混合量化器的多无人机主动抗干扰时变轨迹跟踪控制方法
CN110084414B (zh) 一种基于k次控制深度强化学习的空管防冲突方法
Lu et al. Research on flight training prediction based on incremental online learning: J. Lu et al.
CN114253285B (zh) 一种多飞行器协同队形集结方法
CN113566826A (zh) 一种基于二次规划的无人机路径规划方法
CN118192626A (zh) 一种基于UM-Event模型的融合运行侧向安全间距评估方法
CN104216416B (zh) 飞行器冲突解脱方法及设备
US11487301B2 (en) Method and device for generating an optimum vertical trajectory intended to be followed by an aircraft
He Online detection and modeling of safety boundaries for aerospace applications using active learning and bayesian statistics
Vela et al. A two-stage stochastic optimization model for air traffic conflict resolution under wind uncertainty
CN113535866A (zh) 一种飞行器航迹融合方法、系统和计算机可读存储介质
CN116560401B (zh) 一种无人机编队中僚机控制指令的确定方法及终端设备
Zhang et al. Learning-Based Fault-Tolerant Optimal Formation Control of Helicopters: An Incremental Fully Actuated System Approach

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 19875041

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE

122 Ep: pct application non-entry in european phase

Ref document number: 19875041

Country of ref document: EP

Kind code of ref document: A1

122 Ep: pct application non-entry in european phase

Ref document number: 19875041

Country of ref document: EP

Kind code of ref document: A1

32PN Ep: public notification in the ep bulletin as address of the adressee cannot be established

Free format text: NOTING OF LOSS OF RIGHTS PURSUANT TO RULE 112(1) EPC (EPO FORM 1205A DATED 30/11/2021)

122 Ep: pct application non-entry in european phase

Ref document number: 19875041

Country of ref document: EP

Kind code of ref document: A1