CN110321938B - State space construction method and device of intelligent unmanned cluster - Google Patents
State space construction method and device of intelligent unmanned cluster Download PDFInfo
- Publication number
- CN110321938B CN110321938B CN201910539923.7A CN201910539923A CN110321938B CN 110321938 B CN110321938 B CN 110321938B CN 201910539923 A CN201910539923 A CN 201910539923A CN 110321938 B CN110321938 B CN 110321938B
- Authority
- CN
- China
- Prior art keywords
- intelligent unmanned
- cluster
- state
- node
- intelligent
- 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.)
- Active
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/23—Clustering techniques
- G06F18/232—Non-hierarchical techniques
- G06F18/2321—Non-hierarchical techniques using statistics or function optimisation, e.g. modelling of probability density functions
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/043—Architecture, e.g. interconnection topology based on fuzzy logic, fuzzy membership or fuzzy inference, e.g. adaptive neuro-fuzzy inference systems [ANFIS]
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Evolutionary Computation (AREA)
- Software Systems (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- General Engineering & Computer Science (AREA)
- Automation & Control Theory (AREA)
- Health & Medical Sciences (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Bioinformatics & Computational Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Computational Mathematics (AREA)
- Fuzzy Systems (AREA)
- Mathematical Analysis (AREA)
- Mathematical Optimization (AREA)
- Pure & Applied Mathematics (AREA)
- Probability & Statistics with Applications (AREA)
- Evolutionary Biology (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- General Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- Mathematical Physics (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
- Feedback Control In General (AREA)
Abstract
The invention provides a state space construction method and a state space construction device of an intelligent unmanned cluster, wherein the method comprises the following steps: describing the state of each intelligent unmanned node at any moment when the intelligent unmanned node operates according to an ordered number pair set formed by the relative distance and the relative position relation between each intelligent unmanned node and the rest intelligent unmanned nodes in the intelligent unmanned cluster; and constructing a state space of the intelligent unmanned cluster according to the state of each intelligent unmanned node at any moment during operation. Aiming at the continuous change of the self state of the intelligent unmanned cluster and the combined explosion characteristic of the environmental state, the invention describes the state of each intelligent unmanned node in the intelligent unmanned cluster by adopting an ordinal pair set formed by the relative distance and the relative position relationship between each intelligent unmanned node and the rest intelligent unmanned nodes in the intelligent unmanned cluster, thereby constructing the state space of the intelligent unmanned cluster, defining the relationship between the node state and the cluster state, and being convenient for representation and calculation.
Description
Technical Field
The invention relates to the technical field of intelligent unmanned control, in particular to a state space construction method and device of an intelligent unmanned cluster.
Background
Under a specific natural environment scene, behavior cooperative work among a group of intelligent unmanned nodes which complete a specified task is called as an intelligent unmanned cluster. Along with the continuous expansion of the application field, the difficulty of communication, control and cooperation of the intelligent unmanned cluster is greatly increased, how to efficiently control each intelligent unmanned node to cooperatively complete the task is a problem to be solved urgently at present, and the good cooperative mechanism can improve the flexibility of intelligent unmanned cluster control, improve the communication efficiency and ensure the efficient and reliable completion of the task.
The behavior of the intelligent unmanned cluster is collaborative and diversified, for example, collaborative search and rescue, collaborative navigation, collaborative assembly of mechanical arms, collaborative handling of mobile robots or unmanned planes and the like of the intelligent unmanned plane. At present, most of researches on intelligent unmanned cluster behavior cooperation are solutions proposed based on control theories, a reinforcement learning method is not utilized to solve the problem of cluster behavior cooperation, and the solutions based on reinforcement learning pay more attention to a feedback mechanism and are more beneficial to cooperative control of intelligent unmanned nodes.
Although the types of cooperative work are different, the essence of the cooperative work is the behavior cooperation of the intelligent unmanned nodes, namely the change of the system state of the intelligent unmanned cluster. Therefore, there is a need to determine the state space of a cluster in a reinforcement learning method, and there is no method for constructing the state space of an intelligent unmanned cluster in the prior art.
Disclosure of Invention
The embodiment of the invention provides a state space construction method and device of an intelligent unmanned cluster, and aims to solve the problem that the state space of the intelligent unmanned cluster needs to be constructed when a reinforcement learning method is used for solving the cluster behavior coordination problem in the prior art.
In a first aspect, an embodiment of the present invention provides a state space construction method for an intelligent unmanned cluster, where the method includes:
describing the state of each intelligent unmanned node at any moment when the intelligent unmanned node runs according to an ordinal pair set formed by the relative distance and the relative position relation between each intelligent unmanned node and the rest intelligent unmanned nodes in the intelligent unmanned cluster;
and constructing a state space of the intelligent unmanned cluster according to the state of each intelligent unmanned node at any one time when the intelligent unmanned node runs.
As a preferred mode of the first aspect of the present invention, when describing a state of each intelligent unmanned node at any time when the intelligent unmanned node operates, the state of the ith intelligent unmanned node at time t is described by the following formula:
wherein { (d) 1 ,o 1 ),…,(d i-1 ,o i-1 ),(d i+1 ,o i+1 ),…,(d n ,o n ) D is an ordered number pair set representing the relative distance and the relative position relationship between the ith intelligent unmanned node and the rest n-1 intelligent unmanned nodes in the intelligent unmanned cluster, and j (j =1, \8230;, i-1, i +1, \8230;, n, and j! = i) represents a relative distance between the ith and jth intelligent unmanned nodes in the intelligent unmanned cluster; o j ＝α i +α j (j =1, \8230;, i-1, i +1, \8230;, n, and j! = i) represents a relative positional relationship between the ith and jth intelligent unmanned nodes in the intelligent unmanned cluster, α i An angle alpha representing the speed direction of the ith intelligent unmanned node and the angle of the connecting direction of the ith intelligent unmanned node and the jth intelligent unmanned node along the clockwise direction j And the angle between the speed direction of the jth intelligent unmanned node and the connecting direction of the ith intelligent unmanned node and the jth intelligent unmanned node along the clockwise direction is shown.
As a preferred mode of the first aspect of the present invention, when the state space of the intelligent unmanned cluster is constructed, the state space of the intelligent unmanned cluster is described by the following formula:
as a preferred mode of the first aspect of the present invention, the method further comprises:
and clustering the state space of the intelligent unmanned cluster by an adaptive fuzzy neural network clustering method to generate the clustered state space of the intelligent unmanned cluster.
As a preferred mode of the first aspect of the present invention, the clustering the state space of the intelligent unmanned aerial vehicle cluster by using an adaptive fuzzy neural network clustering method, and the generating the clustered state space of the intelligent unmanned aerial vehicle cluster includes:
determining the relative distance between the state of any intelligent unmanned node outside the currently newly acquired sample state set and the state of each intelligent unmanned node in the sample state set, and determining the weight of the state of each intelligent unmanned node in the sample state set according to each relative distance;
classifying the state of each intelligent unmanned node in the current intelligent unmanned cluster according to the ECM clustering algorithm and the weight of the state of each intelligent unmanned node in the sample state set to generate at least one cluster, wherein the parameters of the cluster comprise a cluster center and a cluster radius;
respectively taking the clustering center and the clustering radius of the cluster as the center and the width of a fuzzy membership function, and solving parameters in the fuzzy membership function by using a gradient descent algorithm;
defuzzifying the fuzzy membership function to generate a state space after the intelligent unmanned cluster is clustered.
In a second aspect, an embodiment of the present invention provides an apparatus for constructing a state space of an intelligent unmanned cluster, where the apparatus includes:
the state description unit is used for describing the state of each intelligent unmanned node at any moment when the intelligent unmanned node runs according to an ordinal number pair set formed by the relative distance and the relative position relation between each intelligent unmanned node and the rest intelligent unmanned nodes in the intelligent unmanned cluster;
and the space construction unit is used for constructing the state space of the intelligent unmanned cluster according to the state of each intelligent unmanned node at any moment when the intelligent unmanned node runs.
As a preferred mode of the second aspect of the present invention, when the state description unit describes the state of each intelligent unmanned node at any time when the intelligent unmanned node operates, the state of the ith intelligent unmanned node at time t is described by the following formula:
wherein { (d) 1 ,o 1 ),…,(d i-1 ,o i-1 ),(d i+1 ,o i+1 ),…,(d n ,o n ) D is an ordered number pair set representing the relative distance and the relative position relationship between the ith intelligent unmanned node and the rest n-1 intelligent unmanned nodes in the intelligent unmanned cluster, and j (j =1, \8230;, i-1, i +1, \8230;, n, and j! = i) represents a relative distance between the ith intelligent unmanned node and the jth intelligent unmanned node in the intelligent unmanned cluster; o j ＝α i +α j (j =1, \8230;, i-1, i +1, \8230;, n, and j! = i) represents a relative positional relationship between the ith and jth intelligent unmanned nodes in the intelligent unmanned cluster, α i The angle alpha of the speed direction of the ith intelligent unmanned node and the connecting direction of the ith intelligent unmanned node and the jth intelligent unmanned node along the clockwise direction j And the angle of the speed direction of the jth intelligent unmanned node and the connecting direction of the ith intelligent unmanned node and the jth intelligent unmanned node along the clockwise direction is shown.
As a preferred mode of the second aspect of the present invention, when the space construction unit constructs the state space of the intelligent unmanned cluster, the state space of the intelligent unmanned cluster is described by the following formula:
as a preferred mode of the second aspect of the present invention, the apparatus further comprises:
and the spatial clustering unit is used for clustering the state space of the intelligent unmanned cluster by an adaptive fuzzy neural network clustering method to generate the clustered state space of the intelligent unmanned cluster.
As a preferable mode of the second aspect of the present invention, the spatial clustering unit is specifically configured to:
determining a relative distance between the state of any intelligent unmanned node outside a currently newly acquired sample state set and the state of each intelligent unmanned node in the sample state set, and determining the weight of the state of each intelligent unmanned node in the sample state set according to each relative distance;
classifying the state of each intelligent unmanned node in the current intelligent unmanned cluster according to the ECM clustering algorithm and the weight of the state of each intelligent unmanned node in the sample state set to generate at least one cluster, wherein the parameters of the cluster comprise a cluster center and a cluster radius;
respectively taking the clustering center and the clustering radius of the cluster as the center and the width of a fuzzy membership function, and solving parameters in the fuzzy membership function by using a gradient descent algorithm;
defuzzifying the fuzzy membership function to generate a state space after the intelligent unmanned cluster is clustered. According to the state space construction method and device of the intelligent unmanned cluster, the ordinal number pair set formed by the relative distance and the relative position relation between each intelligent unmanned node and the rest intelligent unmanned nodes in the intelligent unmanned cluster is adopted for continuous change of the self state of the intelligent unmanned cluster and the combined explosion characteristic of the environmental state, so that the state of each intelligent unmanned node in the intelligent unmanned cluster is described, the state space of the intelligent unmanned cluster is further constructed, the relation between the node state and the cluster state is determined, and the representation and the calculation are convenient.
Therefore, a local behavior cooperation model of the intelligent unmanned cluster can be established based on the method, and a research scheme for solving the behavior cooperation of the intelligent unmanned cluster by using a fuzzy reinforcement learning algorithm is further provided.
Drawings
In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the drawings required to be used in the description of the embodiments are briefly introduced below, and it is obvious that the drawings in the description below are only some embodiments of the present invention, and it is obvious for those skilled in the art that other drawings can be obtained according to the drawings without creative efforts.
Fig. 1 is a schematic flowchart of a state space construction method of an intelligent unmanned cluster according to an embodiment of the present invention;
fig. 2 is a schematic state description diagram of an intelligent unmanned node according to an embodiment of the present invention;
fig. 3 is a schematic diagram of a relative position relationship of an intelligent unmanned node according to an embodiment of the present invention;
fig. 4 is a schematic diagram of initial clustering conditions of a state space of an intelligent unmanned cluster according to an embodiment of the present invention;
fig. 5 is a schematic structural diagram of a state space construction device of an intelligent unmanned cluster according to an embodiment of the present invention.
Detailed Description
In order to make the technical solutions of the present invention better understood, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
It should be noted that: like reference numbers and letters refer to like items in the following figures, and thus, once an item is defined in one figure, it need not be further defined or explained in subsequent figures.
The embodiment of the invention discloses a state space construction method of an intelligent unmanned cluster, which mainly comprises the following steps of:
101. describing the state of each intelligent unmanned node at any moment when the intelligent unmanned node operates according to an ordered number pair set formed by the relative distance and the relative position relation between each intelligent unmanned node and the rest intelligent unmanned nodes in the intelligent unmanned cluster;
102. and constructing a state space of the intelligent unmanned cluster according to the state of each intelligent unmanned node at any moment during operation.
In step 101, the coordination is a process of adjusting the relative relationship of one or some intelligent unmanned nodes in the cluster with respect to the rest intelligent unmanned nodes, which is a dynamic and continuous process, and in order to clearly depict the current state of the cluster and the environment, a set of continuous states S may be used t To represent a collaborative process of change, where t is time and the state is migrated over time.
Preferably, when the state of each intelligent unmanned node at any time when running is described, the state of the ith intelligent unmanned node at the time t is described by the following formula:
wherein { (d) 1 ,o 1 ),…,(d i-1 ,o i-1 ),(d i+1 ,o i+1 ),…,(d n ,o n ) D is an ordered number pair set representing the relative distance and relative position relationship between the ith intelligent unmanned node and the rest n-1 intelligent unmanned nodes in the intelligent unmanned cluster, and d j (j =1, \8230;, i-1, i +1, \8230;, n, and j! = i) represents a relative distance between the ith intelligent unmanned node and the jth intelligent unmanned node in the intelligent unmanned cluster; o. o j ＝α i +α j (j =1, \8230;, i-1, i +1, \8230;, n, and j! = i) represents a relative positional relationship between the ith and jth intelligent unmanned nodes in the intelligent unmanned cluster, α i An angle alpha representing the speed direction of the ith intelligent unmanned node and the angle of the connecting direction of the ith intelligent unmanned node and the jth intelligent unmanned node along the clockwise direction j And the angle between the speed direction of the jth intelligent unmanned node and the connecting direction of the ith intelligent unmanned node and the jth intelligent unmanned node along the clockwise direction is shown.
Referring to fig. 2, fig. 2 shows a relative distance and a relative position relationship between an ith intelligent unmanned node and a jth intelligent unmanned node in an intelligent unmanned cluster. According to the safety state of the intelligent unmanned nodes and the sensible capacity of sensor data, the relative distance d between the two intelligent unmanned nodes is known to be a continuous variable, so that the relative distance d depicts a continuous state.
Referring to fig. 3, fig. 3 specifically shows a relative position relationship between the ith intelligent unmanned node and the jth intelligent unmanned node, which can be divided into four cases, namely a, b, c and d shown in fig. 3. As can be seen from the graph, the description of the relative position relation o in the state of the ith intelligent unmanned node can be intuitively seen as a i And a j Are respectively in the range of { a i ∈[0°，360°)，a j E [0 °,360 °), so o e [0 °,720 °). The relative position between intelligent unmanned nodes is also a continuous variable, and a continuous state is depicted.
The state of the ith intelligent unmanned node in the intelligent unmanned cluster at the time t
Is composed of two successive states (d) k ，o k ) Where k =1,2, \ 8230;, i-1,i +1, \ 8230;, n, and k! = i, therefore
Again in a continuous state.
In step 102, a state space of the intelligent unmanned cluster is further constructed according to the state of each intelligent unmanned node in the intelligent unmanned cluster at any time when the intelligent unmanned node runs.
Preferably, when the state space of the intelligent unmanned cluster is constructed, the state space of the intelligent unmanned cluster at the time t is described by the following formula:
in the above formula, S t Is an n-dimensional vector representing the state space of the intelligent unmanned cluster, wherein
i =1,2, \8230, n is the state of each intelligent unmanned node at any one time while it is running.
The state space of the intelligent unmanned cluster constructed by the method is also continuous, and the size of the intelligent unmanned cluster is continuously changed along with the change of the total number n of the intelligent unmanned nodes, so that the problem of dimension disaster is easily caused. Moreover, the state space of the intelligent unmanned cluster needs to be redefined when the number of intelligent unmanned nodes is changed, which is not feasible in general fuzzy reinforcement learning. Due to the limitation of time and computing resources, the state space of the intelligent unmanned cluster is uncertain, and the behavior space of the intelligent unmanned nodes cannot be reduced, because the reduction of the behavior space is equal to the reduction of the capacity of the intelligent unmanned nodes, the task cannot be guaranteed to be completely executed. Therefore, in order to improve the convergence and the convergence rate of the blur reinforcement learning, the state space thereof needs to be compressed.
Preferably, the method further comprises:
103. and clustering the state space of the intelligent unmanned cluster by using a self-adaptive fuzzy neural network clustering method to generate the clustered state space of the intelligent unmanned cluster.
In step 103, an Adaptive Fuzzy Neural Network Clustering (AFNNC) method is used to compress the state space of the intelligent unmanned cluster. The external environment where the intelligent unmanned cluster is located is complex and changeable, all factors are mutually influenced and cross-coupled, the perception of the intelligent unmanned node on the environment state is difficult to accurately measure and calculate, and the problem can be well described by using fuzzy logic expression in the AFNNC method. In addition, the AFNNC method combines the learning optimization ability of the neural network, so that the network can reconstruct a network structure, adjust parameters and generate corresponding fuzzy rules according to training data.
Preferably, in one possible implementation, step 103 may be implemented as follows:
1031. determining the relative distance between the state of any intelligent unmanned node outside the currently newly acquired sample state set and the state of each intelligent unmanned node in the sample state set, and determining the weight of the state of each intelligent unmanned node in the sample state set according to each relative distance;
1032. classifying the states of all intelligent unmanned nodes in the current intelligent unmanned cluster according to the weight of the states of all intelligent unmanned nodes in the sample state set according to an ECM clustering algorithm to generate at least one cluster, wherein parameters of the cluster comprise a cluster center and a cluster radius;
1033. respectively taking the clustering center and the clustering radius of the cluster as the center and the width of the fuzzy membership function, and solving parameters in the fuzzy membership function by using a gradient descent algorithm;
1034. and defuzzifying the fuzzy membership function to generate a state space after the intelligent unmanned cluster is clustered.
In order to understand the specific implementation process of step 103, the following will describe steps 1031 to 1034 in detail:
(1) First, the euclidean distance between two vectors x and y is defined as:
wherein x, y is ∈ R P Wherein P represents the length of the sequence and is | x-y | ∈ [0,1 |)]。
(2) At the initial stage of operation of the intelligent unmanned cluster, a part of intelligent unmanned nodes generate aggregation to form a part of state set, which can also be called a sample state set. Wherein N is q Representing the number of intelligent unmanned nodes in the sample state set, q is less than n, and the sample state setThe states of all intelligent unmanned nodes within the set are proximity states for intelligent unmanned nodes outside the set of sample states.
Calculating any intelligent unmanned node x outside the newly acquired sample state set by using the Euclidean distance formula (1-1) i Current state x of q And N in the sample state set q Relative distance d = [ d ] between adjacent states of intelligent unmanned nodes 1 ,d 2 ,…,d n ]Where N is q The number of the intelligent unmanned nodes is determined empirically, and the weight of the state of each intelligent unmanned node in the sample state set can be expressed as:
w i ＝1-(d i -min i (d)),i＝1,2,…,N q (1-2)
in the formula (d) i Representing an intelligent unmanned node x outside the sample state set i To N in the sample state set q Relative distance between adjacent states of individual intelligent unmanned nodes, min i (d) Is a relative distance d = [ d = [) 1 ,d 2 ,…,d n ]The minimum value of (d).
(3) Clustering the state of each intelligent unmanned node outside the sample state set by using an ECM (operating Clustering Method) Clustering algorithm through the weight of the state of each intelligent unmanned node in the sample state set, wherein the Method comprises the following steps:
a. firstly, simply selecting an intelligent unmanned node from an intelligent unmanned cluster as a first cluster
Cluster center of
And radius the category at that time
Setting the state of the intelligent unmanned node as 0, continuously executing n times, and determining the clustering of the states of n intelligent unmanned nodes
b. Calculating intelligent unmanned node x outside newly acquired sample state set i With the cluster center C of the n clusters already determined Cj The relative distance d (i, j) can be calculated by the formula (1-1):
d(i,j)＝||x i -C Cj ||，j＝1,2,…,n。 (1-3)
c. if the relative distance d (i, j) calculated by the formula (1-2) is not more than the clustering radius of at least one cluster in the existing clusters, the intelligent unmanned nodes x outside the newly acquired sample state set are used i Merge into the cluster with the shortest distance to it, i.e.
d(i,m)＝‖x i -C Cm ‖＝min(||x i -C Cj ||)，j＝1,2,…,n， (1-4)
In the formula (I), the compound is shown in the specification,
representing intelligent unmanned nodes x in existing clusters and outside newly acquired sample state sets i The smallest cluster radius of distance. After the classification is finished, the intelligent unmanned node x outside the next sample state set is calculated continuously in the step b i+1 。
d. If the intelligent unmanned node x is outside the newly acquired sample state set i If the existing cluster is updated or the existing cluster does not meet the condition, the intelligent unmanned node x needs to be calculated i Comparing the distance between the intelligent unmanned node x and clusters needing to be updated with a threshold value, judging, if the distance is more than two times of the threshold value, establishing a new cluster, otherwise, establishing the intelligent unmanned node x i Into previously classified clusters:
the minimum distance calculated by selecting equation (1-5) is defined as s (i, a) and is determined by cluster C a Cluster radius with it
Is represented as follows:
there are two cases to consider in this case, one is: when s (i, a)>2D, a new cluster C needs to be established new With a cluster radius of
Secondly, the following steps: when s (i, a) ≦ 2D, the cluster C needs to be updated a And its clustering radius
Updated cluster of
Cluster radius of
e. Whether a new cluster is established or an existing cluster is updated, the intelligent unmanned node x of which the cluster center is outside the newly acquired sample state set i To the original cluster center, and the new cluster center to the intelligent unmanned node x outside the newly acquired sample state set i Is equal to the cluster radius, and then the intelligent unmanned node x i Classifying the nodes into the clusters which are divided before, and then turning to the step b to continue to calculate the intelligent unmanned nodes x outside the next sample state set after the classification is finished i+1 And finishing clustering of the intelligent unmanned nodes outside all the sample state sets.
(4) According to the result obtained by the ECM clustering algorithm in the previous step, the clustering center of the cluster is used as the center of the fuzzy membership function, the clustering radius is used as the width of the fuzzy membership function, and the method comprises the following steps:
in the formula (1-7), G ij Is the output of a fuzzy membership function, where x ij Is the jth state value, m, of the ith intelligent unmanned node in the intelligent unmanned cluster ij And σ ij The mean value and the variance of the fuzzy membership function corresponding to the jth state value of the ith intelligent unmanned node in the intelligent unmanned cluster are respectively, n is the number of the intelligent unmanned nodes in the intelligent unmanned cluster, and l is a regular number.
(5) Fuzzy rules are constructed, in the form shown below:
R l :IF x 1 is F l1 andx 2 is l2 and…x n is F lp THEN y＝n l ， (1-8)
in the formula F lj Is a fuzzy set defined by the fuzzy membership function in equations (1-7). The output of which can be expressed as:
n l ＝b l0 +b l1 x 1 +b l2 x 2 +…+b lp x p ， (1-9)
adopting an optimized center averaging method to obtain intelligent unmanned nodes x outside a newly acquired sample state set i State x of i ＝[x 1 ,x 2 ，…,x p ]Defuzzification, the output is:
in the formula, l is a rule number, p is the state number of the intelligent unmanned node outside the newly acquired sample state set, and the parameter alpha in the fuzzy membership function can be solved by using a gradient descent algorithm lj 、m ij And σ ij 。
b l0 (k+1)＝b l0 (k)-η b w i Φ(x i )[f (k) (x i )-t i ] (1-11)
b lj (k+1)＝b lj (k)-η b w i Φ(x i )[f (k) (x i )-t i ] (1-12)
In the formula, phi (x) i ) Comprises the following steps:
eta in the formula b ，η α ，η m And η σ Are respectively the parameter b j ，α lj ，m lj And σ lj The learning rate of (2). The following tables each represent:
wherein i is the number of intelligent unmanned nodes in the intelligent unmanned cluster, i =1,2, \8230, and N;
the variable dimension of the intelligent unmanned node outside the newly acquired sample state set is j, j =1,2, \ 8230;, P;
m represents the number of fuzzy rules, l =1,2, \8230;
the iteration step size is k, k =1,2, \ 8230.
Finally, after the clustering is finished, the generated intelligent unmanned cluster clustered state space is S t ＝(d，o，k)。
In addition, one index that can describe the performance of the current network structure is the cluster population error, using E i It is shown that the weight error function of the intelligent unmanned cluster can be calculated with the following formula.
W in the formula i Calculated from the formula (1-2), w i A weight representing a state of each intelligent unmanned node in the sample state set.
In fact, the compression of the state space is the classification of the state space, that is, some states are combined into one type of state by using a classifier, so that the dynamic division of the state space is realized, and the convergence and convergence speed of the fuzzy reinforcement learning can be promoted, and the learning speed is accelerated.
To further explain the result of the cluster compression of the state space of the intelligent unmanned cluster according to the embodiment of the present invention, a specific example will be described in detail below.
For the ith intelligent unmanned node in the intelligent unmanned cluster, taking the node as a reference, dividing the state space of the ith intelligent unmanned node into S t Where k denotes the number of intelligent unmanned nodes in the state, since the objectives of all intelligent unmanned nodes are the same, all intelligent unmanned nodes in the same state can take the same behavior, i.e. in case of state space determination, the behavior of the intelligent unmanned nodes is determined accordingly.
After the state space of the intelligent unmanned cluster is clustered and compressed, the original continuous state space can be discretized so as to facilitate the description of the problems. The relative distance d between any two intelligent unmanned nodes can be divided into three states according to the state of each intelligent unmanned node, wherein-1 represents a dangerous state, 0 represents a safe state, and 1 represents an adjustable state. As follows:
in the formula, R represents the self-safety distance between two intelligent unmanned nodes.
Then, the relative position relationship is defined as eight intervals, and the angle range is divided as follows:
after clustering is completed, the continuous state of the intelligent unmanned nodes is discretized, and the state size for each intelligent unmanned node is 3 x 8 x (n-1), so the state space size of the intelligent unmanned cluster is (3 x 8 x (n-1)) n 。
Referring to fig. 4 (a), the state space of the intelligent unmanned cluster may be divided into 24, and the number of intelligent unmanned nodes may be different in each state space after clustering. Referring to fig. 4 (B), the state of the intelligent unmanned cluster at a certain time is shown, wherein there are two intelligent unmanned nodes 1 and 2 in class a, two intelligent unmanned nodes 3 and 4 in class B, three intelligent unmanned nodes 5,6 and 7 in class C, and no node in other classes.
The compressed state space after the intelligent unmanned cluster clustering is shown in the following table 1-1, and at this time, the state space for one intelligent unmanned cluster can be clustered into 3 × 8=24 classes as follows.
TABLE 1-1 clustered State spaces for Intelligent unmanned clusters
In table, k i =0,1 \ 8230;, n (i =1,2, \8230;, 24), and has k 1 +k 2 +…+k 24 ＝n。
The clustering result shows that k intelligent unmanned nodes exist in the assumed relative distance and relative position relationship (d, o) at the moment, and the value of k is only two cases: firstly, when k =0, it means that no intelligent unmanned node exists in the state, and no cooperative action is needed; secondly, k is greater than 0 to indicate that k intelligent unmanned nodes exist in the state, and cooperation needs to be performed when k intelligent unmanned nodes exist, and since the intelligent unmanned nodes need to complete a common target and are in the same state, the k intelligent unmanned nodes can be classified into one group and can be indicated by 1, and the k intelligent unmanned nodes can be obtained according to the cluster consistency principleThe same action is taken by the point, and the state space of the intelligent unmanned cluster can be represented as S t And (d, o, k), wherein d is ∈ { -1,0,1}, o is ∈ {1,2,3,4,5,6,7,8}, and k is ∈ {0,1}. Here, the pair (d, o) is different clustering states, and there are 3 × 8=24 states, and k describes whether the intelligent unmanned node exists in the clustering state, that is, k =0 or k =1.
At this time, the state space of the intelligent unmanned cluster can be converted into a 24-bit integer representation, where each bit takes a value of 0 or 1, that is, a 24-bit binary number can be used for representation, that is, the size of the state space is: 224=16,777,216. For n intelligent unmanned nodes in the intelligent unmanned cluster, the relative distance and the relative position relationship are adopted during state description, and the same Q matrix can be adopted for representation:
it should be noted that the above-mentioned embodiments of the method are described as a series of actions for simplicity of description, but those skilled in the art should understand that the present invention is not limited by the described sequence of actions. Further, those skilled in the art will appreciate that the embodiments described in the specification are presently preferred and that no particular act is required to implement the invention.
Based on the same inventive concept, the embodiment of the invention also discloses a state space construction device of the intelligent unmanned cluster, and as shown in fig. 5, the device mainly comprises:
the state description unit 51 is configured to describe a state of each intelligent unmanned node at any time when the intelligent unmanned node operates according to an ordered number pair set formed by relative distances and relative position relationships between each intelligent unmanned node and the other intelligent unmanned nodes in the intelligent unmanned cluster;
and a space constructing unit 52, configured to construct a state space of the intelligent unmanned cluster according to a state of each intelligent unmanned node at any time when the intelligent unmanned node operates.
Preferably, when the state describing unit 51 describes the state of each intelligent unmanned node at any time when the intelligent unmanned node runs, the state of the ith intelligent unmanned node at the time t is described by the following formula:
wherein { (d) 1 ,o 1 ),…,(d i-1 ,o i-1 ),(d i+1 ,o i+1 ),…,(d n ,o n ) D is an ordered number pair set representing the relative distance and relative position relationship between the ith intelligent unmanned node and the rest n-1 intelligent unmanned nodes in the intelligent unmanned cluster, and d j (j =1, \8230;, i-1, i +1, \8230;, n, and j! = i) represents a relative distance between the ith and jth intelligent unmanned nodes in the intelligent unmanned cluster; o. o j ＝α i +α j (j =1, \8230;, i-1, i +1, \8230;, n, and j! = i) represents a relative positional relationship between the ith and jth intelligent unmanned nodes in the intelligent unmanned cluster, α i The angle alpha of the speed direction of the ith intelligent unmanned node and the connecting direction of the ith intelligent unmanned node and the jth intelligent unmanned node along the clockwise direction j And the angle between the speed direction of the jth intelligent unmanned node and the connecting direction of the ith intelligent unmanned node and the jth intelligent unmanned node along the clockwise direction is shown.
Preferably, when the space construction unit 52 constructs the state space of the intelligent unmanned cluster, the state space of the intelligent unmanned cluster is described by the following formula:
preferably, the apparatus further comprises:
and the spatial clustering unit 53 is configured to cluster the state space of the intelligent unmanned cluster by using an adaptive fuzzy neural network clustering method, and generate a clustered state space of the intelligent unmanned cluster.
Preferably, the spatial clustering unit 53 is specifically configured to:
determining the relative distance between the state of any intelligent unmanned node outside the currently newly acquired sample state set and the state of each intelligent unmanned node in the sample state set, and determining the weight of the state of each intelligent unmanned node in the sample state set according to each relative distance;
classifying the states of all intelligent unmanned nodes in the current intelligent unmanned cluster according to the weight of the states of all intelligent unmanned nodes in the sample state set according to an ECM clustering algorithm to generate at least one cluster, wherein parameters of the cluster comprise a cluster center and a cluster radius;
respectively taking the clustering center and the clustering radius of the cluster as the center and the width of the fuzzy membership function, and solving parameters in the fuzzy membership function by using a gradient descent algorithm;
defuzzifying the fuzzy membership function to generate a state space after intelligent unmanned cluster clustering.
In summary, the state space construction method and apparatus for an intelligent unmanned cluster provided in the embodiments of the present invention describe the state of each intelligent unmanned node in the intelligent unmanned cluster by using an ordinal pair set formed by the relative distance and the relative position relationship between each intelligent unmanned node and the other intelligent unmanned nodes in the intelligent unmanned cluster, and further construct the state space of the intelligent unmanned cluster, thereby defining the relationship between the node state and the cluster state, and facilitating representation and calculation, for the continuity change of the state of the intelligent unmanned cluster and the combined explosion characteristics of the environmental state. Therefore, a local behavior cooperation model of the intelligent unmanned cluster can be established based on the method, and a research scheme for solving behavior cooperation of the intelligent unmanned cluster by using a fuzzy reinforcement learning algorithm is further provided.
It should be noted that the state space construction apparatus of an intelligent unmanned cluster according to the embodiment of the present invention and the state space construction method of an intelligent unmanned cluster according to the foregoing embodiment belong to the same technical concept, and the specific implementation process thereof may refer to the description of the method steps in the foregoing embodiment, which is not described herein again.
It should be understood that the above state space building apparatus for an intelligent unmanned cluster includes only units that are logically divided according to functions implemented by the device apparatus, and in practical applications, the above units may be stacked or split. Moreover, functions implemented by the state space construction device for an intelligent unmanned cluster provided in this embodiment correspond to the state space construction method for an intelligent unmanned cluster provided in the foregoing embodiment one by one, and for a more detailed processing flow implemented by the device, detailed description is already made in the foregoing method embodiment, and detailed description is not given here.
According to the state space construction method and device of the intelligent unmanned cluster, provided by the embodiment of the invention, aiming at the continuous change of the self state of the intelligent unmanned cluster and the combined explosion characteristic of the environmental state, the state of each intelligent unmanned node in the intelligent unmanned cluster is described by using an ordinal pair set formed by the relative distance and the relative position relation between each intelligent unmanned node and the rest intelligent unmanned nodes, so that the state space of the intelligent unmanned cluster is constructed, the relation between the node state and the cluster state is defined, and the representation and the calculation are convenient. Therefore, a local behavior cooperation model of the intelligent unmanned cluster can be established based on the method, and a research scheme for solving behavior cooperation of the intelligent unmanned cluster by using a fuzzy reinforcement learning algorithm is further provided.
Those skilled in the art will appreciate that all or a portion of the steps of implementing the above-described method embodiments may be performed by hardware associated with program instructions. The foregoing program may be stored in a computer-readable storage medium. The program executes the steps of the above embodiments of the method when executed, and the storage medium includes various media such as ROM, RAM, magnetic or optical disk, etc. which can store program codes.
In the above embodiments of the present invention, the descriptions of the respective embodiments have respective emphasis, and for parts that are not described in detail in a certain embodiment, reference may be made to related descriptions of other embodiments.
The above description is only for the purpose of illustrating the preferred embodiments of the present invention and is not to be construed as limiting the invention, and any modifications, equivalents, improvements and the like that fall within the spirit and principle of the present invention are intended to be included therein.
Claims (8)
1. A state space construction method of an intelligent unmanned cluster is characterized by comprising the following steps:
describing the state of each intelligent unmanned node at any moment when the intelligent unmanned node runs according to an ordinal pair set formed by the relative distance and the relative position relation between each intelligent unmanned node and the rest intelligent unmanned nodes in the intelligent unmanned cluster;
constructing a state space of the intelligent unmanned cluster according to the state of each intelligent unmanned node at any moment during operation;
and clustering the state space of the intelligent unmanned cluster by an adaptive fuzzy neural network clustering method to generate the clustered state space of the intelligent unmanned cluster.
2. The method according to claim 1, wherein when describing the state of each intelligent unmanned node at any time during operation, the state of the ith intelligent unmanned node at time t is described by the following formula:
n is the total number of intelligent unmanned nodes in the intelligent unmanned cluster;
wherein { (d) 1 ,o 1 ),…,(d i-1 ,p i-1 ),(d i+1 ,o i+1 ),…,(d n ,o n ) D is an ordered number pair set representing the relative distance and the relative position relationship between the ith intelligent unmanned node and the rest n-1 intelligent unmanned nodes in the intelligent unmanned cluster, and j j =1, \8230;, i-1, i +1, \8230;, n, and j! = i represents the relative distance between the ith intelligent unmanned node and the jth intelligent unmanned node in the intelligent unmanned cluster；o j ＝α i +α j J =1, \8230;, i-1, i +1, \8230;, n, and j! = i denotes a relative position relationship between the ith intelligent unmanned node and the jth intelligent unmanned node in the intelligent unmanned cluster, and α i The angle alpha of the speed direction of the ith intelligent unmanned node and the connecting direction of the ith intelligent unmanned node and the jth intelligent unmanned node along the clockwise direction j And the angle between the speed direction of the jth intelligent unmanned node and the connecting direction of the ith intelligent unmanned node and the jth intelligent unmanned node along the clockwise direction is shown.
3. The method of claim 2, wherein the state space of the intelligent unmanned cluster is described by the following formula when the state space of the intelligent unmanned cluster is constructed:
4. the method of claim 1, wherein the clustering the state space of the intelligent unmanned clusters through an adaptive fuzzy neural network clustering method, and the generating the clustered state space of the intelligent unmanned clusters comprises:
determining the relative distance between the state of any intelligent unmanned node outside the currently newly acquired sample state set and the state of each intelligent unmanned node in the sample state set, and determining the weight of the state of each intelligent unmanned node in the sample state set according to each relative distance;
classifying the state of each intelligent unmanned node in the current intelligent unmanned cluster according to the weight of the state of each intelligent unmanned node in the sample state set according to an ECM clustering algorithm to generate at least one cluster, wherein the parameters of the cluster comprise a cluster center and a cluster radius;
respectively taking the clustering center and the clustering radius of the cluster as the center and the width of a fuzzy membership function, and solving parameters in the fuzzy membership function by using a gradient descent algorithm;
defuzzifying the fuzzy membership function to generate a state space after the intelligent unmanned cluster is clustered.
5. An intelligent unmanned cluster state space construction device, the device comprising:
the state description unit is used for describing the state of each intelligent unmanned node at any moment when the intelligent unmanned node runs according to an ordinal number pair set formed by the relative distance and the relative position relation between each intelligent unmanned node and the rest intelligent unmanned nodes in the intelligent unmanned cluster;
the space construction unit is used for constructing a state space of the intelligent unmanned cluster according to the state of any one moment when each intelligent unmanned node runs;
and the spatial clustering unit is used for clustering the state space of the intelligent unmanned cluster by an adaptive fuzzy neural network clustering method to generate the clustered state space of the intelligent unmanned cluster.
6. The apparatus according to claim 5, wherein when the state description unit describes the state of each intelligent unmanned node at any time when it runs, the state of the ith intelligent unmanned node at time t is described by the following formula:
n is the total number of intelligent unmanned nodes in the intelligent unmanned cluster;
wherein { (d) 1 ,o 1 ),…,(d i-1 ,o i-1 ),(d i+1 ,o i+1 ),…,(d n ,o n ) D is an ordered number pair set representing the relative distance and the relative position relationship between the ith intelligent unmanned node and the rest n-1 intelligent unmanned nodes in the intelligent unmanned cluster, and j j =1, \ 8230;, i-1, i +1, \ 8230;, n, and j! = i represents the relative distance between the ith intelligent unmanned node and the jth intelligent unmanned node in the intelligent unmanned cluster; o j ＝α i +α j J =1, \ 8230;, i-1, i +1, \ 8230;, n, and j! = i represents the relative position relationship between the ith intelligent unmanned node and the jth intelligent unmanned node in the intelligent unmanned cluster, and alpha represents i An angle alpha representing the speed direction of the ith intelligent unmanned node and the angle of the connecting direction of the ith intelligent unmanned node and the jth intelligent unmanned node along the clockwise direction j And the angle of the speed direction of the jth intelligent unmanned node and the connecting direction of the ith intelligent unmanned node and the jth intelligent unmanned node along the clockwise direction is shown.
7. The apparatus of claim 6, wherein the space construction unit describes the state space of the intelligent unmanned cluster by the following formula when constructing the state space of the intelligent unmanned cluster:
8. the apparatus according to claim 5, wherein the spatial clustering unit is specifically configured to:
determining a relative distance between the state of any intelligent unmanned node outside a currently newly acquired sample state set and the state of each intelligent unmanned node in the sample state set, and determining the weight of the state of each intelligent unmanned node in the sample state set according to each relative distance;
classifying the state of each intelligent unmanned node in the current intelligent unmanned cluster according to the ECM clustering algorithm and the weight of the state of each intelligent unmanned node in the sample state set to generate at least one cluster, wherein the parameters of the cluster comprise a cluster center and a cluster radius;
respectively taking the clustering center and the clustering radius of the cluster as the center and the width of a fuzzy membership function, and solving parameters in the fuzzy membership function by using a gradient descent algorithm;
defuzzifying the fuzzy membership function to generate a state space after the intelligent unmanned cluster is clustered.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910539923.7A CN110321938B (en) | 2019-06-20 | 2019-06-20 | State space construction method and device of intelligent unmanned cluster |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910539923.7A CN110321938B (en) | 2019-06-20 | 2019-06-20 | State space construction method and device of intelligent unmanned cluster |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN110321938A CN110321938A (en) | 2019-10-11 |
| CN110321938B true CN110321938B (en) | 2022-10-11 |
Family
ID=68121101
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910539923.7A Active CN110321938B (en) | 2019-06-20 | 2019-06-20 | State space construction method and device of intelligent unmanned cluster |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110321938B (en) |
Citations (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101825901A (en) * | 2010-03-31 | 2010-09-08 | 北京航空航天大学 | Multi-agent robot cooperative control method based on artificial physics method |
| CN102096415A (en) * | 2010-12-31 | 2011-06-15 | 重庆邮电大学 | Multi-robot formation method based on Ad-Hoc network and leader-follower algorithm |
| CN102749847A (en) * | 2012-06-26 | 2012-10-24 | 清华大学 | Cooperative landing method for multiple unmanned aerial vehicles |
| CN103197684A (en) * | 2013-04-25 | 2013-07-10 | 清华大学 | Method and system for cooperatively tracking target by unmanned aerial vehicle cluster |
| CN103631141A (en) * | 2013-12-11 | 2014-03-12 | 北京航空航天大学 | Light transmission hypothesis based intensive autonomous aerial vehicle formation control method |
| CN106295613A (en) * | 2016-08-23 | 2017-01-04 | 哈尔滨理工大学 | A kind of unmanned plane target localization method and system |
| CN108983823A (en) * | 2018-08-27 | 2018-12-11 | 安徽农业大学 | A kind of plant protection drone cluster cooperative control method |
| CN109343966A (en) * | 2018-11-01 | 2019-02-15 | 西北工业大学 | A kind of cluster organization method and device of unmanned node |
| CN109445456A (en) * | 2018-10-15 | 2019-03-08 | 清华大学 | A kind of multiple no-manned plane cluster air navigation aid |
| CN109885883A (en) * | 2019-01-21 | 2019-06-14 | 江苏大学 | A kind of control method of the unmanned vehicle transverse movement based on GK clustering algorithm model prediction |
-
2019
- 2019-06-20 CN CN201910539923.7A patent/CN110321938B/en active Active
Patent Citations (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101825901A (en) * | 2010-03-31 | 2010-09-08 | 北京航空航天大学 | Multi-agent robot cooperative control method based on artificial physics method |
| CN102096415A (en) * | 2010-12-31 | 2011-06-15 | 重庆邮电大学 | Multi-robot formation method based on Ad-Hoc network and leader-follower algorithm |
| CN102749847A (en) * | 2012-06-26 | 2012-10-24 | 清华大学 | Cooperative landing method for multiple unmanned aerial vehicles |
| CN103197684A (en) * | 2013-04-25 | 2013-07-10 | 清华大学 | Method and system for cooperatively tracking target by unmanned aerial vehicle cluster |
| CN103631141A (en) * | 2013-12-11 | 2014-03-12 | 北京航空航天大学 | Light transmission hypothesis based intensive autonomous aerial vehicle formation control method |
| CN106295613A (en) * | 2016-08-23 | 2017-01-04 | 哈尔滨理工大学 | A kind of unmanned plane target localization method and system |
| CN108983823A (en) * | 2018-08-27 | 2018-12-11 | 安徽农业大学 | A kind of plant protection drone cluster cooperative control method |
| CN109445456A (en) * | 2018-10-15 | 2019-03-08 | 清华大学 | A kind of multiple no-manned plane cluster air navigation aid |
| CN109343966A (en) * | 2018-11-01 | 2019-02-15 | 西北工业大学 | A kind of cluster organization method and device of unmanned node |
| CN109885883A (en) * | 2019-01-21 | 2019-06-14 | 江苏大学 | A kind of control method of the unmanned vehicle transverse movement based on GK clustering algorithm model prediction |
Non-Patent Citations (1)
| Title |
|---|
| 《基于多智能体的Option自动生成算法》;沈晶等;《智能系统学报》;20060331;第1卷(第1期);摘要 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN110321938A (en) | 2019-10-11 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Wang et al. | A multilayer path planner for a USV under complex marine environments | |
| Zhang et al. | A geometrical representation of McCulloch-Pitts neural model and its applications | |
| CN110782015A (en) | Training method and device for network structure optimizer of neural network and storage medium | |
| Tuba et al. | Mobile robot path planning by improved brain storm optimization algorithm | |
| Le et al. | Optimization via low-rank approximation for community detection in networks | |
| CN111460928A (en) | Human body action recognition system and method | |
| CN111460234B (en) | Graph query method, device, electronic equipment and computer readable storage medium | |
| CN113496247A (en) | Estimating an implicit likelihood of generating a countermeasure network | |
| US20210286375A1 (en) | Systems and methods for multi-agent system control using consensus and saturation constraints | |
| Zhang et al. | An affinity propagation clustering algorithm for mixed numeric and categorical datasets | |
| Toda et al. | Multilayer batch learning growing neural gas for learning multiscale topologies | |
| Talati et al. | Analysis, Simulation and Optimization of LVQ Neural Network Algorithm and Comparison with SOM | |
| CN111983923A (en) | Formation control method, system and equipment for limited multi-agent system | |
| Smith et al. | An investigation of how neural networks learn from the experiences of peers through periodic weight averaging | |
| Yu et al. | Distributed generation and control of persistent formation for multi-agent systems | |
| Munoz-Salinas et al. | Automatic tuning of a fuzzy visual system using evolutionary algorithms: single-objective versus multiobjective approaches | |
| CN110321938B (en) | State space construction method and device of intelligent unmanned cluster | |
| CN111291193B (en) | Application method of knowledge graph in zero-time learning | |
| Elsayed et al. | Memetic multi-topology particle swarm optimizer for constrained optimization | |
| Ghofrani et al. | A new probabilistic classifier based on decomposable models with application to internet traffic | |
| CN108898227A (en) | Learning rate calculation method and device, disaggregated model calculation method and device | |
| Liu et al. | Optimal formation of robots by convex hull and particle swarm optimization | |
| WO2021059527A1 (en) | Learning device, learning method, and recording medium | |
| Dantas et al. | Adaptive batch SOM for multiple dissimilarity data tables | |
| Neumeier et al. | A multidimensional graph fourier transformation neural network for vehicle trajectory prediction |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |