Artificial neural network-based relative positioning method and device to which optimized arrangement of reference nodes is applied in environment where fixed nodes do not exist

Abstract

The present invention relates to an artificial neural network-based method and device for relative positioning to which optimized arrangement of reference nodes is applied in an environment where fixed nodes do not exist. An artificial neural network-based relative positioning method according to one embodiment of the present invention comprises the steps of: collecting distance information between respective nodes with respect to a plurality of nodes forming a relative formation; selecting three reference nodes among the plurality of nodes; rearranging the relative formation of the plurality of nodes according to a preset rule on the basis of the three selected reference nodes; training an artificial neural network using the collected distance information between the respective nodes as input; and predicting the formation by estimating relative coordinate information of the respective nodes on the basis of the trained artificial neural network.

Description

A method and device for relative positioning based on an artificial neural network that applies optimal placement of reference nodes in an environment where fixed nodes do not exist.

[0001] The present invention relates to an optimal arrangement of reference nodes and a relative positioning technology based on an artificial neural network, and more particularly, to a relative positioning technology utilizing an artificial neural network using only distance information through an optimized arrangement of a cluster node system in an environment where no fixed nodes exist.

[0002] A swarm robot system is a system in which multiple intelligent robots collaborate to perform tasks. Typically, swarm robot systems use the Global Positioning System (GPS) to determine each other's absolute positions and avoid collisions while performing tasks. However, in environments where GPS is unavailable, such as indoors, alternative technologies like trilateration can be used to determine the positions of nodes. Trilateration requires three or more fixed anchors and uses linear distances from these anchors to determine the positions of nodes. Here, the anchors represent reference nodes used for large-scale estimation and are referred to as "reference nodes" below.

[0003] However, in situations where all nodes are mobile, it's difficult to establish a fixed reference node. Therefore, in situations where a fixed reference node doesn't exist, the development of relative positioning technology using only distance information is necessary. In this case, if all nodes are mobile and randomly placed in a given space, there's no fixed node with absolute position information, leading to ambiguity in relative positioning due to rotation, translation, and symmetry. This ambiguity can be resolved by applying three rules when operating cluster nodes.

[0004] However, if two or more of the three reference nodes have similar coordinates or the three nodes form a straight line, the accuracy of coordinate estimation deteriorates. To address this, methods that set a minimum distance between nodes may be unsuitable for clustered node systems that perform tasks while maintaining a fixed formation.

[0005] Embodiments of the present invention aim to provide a relative positioning method and device based on an artificial neural network applying an optimized arrangement of three reference nodes selected in an environment where no fixed nodes exist, for accurately predicting a formation by estimating relative coordinate information of each node based on an artificial neural network applying an optimized arrangement of three reference nodes selected in an environment where no fixed nodes exist.

[0006] Embodiments of the present invention aim to provide an artificial neural network-based relative positioning method and device that applies the optimized placement of reference nodes in an environment where no fixed nodes exist, in order to realize the optimized placement of swarm robots in a situation where no fixed nodes exist, and to accurately estimate the relative positions of deep learning-based swarm robots using only distance information.

[0007] Embodiments of the present invention aim to provide an artificial neural network-based relative positioning method and device that applies an optimized arrangement of reference nodes in an environment where no fixed nodes exist, in order to maintain the size of cluster nodes in a cluster node system and improve the performance of a relative positioning system through an optimized arrangement of three reference nodes.

[0008] However, the problem to be solved by the present invention is not limited to this, and may be expanded in various ways in environments that do not deviate from the spirit and scope of the present invention.

[0009] According to one embodiment of the present invention, a relative positioning method performed by a relative positioning device, comprising: a step of collecting distance information between each node of a plurality of nodes forming a relative formation; a step of selecting three reference nodes from among the plurality of nodes; a step of rearranging the relative formation of the plurality of nodes according to a preset rule centered on the three selected reference nodes; a step of training an artificial neural network using the collected distance information between each node as input; and a step of estimating relative coordinate information of each node based on the trained artificial neural network to predict the formation, wherein an artificial neural network-based relative positioning method applying an optimized arrangement of reference nodes in an environment where no fixed nodes exist can be provided.

[0010] The distance information between each node collected above may include distance errors.

[0011] The step of selecting the three reference nodes may select the three reference nodes from among the plurality of nodes by using any one of the sum of distances of three nodes from among the plurality of nodes, the mean absolute deviation (MAD) for three sides between the three nodes, the normalized MAD, and the variance for three sides between the three nodes.

[0012] The step of selecting the three reference nodes may include calculating the sum of distances of three nodes among the plurality of nodes, and selecting the combination of the three nodes with the largest calculated sum of distances as the three reference nodes.

[0013] The step of selecting the three reference nodes may include calculating the mean absolute deviation (MAD) for three edges between three nodes among the plurality of nodes, and selecting a combination of three nodes with the minimum variance for the three edges between the three nodes as the three reference nodes.

[0014] The step of selecting the three reference nodes may include calculating the mean absolute deviation (MAD) for three edges between three nodes among the plurality of nodes, normalizing the calculated mean absolute deviation (MAD) by dividing it by the sum of all edges, and selecting the combination of three nodes with the minimum normalized mean absolute deviation (MAD) as the three reference nodes.

[0015] The step of selecting the three reference nodes may include calculating the variance for three edges between three nodes among the plurality of nodes, normalizing the calculated variance by dividing it by the sum of all edges, and selecting the combination of three nodes with the minimum normalized variance as the three reference nodes.

[0016] Meanwhile, according to another embodiment of the present invention, there is provided an artificial neural network-based relative positioning device that applies an optimized arrangement of reference nodes in an environment where no fixed nodes exist, the device including: a memory that stores one or more programs; and a processor that executes the one or more stored programs, wherein the processor collects distance information between each node forming a relative formation, selects three reference nodes from among the plurality of nodes, rearranges the relative formation of the plurality of nodes according to a preset rule centered on the three selected reference nodes, trains an artificial neural network using the collected distance information between each node as input, and estimates relative coordinate information of each node based on the trained artificial neural network to predict the formation.

[0017] The distance information between each node collected above may include a distance error.

[0018] The processor may select three reference nodes from among the plurality of nodes by using any one of the sum of distances of three nodes from among the plurality of nodes, the mean absolute deviation (MAD) of three sides between the three nodes, the normalized MAD, and the variance of three sides between the three nodes.

[0019] The above processor can calculate the sum of distances of three nodes among the plurality of nodes, and select a combination of three nodes with the largest calculated sum of distances as three reference nodes.

[0020] The above processor can calculate the mean absolute deviation (MAD) for three sides between three nodes among the plurality of nodes, and select a combination of three nodes with the minimum calculated mean absolute deviation (MAD) as three reference nodes.

[0021] The above processor can calculate the mean absolute deviation (MAD) for three edges between three nodes among the plurality of nodes, normalize the calculated mean absolute deviation (MAD) by dividing it by the sum of all edges, and select a combination of three nodes with the minimum normalized mean absolute deviation (MAD) as three reference nodes.

[0022] The above processor can calculate the variance for three edges between three nodes among the plurality of nodes, normalize the calculated variance by dividing it by the sum of all edges, and select a combination of three nodes with the minimum normalized variance as three reference nodes.

[0023] The disclosed technology may have the following effects. However, this does not mean that a particular embodiment must include all or only the following effects, and thus the scope of the disclosed technology should not be construed as being limited thereby.

[0024] Embodiments of the present invention can accurately predict a formation by estimating relative coordinate information of each node based on an artificial neural network that applies an optimized arrangement of three reference nodes selected in an environment where no fixed nodes exist.

[0025] Embodiments of the present invention can realize optimized placement of swarm robots in a situation where there are no fixed nodes, and can accurately estimate the relative positions of swarm robots based on deep learning using only distance information.

[0026] Embodiments of the present invention can improve the accuracy of a relative position estimation system based on an artificial neural network utilizing distance information in an environment where no fixed nodes exist.

[0027] Embodiments of the present invention can improve the performance of a relative positioning system by maintaining the size of cluster nodes in a cluster node system and optimizing the arrangement of three reference nodes.

[0028] Embodiments of the present invention can maximize work efficiency by minimizing collisions between robots through optimized arrangement of cluster nodes.

[0029] Embodiments of the present invention can contribute to improving the performance of robot systems utilized in military operations and various fields by identifying the exact relative positions of cluster nodes.

[0030] FIG. 1 is a drawing for explaining a relative positioning operation performed by an artificial neural network-based relative positioning device according to one embodiment of the present invention.

[0031] FIG. 2 and FIG. 3 are diagrams for explaining factors that reduce the accuracy of coordinate estimation of artificial intelligence according to one embodiment of the present invention.

[0032] FIG. 4 is a flowchart for explaining an artificial neural network-based relative positioning method that applies optimized placement of reference nodes in an environment where no fixed nodes exist according to one embodiment of the present invention.

[0033] FIGS. 5 to 9 are diagrams for explaining a process of selecting an optimal reference node and rearranging a large group in a situation where multiple nodes are randomly arranged according to one embodiment of the present invention.

[0034] FIG. 10 is a configuration diagram of an artificial neural network-based relative positioning device that applies an optimized arrangement of reference nodes in an environment where no fixed nodes exist according to one embodiment of the present invention.

[0035] The present invention is capable of various modifications and embodiments. Specific embodiments are illustrated in the drawings and described in detail in the detailed description. However, this is not intended to limit the present invention to specific embodiments, and it is to be understood that all modifications, equivalents, and alternatives included within the technical spirit and scope of the present invention are included. In describing the present invention, if a detailed description of a related known technology is judged to obscure the gist of the present invention, the detailed description will be omitted.

[0036] Terms like "first" and "second" may be used to describe various components, but these terms do not limit the components themselves. These terms are used solely to distinguish one component from another.

[0037] The terminology used in this invention is solely for the purpose of describing specific embodiments and is not intended to limit the invention. The terminology used in this invention has been selected from widely used, current terms, taking into account the functions of the invention. However, this may vary depending on the intentions of those skilled in the art, precedents, or the emergence of new technologies. Furthermore, in certain cases, the applicant may arbitrarily select terms, in which case their meanings will be described in detail in the relevant description of the invention. Therefore, the terms used in this invention should not be defined simply as names of terms, but rather based on their meanings and the overall content of the invention.

[0038] Singular expressions include plural expressions unless the context clearly dictates otherwise. In the present invention, terms such as "comprise" or "have" are intended to indicate the presence of a feature, number, step, operation, component, part, or combination thereof described in the specification, but should be understood not to preclude the possibility of the presence or addition of one or more other features, numbers, steps, operations, components, parts, or combinations thereof.

[0039] [Explanation of symbols]

[0040] 100: Relative Positioning Device

[0041] 110: Memory

[0042] 120: Processor

[0043] Hereinafter, embodiments of the present invention will be described in detail with reference to the attached drawings. In describing with reference to the attached drawings, identical or corresponding components are assigned the same drawing numbers, and redundant descriptions thereof will be omitted.

[0044] FIG. 1 is a drawing for explaining a relative positioning operation performed by an artificial neural network-based relative positioning device according to one embodiment of the present invention.

[0045] As illustrated in FIG. 1, an artificial neural network-based relative positioning device (100) according to one embodiment of the present invention can derive the relative positions of all nodes as coordinates using only the distance information between nodes in a situation where all nodes are moving in real time. At this time, the relative positioning device (100) can estimate the two-dimensional coordinates of all nodes by using the artificial neural network-based distance information between nodes that applies the optimized arrangement of reference nodes in an environment where fixed nodes do not exist, thereby predicting the size of the nodes.

[0046] Here, the artificial neural network-based relative positioning device (100) collects distance information between nodes as input data in an environment where no fixed nodes exist. When the number of nodes is N, N There are two sets of measured distance information. For example, when the number of nodes is 5, the input distance information between nodes is 10.

[0047] And the artificial neural network-based relative positioning device (100) selects three reference nodes, rearranges the relative size of each node around the three selected reference nodes, estimates the relative coordinate information of each node based on the learned artificial neural network to predict the size, and outputs the relative coordinate information of each node as output data. When the number of nodes is N, (2×N-3) output values ​​can exist. For example, when the number of nodes is 5, the size of the output relative coordinate information is 7 (=2×5-3).

[0048] Meanwhile, in one embodiment of the present invention, ambiguity may arise in the problem of finding coordinates based on distance information. For example, even if nodes are of the same size, various coordinate information may be present during performance evaluation.

[0049] To resolve this ambiguity, the relative positioning device (100) according to one embodiment of the present invention performs a data generation rule that eliminates ambiguity and minimizes restrictions.

[0050] The relative positioning device (100) can rearrange the relative size of nodes according to the following rules 1 to 3 centered on three selected reference nodes.

[0051] Rule 1: Assume that the reference node 1 is the origin (x1= y1= 0).

[0052] Rule 2: Assume that the x-value of the 2nd reference node is positive and the y-value is 0 (x2> 0, y2= 0)

[0053] Rule 3: Assuming that the 3rd reference node exists in quadrants 1 and 2 (y3 > 0)

[0054] Here, x n , y n Each represents the x and y values ​​of the nth node.

[0055] FIG. 2 and FIG. 3 are diagrams for explaining factors that reduce the accuracy of coordinate estimation of artificial intelligence according to one embodiment of the present invention.

[0056] When the coordinates of the reference node are similar or the coordinates of three nodes are in a straight line, the accuracy of the coordinate estimation of the artificial intelligence according to one embodiment of the present invention may be reduced.

[0057] For example, let's explain by assuming that nodes 1 to 5 exist and the reference node is node 1 to 3.

[0058] As illustrated in Figure 2, in a situation where nodes 1 through 5 exist, nodes 1 and 2, among the reference nodes 1 through 3, are close together, and thus the coordinates of nodes 1 and 2 are similar. In this case, the accuracy of the artificial intelligence's coordinate estimation may be reduced.

[0059] As illustrated in Figure 3, in a situation where nodes 1 to 5 exist, nodes 1 to 3, which are reference nodes, form a straight line, so the y-coordinate is 0. In this case, the accuracy of the artificial intelligence's coordinate estimation may be reduced.

[0060] To address the factors that reduce the coordinate estimation accuracy illustrated in FIGS. 2 and 3, the relative positioning device (100) according to one embodiment of the present invention can select optimal reference nodes and arrange them according to data rules. Here, the optimal reference node combination may include a combination in which the reference nodes are far apart from each other, or a combination that is as close as possible to an equilateral triangle rather than a straight line.

[0061] FIG. 4 is a flowchart for explaining an artificial neural network-based relative positioning method that applies optimized placement of reference nodes in an environment where no fixed nodes exist according to one embodiment of the present invention.

[0062] In step S101, the artificial neural network-based relative positioning device (100) collects distance information between each node for a plurality of nodes forming a relative large area. Here, the artificial neural network-based relative positioning device (100) can collect and store distance information between all existing nodes. At this time, the method or technology for collecting distance information is not limited. The collected distance information may include distance errors. For example, the distance measurement error between nodes may be Gaussian noise following a normal distribution. In this case, in one embodiment of the present invention, the influence of distance errors on prediction accuracy can be minimized by using a dataset containing Gaussian noise.

[0063] In step S102, the artificial neural network-based relative positioning device (100) selects three reference nodes from among a plurality of nodes. Here, the artificial neural network-based relative positioning device (100) can select three optimal nodes that can serve as reference nodes from among all nodes. For example, when three reference nodes are selected, the node numbers can be designated as nodes 1, 2, and 3 in descending order.

[0064] At this time, according to embodiments of the present invention, there may be a total of four methods for selecting an optimal reference node.

[0065] First, the artificial neural network-based relative positioning device (100) can select the combination with the largest distance value between three reference nodes. The artificial neural network-based relative positioning device (100) can select the reference node by calculating the distance for combinations of three nodes that can exist in all nodes and then selecting the case with the largest distance value. At this time, when the three sides of the triangle formed by the combination of three nodes are a, b, and c, respectively, the artificial neural network-based relative positioning device (100) can select the case with the largest three sides of the triangle formed by the combination of three nodes according to the following [Mathematical Formula 1].

[0066]

[0067] Here, a, b, and c represent the three sides of the triangle.

[0068] Second, the artificial neural network-based relative positioning device (100) can select the number of cases in which the combination of three nodes that can exist in all nodes is closest to an equilateral triangle. There are three sides for each combination of three nodes. In order to select three nodes close to an equilateral triangle, the artificial neural network-based relative positioning device (100) can calculate the average again for the difference between each of the three sides and the average of the three sides, and select the number of cases in which the value is the smallest. Here, calculating the average of this difference can be referred to as MAD (Mean Absolute Deviation). The artificial neural network-based relative positioning device (100) can calculate the average m of the three sides as in [Mathematical Formula 2] below, and calculate the average again for the difference between each of the three sides and the average of the three sides as in [Mathematical Formula 3] below.

[0069]

[0070] Here, m represents the average of the three sides, and a, b, and c represent the three sides of the triangle formed by the combination of the three nodes.

[0071]

[0072] MAD represents the average of the differences between each of the three sides and the average of the three sides.

[0073] Third, the artificial neural network-based relative positioning device (100) first calculates the average of the three sides of a triangle formed by a combination of three nodes as in [Mathematical Formula 2], and then calculates the average (MAD) of the difference between the average and the three sides as in [Mathematical Formula 3]. This is the same as the second method.

[0074] Afterwards, the artificial neural network-based relative positioning device (100) can perform normalization by dividing the mean absolute deviation (MAD) obtained through the above process by the sum of all sides to select the combination of three nodes with the smallest normalized value. Here, the value calculated by performing normalization is MAD normalization That is, the artificial neural network-based relative positioning device (100) can perform normalization by dividing the mean absolute deviation (MAD) by the sum of all sides as in [Mathematical Formula 4] below.

[0075]

[0076] Here, a, b, and c represent the three sides of a triangle formed by the combination of three nodes. MAD normalization The mean absolute deviation (MAD) is the normalized value divided by the sum of all variables.

[0077] Fourth, the artificial neural network-based relative positioning device (100) is the same as the third method, but performs normalization by calculating the variance of the three sides between the three nodes, rather than the MAD, for the average of the three sides of the triangle formed by the three nodes. Here, the artificial neural network-based relative positioning device (100) can calculate the average of the three sides of the triangle formed by the three nodes as in [Mathematical Formula 2], and then calculate the variance of the three sides between the average and the three nodes as in [Mathematical Formula 5]. In addition, the artificial neural network-based relative positioning device (100) can perform normalization by calculating the calculated variance as in [Mathematical Formula 6].

[0078]

[0079] Here, represents the variance for three sides, m represents the mean of the three sides, and a, b, and c represent the three sides of the triangle formed by the combination of the three nodes.

[0080]

[0081] Here, Normalization represents the regularization of the variance of the three edges between the three nodes.

[0082] Meanwhile, in step S103, the artificial neural network-based relative positioning device (100) rearranges the relative formation of multiple nodes according to a preset rule centered on the three selected reference nodes. After the step of selecting the optimal reference nodes, the artificial neural network-based relative positioning device (100) can rearrange the relative formation centered on the three optimal nodes. Here, the artificial neural network-based relative positioning device (100) can apply at least one of origin (parallel) movement, rotational movement, and symmetrical movement according to the data generation rule applied to resolve the ambiguity of the formation. At this time, the ambiguity of the formation means that when the nodes are randomly arranged and only the coordinate information of the nodes is changed, such as rotational symmetry of the same formation, the artificial intelligence does not recognize it as the correct answer, resulting in a deterioration in coordinate estimation performance. At this time, even if the coordinate information of the nodes forming the formation is changed, the distance information between each node does not change.

[0083] In step S104, the artificial neural network-based relative positioning device (100) trains the artificial neural network using the collected distance information between each node as input. The artificial neural network-based relative positioning device (100) can train the artificial neural network using the measured distance information between each node as input. At this time, only the distance information between each existing node is used as input data, and the output data of the artificial intelligence is the coordinate information of each node.

[0084] In step S105, the artificial neural network-based relative positioning device (100) estimates the relative coordinate information of each node based on the learned artificial neural network to predict the formation.

[0085] Here, the artificial neural network-based relative positioning device (100) can estimate the relative coordinate information of each node based on an artificial neural network that inputs distance information between each node, using a relative estimation method or a group-by-group estimation method, thereby predicting the formation. In both methods, the values ​​of the first and second reference nodes, which are the origins, are excluded from the output data according to the criteria selected when creating the dataset.

[0086] First, regarding the simultaneous estimation method, the artificial neural network-based relative positioning device (100) can output the estimated x, y values ​​for all nodes when the distance information between all nodes is input to the artificial neural network. Therefore, as the number of nodes to be estimated increases, the sizes of input and output increase, so different models can be applied. The distance information between all nodes existing on the coordinate plane is used as input data of the artificial neural network. Through this distance information, all coordinates of each node are estimated simultaneously, and the estimated coordinate information at this time becomes the output data. For example, when there are N nodes to be estimated, the input data is the result of selecting 2 out of N and generating a combination. N There are C2, and the output data is (2N-3).

[0087] Second, regarding the group-by-group estimation method, the artificial neural network-based relative positioning device (100) can sequentially estimate coordinates by grouping the nodes to be estimated one by one with three reference nodes. At this time, the artificial neural network-based relative positioning device (100) uses a total of two artificial neural networks, and when the number of nodes is N, (N-2) coordinate estimations are required.

[0088] First, the artificial neural network-based relative positioning device (100) obtains the coordinates of the second reference node by utilizing the distance information from the first reference node. The artificial neural network-based relative positioning device (100) can estimate the coordinates of the third reference node based on the first artificial neural network. At this time, the input data is distance information between all reference nodes. The artificial neural network-based relative positioning device (100) can sequentially estimate the coordinates of the nodes one by one for the remaining nodes excluding the reference node based on the second artificial neural network. The input of the second artificial neural network is the distance information between the reference node and the node to be estimated and the coordinate information of the third reference node estimated through the first artificial neural network. The group-by-group estimation method does not require each model even if the number of nodes changes, and can estimate the coordinates of all nodes through two deep neural networks.

[0089] FIGS. 5 to 9 are diagrams for explaining a process of selecting an optimal reference node and rearranging a large group in a situation where multiple nodes are randomly arranged according to one embodiment of the present invention.

[0090] For example, referring to FIGS. 5 through 9, the process of selecting an optimal reference node and rearranging the large-scale nodes in a situation where nodes 1 through 5 are randomly placed will be described. Note that one embodiment of the present invention is not limited to a specific number of nodes.

[0091] As illustrated in Figure 5, nodes 1 to 5 may be randomly placed. That is, the formation illustrated in Figure 5 may be an initial formation in which nodes 1 to 5 are randomly placed. Coordinates for each node may be randomly generated. Here, the number of existing nodes is not limited.

[0092] As illustrated in FIG. 6, the artificial neural network-based relative positioning device (100) can select three reference nodes by selecting one or more of the four reference node selection methods described above. In the example illustrated in FIG. 6, the artificial neural network-based relative positioning device (100) can select nodes 1, 2, and 4 as the three reference nodes.

[0093] Afterwards, the artificial neural network-based relative positioning device (100) can change the selected reference nodes 1, 2, and 4 according to the rules.

[0094] First, as illustrated in Fig. 7, the artificial neural network-based relative positioning device (100) can move a large origin (parallel) to place the selected reference node No. 1 at the origin.

[0095] After the first reference node is positioned as the origin, the artificial neural network-based relative positioning device (100) can change the selected second reference node according to the rule, as illustrated in FIG. 8. Here, the artificial neural network-based relative positioning device (100) can perform a large rotational movement to obtain the angle of the second reference node. Through the large rotational movement, the second reference node can be positioned in the fourth quadrant. Subsequently, the artificial neural network-based relative positioning device (100) can perform a large re-rotation to apply the second reference node to the rule. Through this, the artificial neural network-based relative positioning device (100) can position the second reference node on the positive x-axis.

[0096] Next, as illustrated in FIG. 9, the artificial neural network-based relative positioning device (100) can perform a large-scale symmetrical movement to change the selected reference node number 4 according to the rule. Through this, the artificial neural network-based relative positioning device (100) can change the y-value of the selected reference node number 4 to a positive number.

[0097] Here, when comparing the randomly arranged initial formation shown in Fig. 5 with the rearranged formation based on the optimal three reference nodes shown in Fig. 9, the distance information between each node remains unchanged.

[0098] Thereafter, the artificial neural network-based relative positioning device (100) learns the artificial neural network by inputting distance information between each node in the rearranged formation based on three reference nodes, and estimates the relative coordinate information of each node based on the learned artificial neural network to predict the formation.

[0099] FIG. 10 is a configuration diagram of an artificial neural network-based relative positioning device that applies an optimized arrangement of reference nodes in an environment where no fixed nodes exist according to one embodiment of the present invention.

[0100] As illustrated in FIG. 10, an artificial neural network-based relative positioning device (100) that applies an optimized arrangement of reference nodes in an environment without fixed nodes according to an embodiment of the present invention includes a memory (110) and a processor (120). However, not all of the illustrated components are essential components. The artificial neural network-based relative positioning device (100) may be implemented with more components than the illustrated components, or may be implemented with fewer components.

[0101] Below, the specific configuration and operation of each component of the artificial neural network-based relative positioning device (100) of Fig. 10 will be described.

[0102] The memory (110) stores one or more programs related to an artificial neural network-based relative positioning method that applies an optimized arrangement of reference nodes in an environment where no fixed nodes exist.

[0103] The processor (120) executes one or more programs stored in the memory (110). The processor (120) collects distance information between each node for a plurality of nodes forming a relative formation, selects three reference nodes from among the plurality of nodes, rearranges the relative formation of the plurality of nodes according to a preset rule centered on the three selected reference nodes, trains an artificial neural network using the collected distance information between each node as input, and estimates relative coordinate information of each node based on the trained artificial neural network to predict the formation.

[0104] According to embodiments, the distance information between each collected node may include a distance error.

[0105] According to embodiments, the processor (120) may select three reference nodes from among the plurality of nodes by using any one of the sum of distances of three nodes from among the plurality of nodes, the mean absolute deviation (MAD) for three sides between the three nodes, the normalized MAD, and the variance for three sides between the three nodes.

[0106] According to embodiments, the processor (120) may calculate the sum of distances of three nodes among a plurality of nodes, and select a combination of three nodes with the largest calculated sum of distances as three reference nodes.

[0107] According to embodiments, the processor (120) may calculate the mean absolute deviation (MAD) for three sides between three nodes among a plurality of nodes, and select a combination of three nodes with the minimum calculated mean absolute deviation (MAD) as three reference nodes.

[0108] According to embodiments, the processor (120) may calculate the mean absolute deviation (MAD) for three edges between three nodes among a plurality of nodes, normalize the calculated mean absolute deviation (MAD) by dividing it by the sum of all edges, and select a combination of three nodes with the minimum normalized mean absolute deviation (MAD) as three reference nodes.

[0109] According to embodiments, the processor (120) may calculate variance for three edges among three nodes among a plurality of nodes, normalize the calculated variance by dividing it by the sum of all edges, and select a combination of three nodes with the minimum normalized variance as three reference nodes.

[0110] Meanwhile, according to one embodiment of the present invention, the various embodiments described above can be implemented as software including instructions stored in a machine-readable storage medium that can be read by a machine (e.g., a computer). The device is a device that can call instructions stored from the storage medium and operate according to the called instructions, and may include an electronic device (e.g., electronic device (A)) according to the disclosed embodiments. When an instruction is executed by a processor, the processor can perform a function corresponding to the instruction directly or by using other components under the control of the processor. The instruction may include code generated or executed by a compiler or interpreter. The machine-readable storage medium may be provided in the form of a non-transitory storage medium. Here, 'non-transitory' means that the storage medium does not contain a signal and is tangible, but does not distinguish between data being stored semi-permanently or temporarily in the storage medium.

[0111] Furthermore, according to one embodiment of the present invention, the method according to the various embodiments described above may be provided as included in a computer program product. The computer program product may be traded as a commodity between sellers and buyers. The computer program product may be distributed in the form of a device-readable storage medium (e.g., compact disc read-only memory (CD-ROM)) or online through an application store (e.g., Play Store™). In the case of online distribution, at least a portion of the computer program product may be temporarily stored or temporarily generated in a storage medium, such as the memory of a manufacturer's server, an application store's server, or a relay server.

[0112] Furthermore, according to one embodiment of the present invention, the various embodiments described above may be implemented in a computer-readable recording medium or a similar device using software, hardware, or a combination thereof. In some cases, the embodiments described herein may be implemented by the processor itself. In a software implementation, embodiments such as the procedures and functions described herein may be implemented as separate software modules. Each of the software modules may perform one or more functions and operations described herein.

[0113] Meanwhile, computer instructions for performing processing operations of a device according to the various embodiments described above may be stored in a non-transitory computer-readable medium. The computer instructions stored in such a non-transitory computer-readable medium, when executed by a processor of a specific device, cause the specific device to perform processing operations in the device according to the various embodiments described above. A non-transitory computer-readable medium refers to a medium that stores data semi-permanently and can be read by a device, rather than a medium that stores data for a short period of time, such as a register, cache, or memory. Specific examples of non-transitory computer-readable media may include a CD, DVD, hard disk, Blu-ray disk, USB, memory card, or ROM.

[0114] In addition, each of the components (e.g., modules or programs) according to the various embodiments described above may be composed of a single or multiple entities, and some of the sub-components described above may be omitted, or other sub-components may be further included in various embodiments. Alternatively or additionally, some components (e.g., modules or programs) may be integrated into a single entity, which may perform the same or similar functions as those performed by each of the respective components prior to integration. Operations performed by modules, programs or other components according to various embodiments may be executed sequentially, in parallel, iteratively or heuristically, or at least some operations may be executed in a different order, omitted, or other operations may be added.

[0115] Although the preferred embodiments of the present invention have been illustrated and described above, the present invention is not limited to the specific embodiments described above, and various modifications can be made by those skilled in the art without departing from the gist of the present invention as claimed in the claims. Furthermore, such modifications should not be understood individually from the technical idea or prospect of the present invention.

Claims

1. In a relative positioning method performed by a relative positioning device, A step of collecting distance information between each node for a plurality of nodes forming a relative large size; A step of selecting three reference nodes from among the above plurality of nodes; A step of rearranging the relative sizes of the plurality of nodes according to a preset rule centered on the three selected reference nodes; A step of learning an artificial neural network by inputting the distance information between each node collected above; and A relative positioning method based on an artificial neural network that applies optimal arrangement of reference nodes in an environment where fixed nodes do not exist, including a step of predicting a formation by estimating relative coordinate information of each node based on the learned artificial neural network.

2. In paragraph 1, A relative positioning method based on an artificial neural network that applies optimal placement of reference nodes in an environment where fixed nodes do not exist, and where the distance information between each node collected above includes a distance error.

3. In paragraph 1, The steps for selecting the three reference nodes are: A relative positioning method based on an artificial neural network that applies optimal arrangement of reference nodes in an environment where fixed nodes do not exist, wherein three reference nodes are selected from among the plurality of nodes by using any one of the sum of distances of three nodes, the mean absolute deviation (MAD) of three sides between the three nodes, the normalized MAD, and the variance of three sides between the three nodes.

4. In paragraph 1, The steps for selecting the three reference nodes are: A relative positioning method based on an artificial neural network that applies optimal arrangement of reference nodes in an environment where there are no fixed nodes, wherein the sum of distances of three nodes among the above-mentioned multiple nodes is calculated, and the combination of three nodes with the largest calculated sum of distances is selected as three reference nodes.

5. In paragraph 1, The steps for selecting the three reference nodes are: A relative positioning method based on an artificial neural network that applies an optimal arrangement of reference nodes in an environment where there are no fixed nodes, wherein the mean absolute deviation (MAD) is calculated for three sides between three nodes among the above-mentioned multiple nodes, and the combination of three nodes with the minimum calculated mean absolute deviation (MAD) is selected as three reference nodes.

6. In paragraph 1, The steps for selecting the three reference nodes are: A relative positioning method based on an artificial neural network that applies the optimal arrangement of reference nodes in an environment where there are no fixed nodes, wherein the mean absolute deviation (MAD) is calculated for three sides between three nodes among the above-mentioned multiple nodes, the calculated mean absolute deviation (MAD) is normalized by dividing the calculated mean absolute deviation (MAD) by the sum of all sides, and the combination of three nodes with the minimum normalized mean absolute deviation (MAD) is selected as three reference nodes.

7. In paragraph 1, The steps for selecting the three reference nodes are: A relative positioning method based on an artificial neural network that applies optimal arrangement of reference nodes in an environment where there are no fixed nodes, wherein the method calculates variance for three edges between three nodes among the above-mentioned multiple nodes, normalizes the calculated variance by dividing it by the sum of all edges, and selects a combination of three nodes with a minimum normalized variance as three reference nodes.

8. Memory for storing one or more programs; and comprising a processor for executing one or more of the stored programs; The above processor, Collect distance information between each node for multiple nodes that form a relatively large size, Select three reference nodes from the above multiple nodes, The relative sizes of the plurality of nodes are rearranged according to preset rules centered on the three selected reference nodes. The artificial neural network is trained using the distance information between each node collected above as input, An artificial neural network-based relative positioning device that estimates relative coordinate information of each node based on the learned artificial neural network and applies optimal arrangement of reference nodes in an environment where no fixed nodes exist, predicting a formation.

9. In paragraph 8, A relative positioning device based on an artificial neural network that applies optimal placement of reference nodes in an environment where fixed nodes do not exist, and where the distance information between each node collected above includes a distance error.

10. In paragraph 8, The above processor, An artificial neural network-based relative positioning device that applies optimal arrangement of reference nodes in an environment where fixed nodes do not exist, wherein three reference nodes are selected from among the plurality of nodes by using any one of the sum of distances of three nodes, the mean absolute deviation (MAD) of three sides between the three nodes, the normalized MAD, and the variance of the three sides between the three nodes.

11. In paragraph 8, The above processor, An artificial neural network-based relative positioning device that applies optimal arrangement of reference nodes in an environment where there are no fixed nodes, wherein the sum of distances of three nodes among the above-mentioned multiple nodes is calculated, and the combination of three nodes with the largest calculated sum of distances is selected as three reference nodes.

12. In paragraph 8, The above processor, A relative positioning device based on an artificial neural network that applies an optimal arrangement of reference nodes in an environment where there are no fixed nodes, wherein the mean absolute deviation (MAD) is calculated for three sides between three nodes among the above-mentioned plurality of nodes, and the combination of three nodes with the minimum calculated mean absolute deviation (MAD) is selected as three reference nodes.

13. In paragraph 8, The above processor, A relative positioning device based on an artificial neural network that applies an optimal arrangement of reference nodes in an environment where there are no fixed nodes, wherein the mean absolute deviation (MAD) is calculated for three sides between three nodes among the above-mentioned plurality of nodes, the calculated mean absolute deviation (MAD) is normalized by dividing the calculated mean absolute deviation (MAD) by the sum of all sides, and the combination of three nodes with the minimum normalized mean absolute deviation (MAD) is selected as three reference nodes.

14. In paragraph 8, The above processor, A relative positioning device based on an artificial neural network that applies an optimal arrangement of reference nodes in an environment where there are no fixed nodes, wherein the variance is calculated for three edges between three nodes among the above-mentioned multiple nodes, the calculated variance is normalized by dividing it by the sum of all edges, and the combination of three nodes with the minimum normalized variance is selected as three reference nodes.

Images