CN113552881A - Multi-path planning data set generation method for neural network training - Google Patents
Multi-path planning data set generation method for neural network training Download PDFInfo
- Publication number
- CN113552881A CN113552881A CN202110802240.3A CN202110802240A CN113552881A CN 113552881 A CN113552881 A CN 113552881A CN 202110802240 A CN202110802240 A CN 202110802240A CN 113552881 A CN113552881 A CN 113552881A
- Authority
- CN
- China
- Prior art keywords
- matrix
- points
- generating
- map
- neural network
- 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.)
- Granted
Links
- 238000000034 method Methods 0.000 title claims abstract description 37
- 238000012549 training Methods 0.000 title claims abstract description 32
- 238000013528 artificial neural network Methods 0.000 title claims abstract description 29
- 239000011159 matrix material Substances 0.000 claims abstract description 162
- 238000003491 array Methods 0.000 claims description 6
- 230000010354 integration Effects 0.000 claims description 6
- 238000004519 manufacturing process Methods 0.000 abstract description 2
- 230000006870 function Effects 0.000 description 9
- 238000011160 research Methods 0.000 description 3
- 238000013135 deep learning Methods 0.000 description 2
- 238000011161 development Methods 0.000 description 2
- 230000018109 developmental process Effects 0.000 description 2
- 238000013473 artificial intelligence Methods 0.000 description 1
- 230000006399 behavior Effects 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 230000002068 genetic effect Effects 0.000 description 1
- 238000003709 image segmentation Methods 0.000 description 1
- 238000003058 natural language processing Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G05—CONTROLLING; REGULATING
- G05D—SYSTEMS FOR CONTROLLING OR REGULATING NON-ELECTRIC VARIABLES
- G05D1/00—Control of position, course or altitude of land, water, air, or space vehicles, e.g. automatic pilot
- G05D1/02—Control of position or course in two dimensions
- G05D1/021—Control of position or course in two dimensions specially adapted to land vehicles
- G05D1/0212—Control of position or course in two dimensions specially adapted to land vehicles with means for defining a desired trajectory
- G05D1/0223—Control of position or course in two dimensions specially adapted to land vehicles with means for defining a desired trajectory involving speed control of the vehicle
-
- G—PHYSICS
- G05—CONTROLLING; REGULATING
- G05D—SYSTEMS FOR CONTROLLING OR REGULATING NON-ELECTRIC VARIABLES
- G05D1/00—Control of position, course or altitude of land, water, air, or space vehicles, e.g. automatic pilot
- G05D1/02—Control of position or course in two dimensions
- G05D1/021—Control of position or course in two dimensions specially adapted to land vehicles
- G05D1/0212—Control of position or course in two dimensions specially adapted to land vehicles with means for defining a desired trajectory
- G05D1/0221—Control of position or course in two dimensions specially adapted to land vehicles with means for defining a desired trajectory involving a learning process
-
- G—PHYSICS
- G05—CONTROLLING; REGULATING
- G05D—SYSTEMS FOR CONTROLLING OR REGULATING NON-ELECTRIC VARIABLES
- G05D1/00—Control of position, course or altitude of land, water, air, or space vehicles, e.g. automatic pilot
- G05D1/02—Control of position or course in two dimensions
- G05D1/021—Control of position or course in two dimensions specially adapted to land vehicles
- G05D1/0276—Control of position or course in two dimensions specially adapted to land vehicles using signals provided by a source external to the vehicle
Abstract
The invention relates to a multipath planning data set generation method for neural network training, which is characterized in that a random matrix generation algorithm is used for generating a 01 matrix map, the accuracy of the matrix map is improved through secondary loop traversal, a random function is used for setting a starting point and an end point, and a time dimension is introduced to perform multipath generation on the basis of a single path generated by a traditional path-finding A-x algorithm, so that collision avoidance is realized. Considering the richness of the data sets, corresponding data sets are respectively manufactured according to three conditions of different starting points and end points, the same end point and the same starting point, and finally the number of cycles is set to finish the manufacturing of the data sets.
Description
Technical Field
The invention relates to a path planning method, in particular to a multipath planning data set generation method based on neural network training.
Background
With the rapid development of modern robots, path planning technology has gained wide attention and application as an important branch and focus in the field of robot research. The existing path planning method comprises a genetic algorithm, an artificial potential field method, a random expanding tree algorithm, a free space method, a grid method and other intelligent heuristic algorithms, and due to inherent defects of the algorithms, a local optimal value is easy to fall into in the path planning and path searching process, and a global optimal solution cannot be obtained.
In recent years, with the great development of artificial intelligence industry and machine computing power, deep learning has been widely applied in the aspects of computer vision, image recognition, voice recognition, image segmentation, natural language processing and the like, and how to plan a path by using a neural network is also a research focus gradually. The success of deep learning depends greatly on the existence of a sufficient number of excellent data set samples for training, and currently, a large number of data sets uniformly used for robot path planning are not available for researchers to research and evaluate worldwide, and the goal of simultaneous multi-path prediction can be achieved.
Therefore, it is necessary to provide a method for generating a multipath planning data set for neural network training to achieve the purpose of path planning of the neural network.
Disclosure of Invention
The invention solves the problems in the prior art and provides a multi-path planning data set generation method for neural network training.
The technical scheme adopted by the invention is that a multipath planning data set generation method for neural network training comprises the following steps:
step 1: generating a map matrix only containing 0 and 1 by using a random matrix generation algorithm, setting the transverse and longitudinal sizes of the map matrix as positive integers n, wherein the probability of 0 generation in the map matrix is p, and the ratio of 0 to p is less than 1; 1 represents an obstacle and 0 represents no obstacle in the matrix, and the finally generated result matrix is saved as n x n and is marked as D1;
step 2: performing secondary traversal on the generated map matrix map by using a circulation method to ensure the accuracy of the data set;
and step 3: setting the positions of a starting point and an end point in the generated map matrix by using a random function, respectively constructing the positions into a starting point matrix n x n and an end point matrix n x n with the same size as the map matrix, and marking the starting point matrix as D2 and the end point matrix as D3;
and 4, step 4: generating a single path by combining a starting point matrix, an end point matrix and map matrix information by using a traditional A-path searching algorithm;
and 5: introducing a time dimension to perform multi-path generation, performing secondary traversal on the single-path information generated in the step 4 to prevent collision, and obtaining final path matrix information, thereby completing multi-path generation and storing the multi-path information as n x n;
step 6: aiming at three conditions of different starting points and end points, the same end point and the same starting point, adding corresponding constraint conditions in the step 3 and continuing to carry out the steps 4 and 5 to obtain and store multipath matrixes under the three conditions, wherein the different starting points and the end points are marked as D4, the same end points are marked as D5 and the same starting points are marked as D6;
and 7: and (3) performing circulation of the steps 1-6M times, storing the result of each operation, performing dimensionality integration on the D1, 2 and 3 matrixes to obtain three-dimensional input M (n) n 3, and selecting any two-dimensional matrix of the D4, 5 and 6 as output M (n) n 1 according to specific training requirements and transmitting the output M (n) n 1 to a neural network for training.
Preferably, in step 1, in the step 1, the random matrix generation algorithm is to firstly generate a matrix of all 1 s by using a numpy extended library of python, then select the matrix, change a part of 1 s to 0 s based on the set probability p generated by 0 s, and finally randomly shuffle the matrix by using a shuffle function in the numpy library to obtain an initial map matrix. In the step 1, n is a multiple of 5, and p is 0.6;
preferably, in the step 2, the loop method adopts while loop, and if the condition of quadratic traversal is not met, the loop will be continued, where quadratic traversal is satisfied by a matrix block traversal pair of 2 × 2Of (2) matrixAnd satisfyOf (2) matrixPerforming a replacement of matrix elements from 0 to 1 or from 1 to 0; wherein i and j respectively refer to an x-axis coordinate position and a y-axis coordinate position of the matrix;
preferably, in step 3, a method for generating a multipath planning data set for neural network training is provided. The method specifically comprises the following steps:
step 3.1: determining the number of starting points and end points, wherein the starting points and the end points are one group or multiple groups, and can be set by self according to training requirements, one group can be set in a single path, and multiple groups are set in multiple paths;
step 3.2: randomly generating one or more different points in the generated matrix map by using a random function, setting the points as starting points, generating a starting point matrix with the same size as the map matrix, setting the position of each starting point as 1, setting other positions as 0, and saving the matrix as n × n and marking the matrix as D2;
step 3.3: newly building a matrix which is completely the same as the map matrix, and setting the corresponding position values of the one or more different starting points in the matrix as 1 to obtain a matrix N;
step 3.4: randomly generating one or more different points in the matrix N generated in step 3.3 by using a random function, setting the points as end points, generating a matrix with the same size as the map matrix, setting the positions of the one or more different end points as 1, setting the other positions as 0, and saving the matrix as N × N, and marking the matrix as D3;
preferably, in the step 4, a method for generating a multipath planning data set for neural network training is provided. The method specifically comprises the following steps:
step 4.1: analyzing and operating the map matrix information, any starting point and corresponding end point information by using an A-star algorithm to generate a single path;
step 4.2: storing the position information of each path point under the single path into an array;
step 4.3: returning to the step 4.1 to select another group of starting points and end points until all the starting points and the end points under the current map matrix find corresponding single paths between the starting points and the end points and store the single paths in corresponding arrays;
preferably, in the step 5, a method for generating a multipath planning data set for neural network training is provided. The method specifically comprises the following steps:
step 5.1: introducing a time dimension, and performing secondary traversal comparison on the arrays stored in all the single paths generated in the step 4.3 under the condition of constant speed or variable speed;
step 5.2: if at the same moment, two or more single-path position information are overlapped to indicate that collision occurs, returning to the step 3 to select the starting point and the end point again;
step 5.3: generating a matrix with the same size as the matrix map until all the single-path information is not overlapped at any time under the condition of the specified speed, setting the position information corresponding to all the generated single paths as 1 in the matrix, setting other positions as 0, and storing the matrix as n x n;
and 5, introducing a time dimension, and setting collision avoidance at different speeds by changing the length of the array intercepted during traversal of the single-path information, so that the target of variable speed is realized, and the actual application scene is more fit.
Preferably, in the step 6, for better expanding the type and practicability of the data set, in the step 6, considering three cases of different starting points and end points, the same end point, and the same starting point, a constraint is added in the step 3: randomly generating an end point and a plurality of starting points aiming at the same end point; randomly generating a starting point and a plurality of end points aiming at the same starting point; ensuring that each point is not repeated aiming at different starting points and end points;
preferably, in step 7, the loop generation is performed on steps 1 to 6, the number of loops is set to M, the result of each operation is stored, the matrices D1, 2, 3 are subjected to dimension integration to obtain three-dimensional input M × n × 3, wherein one-dimensional n × 1 corresponds to the matrix D1, two-dimensional n × n 2 corresponds to the matrix D2, three-dimensional n × 3 corresponds to the matrix D3, any two-dimensional matrix of D4, 5, 6 is taken as output M × n and is transmitted to the neural network for training, the loop is completed, and M (n × 3) matrices are obtained as input, and M (n × n) matrices are obtained as output.
The invention provides a multipath planning data set generation method for neural network training, which is characterized in that a random matrix generation algorithm is used for generating a 01 matrix map, the accuracy of the matrix map is improved through secondary loop traversal, and then a random function is used for setting a starting point and an end point. And a time dimension is introduced to generate multiple paths on the basis of a single path generated by the traditional path-finding A-x algorithm, so that collision avoidance is realized. Considering the richness of the data sets, corresponding data sets are respectively manufactured according to three conditions of different starting points and end points, the same end point and the same starting point, and finally the number of cycles is set to finish the manufacturing of the data sets.
Drawings
Fig. 1 is a matrix corresponding to the behavior of the through obstacle according to the present invention.
Fig. 2 is a flow chart of a multipath planning data set generation method of the present invention.
Detailed Description
The present invention is described in further detail with reference to the following examples, but the scope of the present invention is not limited thereto.
The invention relates to a multipath planning data set generation method for neural network training, comprising the following steps.
Step 1: generating a map matrix only containing 0 and 1 by using a random matrix generation algorithm, setting the transverse and longitudinal sizes of the map matrix as positive integers n, wherein the probability of 0 generation in the map matrix is p, and the ratio of 0 to p is less than 1; 1 represents an obstacle and 0 represents no obstacle in the matrix, and the finally generated result matrix is saved as n x n and is marked as D1;
in the step 1, the random matrix generation algorithm is to firstly generate a matrix of 1 by using a numpy extended library of python, then select the matrix, change a part of 1 values into 0 values based on the set probability p generated by 0, and finally randomly disorder the matrix by using a shuffle function in the numpy extended library to obtain an initial map matrix. In the step 1, n is a multiple of 5, and p is 0.6;
step 2: performing secondary traversal on the generated map matrix map by using a circulation method to ensure the accuracy of the data set;
in the step 2, the loop method adopts while loop, and if the condition of secondary traversal is not met, the loop will be continued all the time, and the secondary traversal is that the matrix block traversal pair of 2x2 meets the requirementOf (2) matrixAnd satisfy The matrix Ii, je performs a replacement of matrix elements from 0 to 1 or from 1 to 0; wherein i and j respectively refer to an x-axis coordinate position and a y-axis coordinate position of the matrix;
and step 3: setting the positions of a starting point and an end point in the generated map matrix by using a random function, respectively constructing the positions into a starting point matrix n x n and an end point matrix n x n with the same size as the map matrix, and marking the starting point matrix as D2 and the end point matrix as D3;
the step 3 comprises the following steps:
step 3.1: determining the number of starting points and end points, wherein the starting points and the end points are one group or multiple groups, and can be set by self according to training requirements, one group can be set in a single path, and multiple groups are set in multiple paths;
step 3.2: randomly generating one or more different points in the generated matrix map by using a random function, setting the points as starting points, generating a starting point matrix with the same size as the map matrix, setting the position of each starting point as 1, setting other positions as 0, and saving the matrix as n × n and marking the matrix as D2;
step 3.3: newly building a matrix which is completely the same as the map matrix, and setting the corresponding position values of the one or more different starting points in the matrix as 1 to obtain a matrix N;
step 3.4: randomly generating one or more different points in the matrix N generated in step 3.3 by using a random function, setting the points as end points, generating a matrix with the same size as the map matrix, setting the positions of the one or more different end points as 1, setting the other positions as 0, and saving the matrix as N × N, and marking the matrix as D3;
and 4, step 4: generating a single path by combining a starting point matrix, an end point matrix and map matrix information by using a traditional A-path searching algorithm;
the step 4 comprises the following steps:
step 4.1: analyzing and operating the map matrix information, any starting point and corresponding end point information by using an A-path searching algorithm to generate a single path;
step 4.2: storing the position information of each path point under the single path into an array;
step 4.3: returning to the step 4.1 to select another group of starting points and end points until all the starting points and the end points under the current map matrix find corresponding single paths between the starting points and the end points and store the single paths in corresponding arrays;
and 5: introducing a time dimension to perform multi-path generation, performing secondary traversal on the path information generated in the step 4 to prevent collision, and obtaining final path matrix information, thereby completing multi-path generation and storing the multi-path information as n × n;
the step 5 comprises the following steps:
step 5.1: introducing a time dimension, and performing secondary traversal comparison on the arrays stored in all the single paths generated in the step 4.3 under the condition of constant speed or variable speed;
step 5.2: if two or more single-path position information are overlapped at the same moment, the collision is indicated, and the step 3 is returned to select the starting point and the end point again;
step 5.3: generating a matrix with the same size as the matrix map until all the single-path information is not overlapped at any time under the condition of the specified speed, setting the position information corresponding to all the generated single paths as 1 in the matrix, setting other positions as 0, and storing the matrix as n x n;
and 5, introducing a time dimension, and setting collision avoidance at different speeds by changing the length of the array intercepted during traversal of the single-path information, so that the target of variable speed is realized, and the actual application scene is more fit.
Step 6: aiming at three conditions of different starting points and end points, the same end point and the same starting point, adding corresponding constraint conditions in the step 3 and continuing to carry out the steps 4 and 5 to obtain and store multipath matrixes under the three conditions, wherein the different starting points and the end points are marked as D4, the same end points are marked as D5 and the same starting points are marked as D6;
in step 6, for better expanding the type and practicability of the data set, in step 6, considering three cases of different starting points and end points, the same end point, and the same starting point, constraints are added in step 3: randomly generating an end point and a plurality of starting points aiming at the same end point; randomly generating a starting point and a plurality of end points aiming at the same starting point; different starting point and end point conditions ensure that all points are not repeated;
and 7: and (3) performing circulation of the steps 1-6M times, storing the result of each operation, performing dimensionality integration on the D1, 2 and 3 matrixes to obtain three-dimensional input M (n) n 3, and selecting one group of two-dimensional matrixes in D4, 5 and 6 as output M (n) n 1 to be transmitted into the neural network for training according to specific training requirements.
In the step 7, the loop generation is performed on the steps 1 to 6, the number of the loop is set to be N, the result of each operation is stored, the matrices of D1, 2 and 3 are subjected to dimension integration to obtain three-dimensional input M (N) N3, wherein one-dimensional N1 corresponds to the matrix of D1, two-dimensional N2 corresponds to the matrix of D2, three-dimensional N3 corresponds to the matrix of D3, any two-dimensional matrix of D4, 5 and 6 is taken as output M (N) N and is transmitted to the neural network for training, the loop is finished, and M (N) N3 matrices are obtained as input, and M (N) matrices are taken as output.
Claims (9)
1. A method of generating a multipath planning dataset for neural network training, comprising: the method comprises the following steps:
step 1: generating a map matrix only containing 0 and 1 by using a random matrix generation algorithm, setting the transverse and longitudinal sizes of the map matrix as positive integers n, wherein the probability of 0 generation in the map matrix is p, 0< p <1, 1 represents an obstacle and 0 represents no obstacle in the matrix, and storing the finally generated matrix as n x n and marking the matrix as D1;
step 2: performing secondary traversal on the generated map matrix by using a circulation method to ensure the accuracy of the data set;
and step 3: setting the positions of a starting point and an end point in the generated map matrix by using a random function, respectively constructing the positions into a starting point matrix n x n and an end point matrix n x n with the same size as the map matrix, and marking the starting point matrix as D2 and the end point matrix as D3;
and 4, step 4: generating a single path by combining a starting point matrix, an end point matrix and map matrix information by using a traditional A-path searching algorithm;
and 5: introducing a time dimension to generate multiple paths, and performing secondary traversal on the path information of the single path generated in the step 4 to prevent collision to obtain final path matrix information, so that the multiple paths are generated and stored as n × n;
step 6: aiming at three conditions of different starting points and end points, the same end point and the same starting point, adding corresponding constraint conditions in the step 3 and continuing to carry out the steps 4 and 5 to obtain and store multipath matrixes under the three conditions, wherein the different starting points and the end points are marked as D4, the same end points are marked as D5 and the same starting points are marked as D6;
and 7: and (3) performing circulation of the steps 1-6M times, storing the result of each operation, performing dimension integration on the matrixes D1, D2 and D3 to obtain three-dimensional input M (n) n 3, taking any two-dimensional matrix of the groups D4, D5 and D6 as output M (n) n 1, and transmitting the output M (n) n 1 into a neural network for training.
2. A method of generating a multipath planning data set for use in neural network training as claimed in claim 1, wherein: in the step 1, the random matrix generation algorithm is to firstly generate a matrix of 1 by using a numpy extended library of python, then select the matrix, change a part of 1 values into 0 values based on the set probability p generated by 0, and finally randomly disorder the matrix by using a shuffle function in the numpy extended library to obtain an initial map matrix.
3. A method of generating a multipath planning data set for use in neural network training as claimed in claim 1, wherein: in the step 1, n is a multiple of 5, and p is 0.6.
4. A method of generating a multipath planning data set for use in neural network training as claimed in claim 1, wherein: in step 2, the second traversal is a 2 × 2 matrix block traversal, and the two-dimensional traversal satisfies the following condition Of (2) matrixAnd satisfy Of (2) matrixMaking matrix elements from 0 to 1 or from 1 to 10, where i and j refer to the x-axis coordinate position and the y-axis coordinate position of the matrix, respectively.
5. A method of generating a multipath planning data set for use in neural network training as claimed in claim 1, wherein: the step 3 comprises the following steps:
step 3.1: determining the number of starting points and end points, wherein the starting points and the end points are one group or a plurality of groups;
step 3.2: randomly generating one or more different points in the generated matrix map by using a random function, setting the points as starting points, generating a starting point matrix with the same size as the map matrix, setting the position of each starting point as 1, setting other positions as 0, and saving the starting point matrix as n × n and marking the starting point matrix as D2;
step 3.3: newly building a matrix which is completely the same as the map matrix, and setting the corresponding position values of the one or more different starting points in the matrix as 1 to obtain a matrix N;
step 3.4: one or more different points are randomly generated in the matrix N generated in step 3.3 using a random function, set as endpoints, a matrix of endpoints of the same size as the map matrix is generated, the position of each endpoint is set to 1, the other positions are set to 0, the endpoint matrix is saved as N x N and is labeled D3.
6. A method of generating a multipath planning data set for use in neural network training as claimed in claim 1, wherein: the step 4 comprises the following steps:
step 4.1: analyzing and operating the map matrix information, any starting point and corresponding end point information by using an A-path searching algorithm to generate a single path;
step 4.2: storing the position information of each path point under the single path into an array;
step 4.3: and returning to the step 4.1 to select another group of starting points and end points until all the starting points and the end points under the current map matrix find corresponding single paths between the starting points and the end points and store the single paths in corresponding arrays.
7. A method of generating a multipath planning data set for use in neural network training as claimed in claim 6, wherein: the step 5 comprises the following steps:
step 5.1: introducing a time dimension, and performing secondary traversal comparison on the arrays stored in all the single paths generated in the step 4.3 under the condition of constant speed or variable speed;
step 5.2: if two or more single-path position information are overlapped at the same moment, the collision is indicated, and the step 3 is returned to select the starting point and the end point again;
step 5.3: and generating a matrix with the same size as the matrix map until all the single-path information is not overlapped at any time under the condition of the specified speed, setting the position information corresponding to all the generated single paths as 1 in the matrix, setting other positions as 0, and storing the matrix as n x n.
8. A method of generating a multipath planning data set for use in neural network training as claimed in claim 1, wherein: in the step 6, for three conditions of different starting points and end points, the same end point and the same starting point, adding constraints in the step 3;
randomly generating an end point and a plurality of starting points aiming at the same end point;
randomly generating a starting point and a plurality of end points aiming at the same starting point;
and aiming at different starting points and end points, ensuring that each point does not generate repetition.
9. A method of generating a multipath planning data set for use in neural network training as claimed in claim 1, wherein: in step 7, the results of each operation are stored, and the matrices of D1, 2, and 3 are subjected to dimension integration to obtain three-dimensional input M (n × 3), where one-dimensional n × 1 corresponds to the matrix of D1, two-dimensional n × 2 corresponds to the matrix of D2, three-dimensional n × n 3 corresponds to the matrix of D3, any two-dimensional matrix of D4, 5, and 6 is taken as output M (n × n) and is transmitted to the neural network for training, and a cycle is completed to obtain M (n × 3) matrices as input, and M (n × n) matrices as output.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202110802240.3A CN113552881B (en) | 2021-07-15 | 2021-07-15 | Multipath planning data set generation method for neural network training |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202110802240.3A CN113552881B (en) | 2021-07-15 | 2021-07-15 | Multipath planning data set generation method for neural network training |
Publications (2)
Publication Number | Publication Date |
---|---|
CN113552881A true CN113552881A (en) | 2021-10-26 |
CN113552881B CN113552881B (en) | 2024-03-26 |
Family
ID=78131918
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202110802240.3A Active CN113552881B (en) | 2021-07-15 | 2021-07-15 | Multipath planning data set generation method for neural network training |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN113552881B (en) |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN114779780A (en) * | 2022-04-26 | 2022-07-22 | 四川大学 | Path planning method and system under random environment |
Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN103402235A (en) * | 2013-07-22 | 2013-11-20 | 上海交通大学 | Directional antenna-based multi-rate multi-path route optimization method |
CN109443363A (en) * | 2018-11-09 | 2019-03-08 | 厦门大学 | Certainty of dividing and ruling path optimizing algorithm |
CN110333739A (en) * | 2019-08-21 | 2019-10-15 | 哈尔滨工程大学 | A kind of AUV conduct programming and method of controlling operation based on intensified learning |
CN110345960A (en) * | 2019-06-13 | 2019-10-18 | 福建工程学院 | Route planning intelligent optimization method for avoiding traffic obstacles |
CN113093787A (en) * | 2021-03-15 | 2021-07-09 | 西北工业大学 | Unmanned aerial vehicle trajectory planning method based on velocity field |
-
2021
- 2021-07-15 CN CN202110802240.3A patent/CN113552881B/en active Active
Patent Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN103402235A (en) * | 2013-07-22 | 2013-11-20 | 上海交通大学 | Directional antenna-based multi-rate multi-path route optimization method |
CN109443363A (en) * | 2018-11-09 | 2019-03-08 | 厦门大学 | Certainty of dividing and ruling path optimizing algorithm |
CN110345960A (en) * | 2019-06-13 | 2019-10-18 | 福建工程学院 | Route planning intelligent optimization method for avoiding traffic obstacles |
CN110333739A (en) * | 2019-08-21 | 2019-10-15 | 哈尔滨工程大学 | A kind of AUV conduct programming and method of controlling operation based on intensified learning |
JP2021034050A (en) * | 2019-08-21 | 2021-03-01 | 哈爾浜工程大学 | Auv action plan and operation control method based on reinforcement learning |
CN113093787A (en) * | 2021-03-15 | 2021-07-09 | 西北工业大学 | Unmanned aerial vehicle trajectory planning method based on velocity field |
Non-Patent Citations (1)
Title |
---|
孟海东;王祺;慕春棣;陈丽萍;: "基于激光雷达的多路径地形匹配算法", 计算机仿真, no. 02, pages 69 * |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN114779780A (en) * | 2022-04-26 | 2022-07-22 | 四川大学 | Path planning method and system under random environment |
Also Published As
Publication number | Publication date |
---|---|
CN113552881B (en) | 2024-03-26 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN108803332B (en) | Path planning method based on improved biophysics | |
CN111696345A (en) | Intelligent coupled large-scale data flow width learning rapid prediction algorithm based on network community detection and GCN | |
CN108304926B (en) | Pooling computing device and method suitable for neural network | |
CN112464611B (en) | Automatic PCB wiring system based on cloud-end collaborative intelligent processing | |
CN112528591A (en) | Automatic PCB wiring method based on joint Monte Carlo tree search | |
Garcia et al. | GPU-based dynamic search on adaptive resolution grids | |
CN106202224B (en) | Search processing method and device | |
CN107451617B (en) | Graph transduction semi-supervised classification method | |
CN111738276A (en) | Image processing method, device and equipment based on multi-core convolutional neural network | |
CN115390565A (en) | Unmanned ship dynamic path planning method and system based on improved D-star algorithm | |
KR102305575B1 (en) | Method and system for highlighting similar areas using similarity between images | |
CN113552881A (en) | Multi-path planning data set generation method for neural network training | |
CN114756974A (en) | Wall distance calculation method considering object plane normal information | |
CN109074348A (en) | For being iterated the equipment and alternative manner of cluster to input data set | |
CN110728359A (en) | Method, device, equipment and storage medium for searching model structure | |
CN116128019A (en) | Parallel training method and device for transducer model | |
CN103268614B (en) | A kind of for many prospects be divided into cut prospect spectrum drawing generating method | |
CN113886226B (en) | Test data generation method of confrontation generation model based on twin network | |
CN106445960A (en) | Data clustering method and device | |
CN110188098B (en) | High-dimensional vector data visualization method and system based on double-layer anchor point map projection optimization | |
CN110598159A (en) | Local grid space analysis parallel computing method based on effective computing quantity | |
CN117312579B (en) | Method and system for generating data model search analysis text | |
CN114610034B (en) | Mobile robot path planning method | |
CN115858629B (en) | KNN query method based on learning index | |
Zhai et al. | A Novel Teaching-Learning-Based Optimization with Laplace Distribution and Experience Exchange |
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 |