CN116673947A - Mobile robot travel path point prediction method - Google Patents
Mobile robot travel path point prediction method Download PDFInfo
- Publication number
- CN116673947A CN116673947A CN202310605169.9A CN202310605169A CN116673947A CN 116673947 A CN116673947 A CN 116673947A CN 202310605169 A CN202310605169 A CN 202310605169A CN 116673947 A CN116673947 A CN 116673947A
- Authority
- CN
- China
- Prior art keywords
- space
- module
- time
- output
- self
- 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.)
- Pending
Links
- 238000000034 method Methods 0.000 title claims abstract description 36
- 230000007246 mechanism Effects 0.000 claims abstract description 25
- 238000000605 extraction Methods 0.000 claims description 57
- 239000011159 matrix material Substances 0.000 claims description 29
- 230000004927 fusion Effects 0.000 claims description 22
- 238000009826 distribution Methods 0.000 claims description 8
- 238000009827 uniform distribution Methods 0.000 claims description 8
- 230000008569 process Effects 0.000 claims description 7
- 238000011176 pooling Methods 0.000 claims description 6
- 230000002123 temporal effect Effects 0.000 claims description 6
- 230000009466 transformation Effects 0.000 claims description 3
- 238000012545 processing Methods 0.000 claims description 2
- 238000010606 normalization Methods 0.000 description 4
- 238000010586 diagram Methods 0.000 description 3
- 230000009286 beneficial effect Effects 0.000 description 2
- 230000007547 defect Effects 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 238000005096 rolling process Methods 0.000 description 1
Classifications
-
- B—PERFORMING OPERATIONS; TRANSPORTING
- B25—HAND TOOLS; PORTABLE POWER-DRIVEN TOOLS; MANIPULATORS
- B25J—MANIPULATORS; CHAMBERS PROVIDED WITH MANIPULATION DEVICES
- B25J9/00—Programme-controlled manipulators
- B25J9/16—Programme controls
- B25J9/1656—Programme controls characterised by programming, planning systems for manipulators
- B25J9/1664—Programme controls characterised by programming, planning systems for manipulators characterised by motion, path, trajectory planning
-
- B—PERFORMING OPERATIONS; TRANSPORTING
- B25—HAND TOOLS; PORTABLE POWER-DRIVEN TOOLS; MANIPULATORS
- B25J—MANIPULATORS; CHAMBERS PROVIDED WITH MANIPULATION DEVICES
- B25J9/00—Programme-controlled manipulators
- B25J9/16—Programme controls
- B25J9/1679—Programme controls characterised by the tasks executed
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01C—MEASURING DISTANCES, LEVELS OR BEARINGS; SURVEYING; NAVIGATION; GYROSCOPIC INSTRUMENTS; PHOTOGRAMMETRY OR VIDEOGRAMMETRY
- G01C21/00—Navigation; Navigational instruments not provided for in groups G01C1/00 - G01C19/00
- G01C21/20—Instruments for performing navigational calculations
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/25—Fusion techniques
-
- 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
- G06N3/042—Knowledge-based neural networks; Logical representations of neural networks
-
- 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
- G06N3/045—Combinations of networks
- G06N3/0455—Auto-encoder networks; Encoder-decoder networks
-
- 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
- G06N3/0464—Convolutional networks [CNN, ConvNet]
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- General Physics & Mathematics (AREA)
- Evolutionary Computation (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- General Engineering & Computer Science (AREA)
- Computing Systems (AREA)
- Mathematical Physics (AREA)
- Remote Sensing (AREA)
- Radar, Positioning & Navigation (AREA)
- Software Systems (AREA)
- Molecular Biology (AREA)
- Health & Medical Sciences (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- General Health & Medical Sciences (AREA)
- Mechanical Engineering (AREA)
- Robotics (AREA)
- Evolutionary Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Automation & Control Theory (AREA)
- Control Of Position, Course, Altitude, Or Attitude Of Moving Bodies (AREA)
Abstract
The invention is applicable to the technical field of mobile robots, and provides a method and a system for predicting a traveling path point of a mobile robot, wherein the method comprises the following steps: the historical space-time stream data is used as input of an encoder to extract time features and space features, output of the encoder is obtained, the output is integrated with the latest space-time data after being encoded, the traffic flow prediction results of all nodes of the target time step are obtained through space decoding and output layers, and the optimal path for the mobile robot to travel is selected from the paths with the shortest distances according to the minimum value of the sum of traffic flows. According to the method, the time complexity and the space complexity are reduced by introducing a ProbSparse self-attention mechanism into the space-time model, and the problem that the calculated time complexity and space complexity are higher when the space-time model is used for extracting the features is solved.
Description
Technical Field
The invention belongs to the technical field of mobile robots, and particularly relates to a method and a system for predicting a traveling path point of a mobile robot.
Background
With the development of economy and the continuous progress of robot technology, various mobile robots have been applied to various industries, and the planning of the travel path of the mobile robots is the key point of research. In a complex urban traffic road, the premise of planning a travel path of a mobile robot is that traffic flow needs to be accurately predicted. Many conventional traffic flow prediction methods use a physical model or make timing predictions based on a single road. The prediction method based on the physical model drive sets a plurality of variables and assumptions in the modeling process, and has the advantages of complex calculation process and low reusability. The time sequence prediction based on a single road only focuses on feature extraction in the time dimension, but ignores the space topological relation existing between roads, so that the defect of the time sequence prediction can be effectively overcome by adopting a space-time model, and the prediction precision is improved. However, the space-time model needs to extract the time dimension and the space dimension simultaneously, so that the model is large, the parameters are more, and the calculated time complexity and space complexity are high.
Disclosure of Invention
The invention aims to provide a method for predicting a traveling path point of a mobile robot. The method aims at solving the problems that when a space-time model is used for predicting the path point of the mobile robot, the space-time model is large in size, more in parameters and high in calculation time complexity and space complexity because the time dimension and the space dimension characteristics are required to be extracted simultaneously.
In one aspect, the present invention provides a method for predicting a travel path point of a mobile robot, the method comprising the steps of:
step one, encoding historical traffic flow space-time data of near N days to obtain a first code, wherein the first code is used as input of an encoder;
sequentially carrying out time dimension feature extraction and space dimension feature extraction on the input of the encoder through a time feature extraction module and a space feature module to obtain the output of the encoder, wherein the time feature extraction module comprises a ProbSparse self-attention mechanism, a one-dimensional convolution layer and a maximum pooling layer; the spatial feature extraction module comprises a ProbSparse self-attention mechanism and a GCN graph roll-up layer;
thirdly, encoding traffic flow space-time information of the latest M (M < N) days according to the encoding mode of the first step to obtain a second encoding;
inputting the output of the encoder and the second code into a fusion module, performing self-attention calculation on the second code for one time to obtain an intermediate result, and performing self-attention calculation on the output of the encoder and the intermediate result to obtain the output of the fusion module;
step five, performing space decoding on the output of the fusion module through a space decoding module, wherein the calculation flow of the space decoding module is the same as that of the space feature extraction module in the step two;
and step six, the output of the space decoding module passes through an output layer consisting of a one-dimensional convolution layer and a full connection layer to obtain a traffic flow prediction result of each node of the target time step, and the optimal path for the mobile robot to travel is selected from the shortest paths of the first several distances according to the minimum value of the traffic flow sum.
Further, the probspark self-attention mechanism adopts KL divergence to measure the similarity between self-attention distribution and uniform distribution, and the dot product pair distribution with low similarity is screened out, and the dot product pairs close to the uniform distribution are directly and averagely added, wherein the calculation formula of the probspark self-attention mechanism is as follows:
Q＝W q X，K＝W k X，V＝W v X
wherein X is input, d is feature vector length,is a sparse matrix of the same size as Q and contains only Q of the first u highest M values, where u=clnl is set Q C is a super parameter, and the score M is as follows:
further, in the spatial feature extraction module, weighting addition is performed by using half-sine distance between nodes and the normalized self-attention system, and finally normalization is performed to form a relation matrix, the relation matrix is used for replacing an adjacent matrix in the GCN graph convolution model to perform update calculation of node features, and the calculation process of the relation matrix A is as follows:
wherein E is a node distance matrix, lambda is a balance coefficient, and the element node distance weights E in the matrix ij The calculation formula of (2) is as follows:
wherein x is i ,x j Represents the longitude and latitude positions of two nodes, h (x i ,x j ) Is the half sine distance of node i, j.
Further, the method also comprises the step of repeating the time feature extraction and the space feature extraction once in the step two.
In the spatial feature extraction module, a GCN graph convolution model is adopted to update node features, wherein an adjacent matrix is replaced by a relation matrix, and the calculation process of the relation matrix A is as follows:
wherein E is a node distance matrix, lambda is a balance coefficient, and the element node distance weights E in the matrix ij The calculation formula of (2) is as follows:
wherein x is i ,x j Represents the longitude and latitude positions of two nodes, h (x i ,x j ) Is the half sine distance of node i, j.
Further, the method also comprises repeating the second step once again.
Further, the encoding includes projection encoding, position encoding and time encoding, and the first encoding is formed by adding three codes after linear transformation to obtain the same matrix size.
Further, the time code is obtained from 5 calculated values of a minute-hour duty cycle, an hour-day duty cycle, a day-week duty cycle, a day-month duty cycle, and a day-year duty cycle.
Further, the projection code is obtained using one-dimensional convolution.
Further, the position code is obtained by adopting a sine and cosine coding mode.
In another aspect, there is provided herein a mobile robot travel path point prediction system, the system comprising:
the coding module is used for respectively coding the historical traffic flow space-time data of the near N days and the traffic flow space-time data of the latest M (M < N) days to obtain a first code and a second code;
the device comprises a characteristic extraction module, a characteristic extraction module and a characteristic processing module, wherein the characteristic extraction module is used for extracting the characteristics of the time dimension and the characteristics of the space dimension of the input of an encoder to obtain the output of the encoder, and comprises a temporal characteristic extraction module and a spatial characteristic extraction module, and the temporal characteristic extraction module comprises a ProbSparse self-attention mechanism, a one-dimensional convolution layer and a maximum pooling layer; the spatial feature extraction module comprises a ProbSparse self-attention mechanism and a GCN graph roll-up layer;
the fusion module is used for inputting the encoder output and the second code into the fusion module, performing self-attention calculation on the second code once to obtain an intermediate result, and performing self-attention calculation on the encoder output and the intermediate result to obtain a fusion module output;
the space decoding module is used for performing space decoding on the output of the fusion module, and the calculation flow of the space decoding module is the same as that of the space feature extraction module;
and the output module is used for outputting the space decoding module to pass through an output layer formed by a one-dimensional convolution layer and a full connection layer to obtain the traffic flow prediction result of each node of the target time step, and selecting the optimal path for the mobile robot to travel from the shortest paths of the first several distances according to the minimum value of the traffic flow sum.
The invention has the beneficial effects that: the invention provides a method and a system for predicting a mobile robot point, aiming at the problems that a space-time model for simultaneously extracting time dimension and space dimension characteristics is large, parameters are more, and calculated time complexity and space complexity are high in the process of predicting a running path point by using the space-time model of a mobile robot. In the invention, the ProbSparse self-attention mechanism is introduced into the time feature extraction module and the space feature extraction module, the KL divergence is adopted in the ProbSparse self-attention mechanism to measure the similarity of self-attention distribution and uniform distribution, the dot product pair distribution with low similarity is screened out, and the dot product pair close to the uniform distribution is directly and averagely added, so that the calculated time complexity and space complexity can be reduced, and the problem of higher calculated time complexity and space complexity is solved. In addition, when the invention encodes traffic flow space-time data, projection encoding, position encoding and time encoding are adopted, and the three encoding modes enable the model to more pertinently model the relation between input and output by combining the relative positions of the time stamp and the data in the self-attention mechanism and the subsequent feature extraction module, so that the time information and the distance information are better fused into the model.
Drawings
Fig. 1 is a flowchart of a method for predicting a travel path point of a mobile robot according to embodiment 1 of the present invention;
fig. 2 is a schematic structural diagram of a time feature extraction module in a method for predicting a travel path point of a mobile robot according to embodiment 1 of the present invention;
fig. 3 is a schematic structural diagram of a spatial feature extraction module in a method for predicting a travel path point of a mobile robot according to embodiment 1 of the present invention;
fig. 4 is a schematic diagram of a mobile robot travel path point prediction system according to embodiment 2 of the present invention.
Detailed Description
The present invention will be described in further detail with reference to the drawings and examples, in order to make the objects, technical solutions and advantages of the present invention more apparent. It should be understood that the specific embodiments described herein are for purposes of illustration only and are not intended to limit the scope of the invention.
The following describes in detail the implementation of the present invention in connection with specific embodiments:
example 1:
fig. 1 shows a flow of implementation of the method for predicting a traveling path point of a mobile robot according to an embodiment of the present invention, and for convenience of explanation, only the portions related to embodiment 1 of the present invention are shown, which are described in detail below:
and S1, encoding the historical traffic flow space-time data of nearly 7 days to obtain a first code, and taking the first code as the input of an encoder.
Further, the codes comprise projection codes, position codes and time codes, and the first code is formed by adding three codes after linear transformation to obtain the same matrix size. The three coding modes enable the model to more pertinently model the relation between input and output by combining the self-attention mechanism with the subsequent feature extraction module and combining the timestamp and the data relative position, and are also beneficial to better integrating time information and distance information into the model.
Further, in time encoding, the following 5 values are calculated from the time stamp of each time step: the time step global time information can be known by the model, so that the time step global time information is more accurate than the time step global time information which is simply distinguished from the afternoon, the weekend, the working days and the four seasons by the aid of the method, and the numerical value is limited to a certain range.
Further, projection coding is obtained by employing one-dimensional convolution.
Further, the position coding is obtained by adopting a sine and cosine coding mode, and in the self-attention mechanism, since the operation of the self-attention mechanism is undirected, the position coding is required to transmit the position information to the model.
The encoder portion of the model is composed of a temporal feature extraction module and a spatial feature extraction module.
Step S2, performing time dimension feature extraction through the input of a time feature extraction module encoder, wherein the time feature extraction module sequentially comprises a ProbSparse self-attention mechanism, a residual connection normalization layer, a one-dimensional convolution layer, a residual connection normalization layer and a maximum pooling layer as shown in FIG. 2.
Further, the ProbSparse self-attention mechanism adopts KL divergence to measure the similarity of self-attention distribution and uniform distribution, the dot product pair distribution with low similarity is screened, and dot product pairs close to the uniform distribution are directly and averagely added, so that the calculated time complexity and space complexity are reduced, and the calculation formula of the ProbSparse self-attention mechanism is as follows:
Q＝W q X，K＝W k X，V＝W v X
wherein X is input, d is feature vector length,is a sparse matrix of the same size as Q and contains only Q of the first u highest M values, where u=cln L is set Q C is a super parameter, and the score M is as follows:
and S3, performing spatial dimension feature extraction on the input of the encoder through a spatial feature extraction module, wherein the spatial feature extraction module comprises a ProbSparse self-attention mechanism, a GCN graph rolling layer and a residual connection normalization layer as shown in FIG. 3.
The spatial feature extraction module updates node features by adopting a mode of combining self-attention calculation and node distance coefficients, specifically, firstly, weighting and adding the half-sine distance between the nodes and the normalized self-attention system, finally, normalizing to form a relation matrix, and replacing the adjacent matrix in the graph convolution model with the relation matrix to update and calculate the node features.
The calculation process of the relation matrix A is as follows:
wherein E is a node distance matrix, lambda is a balance coefficient, and the element node distance weights E in the matrix ij The calculation formula of (2) is as follows:
wherein x is i ,x j Represents the longitude and latitude positions of two nodes, h (x i ,x j ) Is the half sine distance of node i, j.
S4, reentering the time feature extraction module and the space feature extraction module to obtain the output out1 of the encoder;
s5, encoding the traffic flow space-time information of the last 1 day according to the encoding mode of the step S1 to obtain in1;
s6, inputting the out1 and the in1 into a fusion module, performing self-attention calculation on the in1 once to obtain an in2, and performing self-attention calculation on the out1 and the in2 to obtain an output of the fusion module;
the fusion module calculates attention coefficients between the extra historical time step input of the decoder and the output result of the encoder to carry out weighted addition on the output result of the encoder, and the operation utilizes the newly added input information of the encoder to distribute different attention to the output result of the encoder so as to better fuse the extra input information with the output result of the encoder.
Step S7, performing space decoding on the output of the fusion module through a space decoding module, wherein the calculation flow of the space decoding module is the same as that of the space feature extraction module in the step S3;
and S8, the output of the space decoding module passes through an output layer consisting of a one-dimensional convolution layer and a full connection layer to obtain a traffic flow prediction result of each node of the target time step, and the optimal path of the mobile robot is selected from the shortest paths of the first several distances according to the minimum value of the traffic flow sum.
Example 2:
as shown in fig. 4, there is also provided a mobile robot travel path point prediction system, which is composed of two parts, namely an encoder and a decoder.
Wherein the encoder section comprises the following modules:
the coding module M1 is used for respectively coding historical traffic flow space-time data of nearly 7 days and traffic flow space-time data of nearly 1 day to obtain a first code and a second code;
the time feature extraction module M2 is used for extracting the feature of the time dimension of the input of the encoder, and comprises a ProbSparse self-attention mechanism, a one-dimensional convolution layer and a maximum pooling layer;
the spatial feature extraction module M3 is used for extracting the features of the spatial dimension of the input of the encoder, and comprises a ProbSparse self-attention mechanism and a GCN chart convolution layer;
the decoder section includes the following modules:
the fusion module M4 is used for inputting the encoder output and the second code into the fusion module, performing self-attention computation on the second code once to obtain an intermediate result, and performing self-attention computation on the encoder output and the intermediate result to obtain the output of the fusion module;
the space decoding module M5 is used for performing space decoding on the output of the fusion module, and the calculation flow of the space decoding module is the same as that of the space feature extraction module;
and the output module M6 is used for outputting the space decoding module to pass through an output layer formed by a one-dimensional convolution layer and a full connection layer to obtain the traffic flow prediction result of each node of the target time step, and selecting the optimal path for the mobile robot to travel from the shortest paths of the first several distances according to the minimum value of the traffic flow sum.
The foregoing description of the preferred embodiments of the invention is not intended to be limiting, but rather is intended to cover all modifications, equivalents, and alternatives falling within the spirit and principles of the invention.
Claims (9)
1. A method for predicting a travel path point of a mobile robot, the method comprising:
step one, encoding historical traffic flow space-time data of near N days to obtain a first code, wherein the first code is used as input of an encoder;
sequentially carrying out time dimension feature extraction and space dimension feature extraction on the input of the encoder through a time feature extraction module and a space feature module to obtain the output of the encoder, wherein the time feature extraction module comprises a ProbSparse self-attention mechanism, a one-dimensional convolution layer and a maximum pooling layer; the spatial feature extraction module comprises a ProbSparse self-attention mechanism and a GCN graph roll-up layer;
thirdly, encoding traffic flow space-time information of the latest M (M < N) days according to the encoding mode of the first step to obtain a second encoding;
inputting the output of the encoder and the second code into a fusion module, performing self-attention calculation on the second code for one time to obtain an intermediate result, and performing self-attention calculation on the output of the encoder and the intermediate result to obtain the output of the fusion module;
step five, performing space decoding on the output of the fusion module through a space decoding module, wherein the calculation flow of the space decoding module is the same as that of the space feature extraction module in the step two;
and step six, the output of the space decoding module passes through an output layer consisting of a one-dimensional convolution layer and a full connection layer to obtain a traffic flow prediction result of each node of the target time step, and the optimal path for the mobile robot to travel is selected from the shortest paths of the previous distances according to the minimum value of the traffic flow sum.
2. The method for predicting the traveling path point of a mobile robot according to claim 1, wherein the probspark self-attention mechanism measures the similarity between the self-attention distribution and the uniform distribution by using KL divergence, and filters dot product pair distribution with low similarity, and the dot product pairs close to the uniform distribution are directly and evenly added, and the calculation formula of the probspark self-attention mechanism is as follows:
Q＝W q X，K＝W k X，V＝W v X
wherein X is input, d is feature vector length,is a sparse matrix of the same size as Q and contains only Q of the first u highest M values, where u=clnl is set Q C is a super parameter, and the score M is as follows:
3. the method for predicting a traveling path point of a mobile robot according to claim 2, wherein in the spatial feature extraction module, a relation matrix is formed by weighting and adding half-sine distances among nodes and a normalized self-attention system and then normalizing, the relation matrix is used for updating and calculating node features instead of an adjacent matrix in a GCN graph convolution model, and the calculation process of the relation matrix a is as follows:
wherein E is a node distance matrix, lambda is a balance coefficient, and the element node distance weights E in the matrix ij The calculation formula of (2) is as follows:
wherein x is i ,x j Represents the longitude and latitude positions of two nodes, h (x i ,x j ) Is the half sine distance of node i, j.
4. The method according to claim 1, further comprising repeating the temporal feature extraction and the spatial feature extraction once in the second step.
5. The method according to claim 1, wherein the codes include projection codes, position codes, and time codes, and the first code is formed by adding three codes after linear transformation to obtain the same matrix size.
6. The method according to claim 5, wherein the time code is obtained from 5 calculated values of a minute-hour ratio, an hour-day ratio, a day-week ratio, a day-month ratio, and a day-year ratio.
7. The method for predicting a traveling path point of a mobile robot according to claim 5, wherein the projection code is obtained by one-dimensional convolution.
8. The method for predicting a traveling path point of a mobile robot according to claim 5, wherein the position code is obtained by sine and cosine coding.
9. A mobile robot travel path point prediction system, comprising:
the coding module is used for respectively coding historical traffic flow space-time data of near N days and traffic flow space-time data of near M (M < N) days to obtain a first code and a second code;
the device comprises a characteristic extraction module, a characteristic extraction module and a characteristic processing module, wherein the characteristic extraction module is used for extracting the characteristics of the time dimension and the characteristics of the space dimension of the input of an encoder to obtain the output of the encoder, and comprises a temporal characteristic extraction module and a spatial characteristic extraction module, and the temporal characteristic extraction module comprises a ProbSparse self-attention mechanism, a one-dimensional convolution layer and a maximum pooling layer; the spatial feature extraction module comprises a ProbSparse self-attention mechanism and a GCN graph roll-up layer;
the fusion module is used for inputting the encoder output and the second code into the fusion module, performing self-attention calculation on the second code once to obtain an intermediate result, and performing self-attention calculation on the encoder output and the intermediate result to obtain a fusion module output;
the space decoding module is used for performing space decoding on the output of the fusion module, and the calculation flow of the space decoding module is the same as that of the space feature extraction module;
and the output module is used for outputting the space decoding module to pass through an output layer formed by a one-dimensional convolution layer and a full connection layer to obtain the traffic flow prediction result of each node of the target time step, and selecting the optimal path for the mobile robot to travel from the shortest paths of the first several distances according to the minimum value of the traffic flow sum.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310605169.9A CN116673947A (en) | 2023-05-26 | 2023-05-26 | Mobile robot travel path point prediction method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310605169.9A CN116673947A (en) | 2023-05-26 | 2023-05-26 | Mobile robot travel path point prediction method |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN116673947A true CN116673947A (en) | 2023-09-01 |
Family
ID=87788298
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202310605169.9A Pending CN116673947A (en) | 2023-05-26 | 2023-05-26 | Mobile robot travel path point prediction method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN116673947A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117096875A (en) * | 2023-10-19 | 2023-11-21 | 国网江西省电力有限公司经济技术研究院 | Short-term load prediction method and system based on ST-transducer model |
-
2023
- 2023-05-26 CN CN202310605169.9A patent/CN116673947A/en active Pending
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117096875A (en) * | 2023-10-19 | 2023-11-21 | 国网江西省电力有限公司经济技术研究院 | Short-term load prediction method and system based on ST-transducer model |
| CN117096875B (en) * | 2023-10-19 | 2024-03-12 | 国网江西省电力有限公司经济技术研究院 | Short-term load prediction method and system based on spatial-Temporal Transformer model |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109887282B (en) | Road network traffic flow prediction method based on hierarchical timing diagram convolutional network | |
| CN111400620A (en) | User trajectory position prediction method based on space-time embedded Self-orientation | |
| CN111832814A (en) | Air pollutant concentration prediction method based on graph attention machine mechanism | |
| CN112489426B (en) | Urban traffic flow space-time prediction scheme based on graph convolution neural network | |
| CN113487061A (en) | Long-time-sequence traffic flow prediction method based on graph convolution-Informer model | |
| CN107180530A (en) | A kind of road network trend prediction method based on depth space-time convolution loop network | |
| CN111401233A (en) | Trajectory prediction method, apparatus, electronic device, and medium | |
| CN113222328A (en) | Air quality monitoring equipment point arrangement and site selection method based on road section pollution similarity | |
| CN110083125B (en) | Machine tool thermal error modeling method based on deep learning | |
| CN111639791A (en) | Traffic flow prediction method, system, storage medium and terminal | |
| CN111554118B (en) | Dynamic prediction method and system for bus arrival time | |
| CN111724586B (en) | Commuting time prediction method, and training method and device of commuting time prediction model | |
| CN116673947A (en) | Mobile robot travel path point prediction method | |
| CN112905379B (en) | Traffic big data restoration method of graph self-encoder based on self-attention mechanism | |
| Lin et al. | Clustering-learning-based long-term predictive localization in 5G-envisioned Internet of connected vehicles | |
| CN116612645A (en) | Expressway service area vehicle flow prediction method | |
| CN114495500B (en) | Traffic prediction method based on dual dynamic space-time diagram convolution | |
| CN110401978A (en) | Indoor orientation method based on neural network and particle filter multi-source fusion | |
| CN111340543A (en) | Fine-grained taxi taking demand prediction method based on attention mechanism | |
| CN111461426A (en) | High-precision travel time length prediction method based on deep learning | |
| CN117153294B (en) | Molecular generation method of single system | |
| CN115049009A (en) | Track prediction method based on semantic fusion representation | |
| CN114462664A (en) | Short-range branch flight scheduling method integrating deep reinforcement learning and genetic algorithm | |
| CN114969234A (en) | Vector characterization method for full-quantity city road network | |
| CN113408786B (en) | Traffic characteristic prediction method and system |
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 |