CN113591966A - Multi-unmanned system task decomposition method combining graph processing and logical description - Google Patents
Multi-unmanned system task decomposition method combining graph processing and logical description Download PDFInfo
- Publication number
- CN113591966A CN113591966A CN202110850917.0A CN202110850917A CN113591966A CN 113591966 A CN113591966 A CN 113591966A CN 202110850917 A CN202110850917 A CN 202110850917A CN 113591966 A CN113591966 A CN 113591966A
- Authority
- CN
- China
- Prior art keywords
- task
- entities
- unmanned
- entity
- dimensional map
- 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 41
- 238000000354 decomposition reaction Methods 0.000 title claims abstract description 21
- 238000012545 processing Methods 0.000 title claims abstract description 14
- 230000009471 action Effects 0.000 claims abstract description 34
- 238000010586 diagram Methods 0.000 claims abstract description 20
- 238000004891 communication Methods 0.000 claims abstract description 5
- 238000013527 convolutional neural network Methods 0.000 claims description 8
- 230000008569 process Effects 0.000 claims description 8
- 238000001514 detection method Methods 0.000 claims description 5
- 230000007717 exclusion Effects 0.000 claims description 3
- 230000006870 function Effects 0.000 claims description 3
- 238000012417 linear regression Methods 0.000 claims description 3
- 238000005192 partition Methods 0.000 claims description 3
- 230000006872 improvement Effects 0.000 description 9
- 238000005516 engineering process Methods 0.000 description 2
- 239000011159 matrix material Substances 0.000 description 2
- 230000002159 abnormal effect Effects 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000002708 enhancing effect Effects 0.000 description 1
- 238000004519 manufacturing process Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000007781 pre-processing Methods 0.000 description 1
- 238000011160 research Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/23—Clustering techniques
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/10—File systems; File servers
- G06F16/11—File system administration, e.g. details of archiving or snapshots
- G06F16/116—Details of conversion of file system types or formats
-
- 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
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T17/00—Three dimensional [3D] modelling, e.g. data description of 3D objects
- G06T17/05—Geographic models
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- General Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- Software Systems (AREA)
- Geometry (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- Evolutionary Computation (AREA)
- Computational Linguistics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Molecular Biology (AREA)
- General Health & Medical Sciences (AREA)
- Mathematical Physics (AREA)
- Biophysics (AREA)
- Biomedical Technology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Computing Systems (AREA)
- Evolutionary Biology (AREA)
- Health & Medical Sciences (AREA)
- Remote Sensing (AREA)
- Computer Graphics (AREA)
- Databases & Information Systems (AREA)
- Control Of Position, Course, Altitude, Or Attitude Of Moving Bodies (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
A multi-unmanned system task decomposition method described in connection with graph processing and logistics, comprising: step S1: space division of the three-dimensional map; step S2: identifying entity three-dimensional coordinates of the three-dimensional map, and adding undirected connectivity among entities to obtain a scene space path logical connection diagram; step S3: calculating and storing the actual distance between the entities, and judging and storing the communication capacity between the entities; step S4: adding an attachment attribute to an entity or an area in the three-dimensional map, adding an attachment attribute to the explored path, and generating a current environment to instantiate and store; step S5: after receiving the current task, the unmanned system generates a prob l em file to be solved by combining with the stored environment instantiation; step S6: generating a domain file of a predefined domain library; step S7: and calculating an adjusting action sequence in real time until the task is completed. The invention has the advantages of better flexibility, universality, higher intelligence and the like.
Description
Technical Field
The invention mainly relates to the technical field of intelligent unmanned equipment control, in particular to a multi-unmanned system task decomposition method combining graph processing and logical description.
Background
With the development of science and technology, more and more intelligent unmanned equipment is applied to various industries. Numerous intelligent unmanned devices (unmanned systems), such as drones, unmanned vehicles, and the like. Most of tasks of the existing unmanned system are to fix the execution sequence of the tasks or to guide the execution of the tasks in an artificial supervision mode, and the tasks are widely used in the fields of factory production, warehouse storage, satellite scheduling and the like, but the task execution method has low fault tolerance and cannot process the tasks in time when abnormity occurs.
In the context of complex tasks, the capability of a single unmanned system is far from meeting the task requirements, so that the cooperative execution of tasks by multiple unmanned systems becomes a focus of attention in the industry. Because the units in the multi-unmanned system have obvious difference, the traditional modeling and planning method is difficult to obtain complete and uniform expression.
Disclosure of Invention
The technical problem to be solved by the invention is as follows: aiming at the technical problems in the prior art, the invention provides a multi-unmanned system task decomposition method which has better flexibility, universality and higher intelligence and is described in combination with graph processing and logicalization.
In order to solve the technical problems, the invention adopts the following technical scheme:
a multi-unmanned system task decomposition method described in connection with graph processing and logistics, comprising:
step S1: space division of the three-dimensional map;
step S2: identifying entity three-dimensional coordinates of the three-dimensional map, and adding undirected connectivity among entities to obtain a scene space path logical connection diagram;
step S3: calculating and storing the actual distance between the entities, and judging and storing the communication capacity between the entities;
step S4: adding an attachment attribute to an entity or an area in the three-dimensional map, adding an attachment attribute to the explored path, and generating a current environment to instantiate and store;
step S5: after receiving the current task, the unmanned system generates a problemm file to be solved by combining with stored environment instantiation and utilizing a PDDL (Planning Domain Definition Language) logical Language;
step S6: generating a domain file of a predefined domain library;
step S7: and calculating and adjusting action sequences in real time according to the stored files until the task is completed.
As a further improvement of the invention: the step S1 includes:
step S101: using each pixel as a group, then calculating the texture of each group, and combining the two closest groups; grouping smaller groups, and continuing to combine the regions until all the regions are combined together;
step S102: the R-CNN utilizes a candidate region method to create interested regions, the interested regions are converted into images with fixed sizes and are respectively sent to a convolutional neural network;
step S103: and classifying the regions by using a clustering algorithm, and correcting the bounding box by using linear regression loss to realize target classification and obtain the bounding box.
As a further improvement of the invention: the clustering algorithm adopts a bottom-up hierarchical clustering method, namely: each data point is first treated as a single cluster and then the distances between all clusters are calculated to merge the clusters until all clusters are aggregated into one cluster.
As a further improvement of the invention: in step S4, an attachment attribute is added to an entity or an area in the three-dimensional map to obtain an entity or area attachment attribute, an attachment attribute is added to the detected path to obtain a path attachment attribute, and a current environment is generated and instantiated and stored.
As a further improvement of the invention: the entity or region attachment attribute comprises one or more of the property of the entity, the state of the partition region and the area of the region; the path adding attachment attribute comprises one or more of the length of the road, the passability of the road and the concealment degree of the road.
As a further improvement of the invention: the to-be-solved problem file comprises object information, an object initial state and a target; the object information comprises all available unmanned systems in the task process; the object initial state refers to one or more of the position, speed, oil quantity, electric quantity and damage condition of all currently available unmanned systems; the goal is a set of desired states that consists of a logically-transformed predicate table and a constraint index.
As a further improvement of the invention: in step S6, the capabilities and actions of the unmanned systems of different categories are expressed in a system logical form, and converted into a structure required by the PDDL format, thereby generating domain files of the domain library.
As a further improvement of the invention: the domain file of the predefined domain library comprises: predicates, functions, constants, actions; after receiving the task target, the pre-defined domain library decomposes the task to be executed based on an LPG-td algorithm to obtain a global execution action sequence. After receiving a task target, the predefined domain library decomposes the task to be executed based on an LPG-td algorithm (Local Search in Planning Graph, Local Search of a Planning Graph) to obtain a global execution action sequence; the LPG defines a data structure of the motion diagram on the basis of the planning diagram, and the search space of the LPG is the space formed by the motion diagram. In the search, the LPG-td randomly generates an action map first, and then modifies the action map so that it is converted into another state. When the unsupported preconditions and mutual exclusion actions no longer exist in the action graph, then the planning solution is said to be found.
As a further improvement of the invention: in the step S1, in an outdoor complex scene, the external environment is sensed through the satellite map and other sensors, the target detection algorithm is used to determine the related entities in the space, and then the clustering algorithm is used to complete the spatial division of the three-dimensional map for the geographic position entity.
As a further improvement of the invention: further comprising step S8: dynamically updating the stored information; the dynamic updating storage information is used for assisting the generation of a next problem file and the task decomposition.
Compared with the prior art, the invention has the advantages that:
1. the invention provides a new planning method based on a domain definition library, which is the leading-edge application of the domain at present. The invention stores the capability, action and other data of all unmanned systems in the PDDL pre-domain defined domain, and when a specific task is decomposed, the task sequence is rapidly searched in the defined domain according to the input of real-time scene conditions. When an effective result cannot be obtained, the inference technology is actively utilized to learn a new environment and the definition domain is updated under the supervision of people, so that the system has good flexibility and universality.
2. The invention adopts image preprocessing aiming at the three-dimensional map, increases the communication relation between the entities, calculates the distance between the entities and judges the communication condition between the entities, thereby effectively increasing the accuracy of subsequent solving; the invention further attaches attributes to the regions or the entities to adapt to different scenes, and is more fit with the real situation, so that the method has stronger performability of the solving result, and can actively update the domain library when the task sequence cannot be decomposed, thereby enhancing the universality and the intelligence of the method. Therefore, the method can realize the collaborative task decomposition of the multiple unmanned systems under the dynamic complex scene, and has good scientific research and engineering values.
Drawings
FIG. 1 is a schematic flow diagram of the process of the present invention.
Fig. 2 is a schematic diagram of search and rescue area division and road connectivity in a specific application example of the present invention.
Fig. 3 is a schematic diagram of the planning result in the specific application example of the present invention.
Detailed Description
The invention will be described in further detail below with reference to the drawings and specific examples.
As shown in fig. 1, the multi-unmanned system task decomposition method described in conjunction with graph processing and logic of the present invention includes:
step S1: space division of the three-dimensional map;
step S2: identifying entity three-dimensional coordinates of the three-dimensional map in the step S1, and adding undirected connectivity among entities according to actual conditions to obtain a scene space path logical connection diagram;
step S3: calculating the actual distance between the entities according to the three-dimensional coordinates of the entities and the scene space path logical connection diagram, and storing the actual distance in a space distance matrix; judging the visibility between the entities, and storing the visibility in a space visibility matrix;
step S4: with the above steps combined, adding an attachment attribute to an entity or an area in the three-dimensional map to obtain an entity or an area attachment attribute, adding an attachment attribute to the detected path to obtain a path attachment attribute, generating a current environment, instantiating and storing the current environment in a Remote Dictionary service (Redis);
step S5: after receiving a current task, the unmanned system updates the current state information of the unmanned system, generates a problemm file to be solved by using a PDDL logical language in combination with stored environment instantiation, wherein the problem file to be solved comprises object information, an object initial state and a target;
step S6: generating a predefined domain library file;
step S7: calculating and adjusting action sequences in real time according to the stored files until the task is completed; in other words, in the process of executing the task by the unmanned system, according to the actual change situation, the environment information, the entity position, the state of the unmanned system and other information stored in the Redis are dynamically updated, and the action sequence is calculated and adjusted in real time until the task is completed;
as a preferred embodiment, the present invention further includes step S8: dynamically updating the stored information; the dynamic updating storage information is used for assisting the generation of a next problem file and the task decomposition.
As can be seen from the above, the multi-unmanned system task decomposition method described in combination with graph processing and logic according to the present invention is to perform multi-unmanned system task decomposition by using the PDDL logical representation method of graph processing to obtain an action sequence, and perform supervised domain updating on abnormal planning conditions according to a real-time scene without setting in advance.
In a specific application example, in step S1, in an outdoor complex scene, the present invention senses an external environment through a satellite map and other sensors, determines related entities in a space by using a target detection algorithm, and then completes spatial division of a three-dimensional map for a geographic position entity by using a clustering algorithm.
In a specific application example, according to the actual requirement, the target detection method may adopt an R-CNN algorithm, that is, a candidate region is used to obtain a region of interest.
In a specific application example, the flow of the R-CNN algorithm includes:
step S101: using each pixel as a group, then calculating the texture of each group, and combining the two closest groups; in the process, as a preferable scheme, in order to avoid the single region from phagocytizing other regions, the invention firstly groups smaller groups and continues to combine the regions until all the regions are combined together;
step S102: the R-CNN creates regions of interest (e.g., about 2000 regions of interest in this example) using the candidate region method, which are converted into fixed size images and sent to the convolutional neural network, respectively;
step S103: regions are classified using a clustering algorithm (SVM) and bounding boxes are corrected using linear regression loss to achieve target classification and obtain bounding boxes.
In the process, as a preferred scheme, the clustering algorithm adopts a bottom-up hierarchical clustering method, namely: each data point is first treated as a single cluster and then the distances between all clusters are calculated to merge the clusters until all clusters are aggregated into one cluster.
In a specific application example, in step S4, the entity or region attachment attribute includes one or more of a property of an entity, a state of a partition, and a region area.
In a specific application example, in the step S4, the path adding attachment attribute includes one or more of a length of a road, a passability of a road, and a degree of concealment of a road.
In a specific application example, in the step S5, the object information includes all available unmanned systems in the task process; the initial state of the object refers to the position, speed, oil quantity, electric quantity, damage condition and the like of all currently available unmanned systems; the goal is a set of desired states that consists of a logically-transformed predicate table and a constraint index.
In a specific application example, in step S6, the method performs system logical expression on the capabilities and actions of different types of unmanned systems, converts the capabilities and actions into a structure required by a PDDL format, and generates a domain file of a predefined domain library; the predefined domain library file comprises: predicates, functions, constants, actions; after receiving a task target, the predefined domain library decomposes the task to be executed based on an LPG-td algorithm (Local Search in Planning Graph, Local Search of a Planning Graph) to obtain a global execution action sequence; the LPG defines a data structure of the motion diagram on the basis of the planning diagram, and the search space of the LPG is the space formed by the motion diagram. In the search, the LPG-td randomly generates an action map first, and then modifies the action map so that it is converted into another state. When the unsupported preconditions and mutual exclusion actions no longer exist in the action graph, then the planning solution is said to be found.
The specific application embodiment is as follows: take unmanned aerial vehicle, unmanned vehicle cooperate ground squad to carry out the emergent search and rescue task after the calamity as an example. The local area input search and rescue force comprises 1 ground team which is provided with a plurality of small search and rescue detection unmanned aerial vehicles and light unmanned vehicles.
Through the report of the superior situation, it is found that E, F areas call for help at two places, one place in the high land of B area and one place in the critical house to be searched in C area. The initial time of search and rescue is 0000. When the search and rescue time is set to be 0100, the area A suddenly appears at one place where the wounded person to be rescued appears.
The search and rescue area division and road connectivity are shown in fig. 2. Area D is a safe area, area B is an uncertain area, and area A, C, E, F is a dangerous area. Road trafficability has been labeled on the graph. The road length and the moving speed of each search and rescue unit are subject to actual data.
After receiving the current task, the unmanned system generates a to-be-solved problemm file by using PDDL in combination with stored environment instantiation, calculates and adjusts a search and rescue action sequence in real time according to the stored file based on a domain file of a predefined domain library until the task is completed. The planning results are shown in fig. 3.
The above is only a preferred embodiment of the present invention, and the protection scope of the present invention is not limited to the above-mentioned embodiments, and all technical solutions belonging to the idea of the present invention belong to the protection scope of the present invention. It should be noted that modifications and embellishments within the scope of the invention may be made by those skilled in the art without departing from the principle of the invention.
Claims (10)
1. A method for task decomposition of a multi-unmanned system described in connection with graph processing and logistics, comprising:
step S1: space division of the three-dimensional map;
step S2: identifying entity three-dimensional coordinates of the three-dimensional map, and adding undirected connectivity among entities to obtain a scene space path logical connection diagram;
step S3: calculating and storing the actual distance between the entities, and judging and storing the communication capacity between the entities;
step S4: adding an attachment attribute to an entity or an area in the three-dimensional map, adding an attachment attribute to the explored path, and generating a current environment to instantiate and store;
step S5: after receiving the current task, the unmanned system generates a proplem file to be solved by using a PDDL logical language in combination with stored environment instantiation;
step S6: generating a domain file of a predefined domain library;
step S7: and calculating and adjusting action sequences in real time according to the stored files until the task is completed.
2. The method for task decomposition of multiple unmanned systems as claimed in claim 1, wherein the step S1 comprises:
step S101: using each pixel as a group, then calculating the texture of each group, and combining the two closest groups; grouping smaller groups, and continuing to combine the regions until all the regions are combined together;
step S102: the R-CNN utilizes a candidate region method to create interested regions, the interested regions are converted into images with fixed sizes and are respectively sent to a convolutional neural network;
step S103: and classifying the regions by using a clustering algorithm, and correcting the bounding box by using linear regression loss to realize target classification and obtain the bounding box.
3. The method for decomposing tasks of a multi-unmanned system described in conjunction with graph processing and logistics according to claim 2, wherein the clustering algorithm employs a bottom-up hierarchical clustering method, namely: each data point is first treated as a single cluster and then the distances between all clusters are calculated to merge the clusters until all clusters are aggregated into one cluster.
4. The method for task decomposition of multiple unmanned systems as claimed in claim 1, 2 or 3, wherein the step S4 comprises adding attachment attributes to entities or areas in the three-dimensional map to obtain entity or area attachment attributes, adding attachment attributes to the detected path to obtain path attachment attributes, and generating and storing the current environment instantiation.
5. The multi-unmanned-system task decomposition method, described in conjunction with graph processing and logistics, of claim 4, wherein the entity or region attachment attributes include one or more of a property of an entity, a state of a partition, a region area; the path adding attachment attribute comprises one or more of the length of the road, the passability of the road and the concealment degree of the road.
6. The method for task decomposition of multiple unmanned systems described in conjunction with graph processing and logic according to claim 1, 2 or 3, wherein the solution file contains object information, object initial state and object; the object information comprises all available unmanned systems in the task process; the object initial state refers to one or more of the position, speed, oil quantity, electric quantity and damage condition of all currently available unmanned systems; the goal is a set of desired states that consists of a logically-transformed predicate table and a constraint index.
7. The method for task decomposition of multiple unmanned systems according to claim 1, 2 or 3, wherein in step S6, the capabilities and actions of different classes of unmanned systems are expressed in a system logic manner and converted into the structure required by PDDL format, and a domain file of a predefined domain library is generated.
8. The method of multi-unmanned-system task decomposition described in connection with graph processing and logicalization according to claim 7, wherein said predefined domain library domain file contains: predicates, functions, constants, actions; after receiving the task target, the predefined domain library decomposes the task to be executed based on an LPG-td algorithm to obtain a global execution action sequence; the LPG defines a data structure of the action diagram on the basis of the planning diagram, and the search space of the LPG is the space formed by the action diagram; during searching, the LPG-td randomly generates an action diagram firstly, and then modifies the action diagram so as to convert the action diagram into another state; when the unsupported preconditions and mutual exclusion actions no longer exist in the action graph, then the planning solution is said to be found.
9. The method for decomposing tasks of multiple unmanned systems according to claim 1, 2 or 3, wherein in the step S1, in the complex outdoor scene, the external environment is sensed through the satellite map and other sensors, the related entities in the space are determined by using the target detection algorithm, and then the spatial division of the three-dimensional map is performed for the geographic position entities by using the clustering algorithm.
10. The method for task decomposition of multiple unmanned systems as claimed in claim 1, 2 or 3, wherein the method further comprises step S8: dynamically updating the stored information; the dynamic updating storage information is used for assisting the generation of a next problem file and the task decomposition.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110850917.0A CN113591966B (en) | 2021-07-27 | 2021-07-27 | Multi-unmanned system task decomposition method combining graph processing and logic description |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110850917.0A CN113591966B (en) | 2021-07-27 | 2021-07-27 | Multi-unmanned system task decomposition method combining graph processing and logic description |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN113591966A true CN113591966A (en) | 2021-11-02 |
| CN113591966B CN113591966B (en) | 2024-06-21 |
Family
ID=78250473
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202110850917.0A Active CN113591966B (en) | 2021-07-27 | 2021-07-27 | Multi-unmanned system task decomposition method combining graph processing and logic description |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113591966B (en) |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20160068268A1 (en) * | 2013-05-02 | 2016-03-10 | Bae Systems Plc | Goal-based planning system |
| CN107924190A (en) * | 2015-06-23 | 2018-04-17 | 阿肯技术公司 | For being supported by recharging station with the system of the autonomous operation of multiple mixing unmanned vehicles of execution service |
| CN112907015A (en) * | 2019-10-15 | 2021-06-04 | 中国人民解放军92578部队 | UUV self-adaptive task planning method under uncertain condition |
-
2021
- 2021-07-27 CN CN202110850917.0A patent/CN113591966B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20160068268A1 (en) * | 2013-05-02 | 2016-03-10 | Bae Systems Plc | Goal-based planning system |
| CN107924190A (en) * | 2015-06-23 | 2018-04-17 | 阿肯技术公司 | For being supported by recharging station with the system of the autonomous operation of multiple mixing unmanned vehicles of execution service |
| US20180188747A1 (en) * | 2015-06-23 | 2018-07-05 | Davide VENTURELLI | System for autonomous operation of multiple hybrid unmanned aerial vehicles supported by recharging stations to perform services |
| CN112907015A (en) * | 2019-10-15 | 2021-06-04 | 中国人民解放军92578部队 | UUV self-adaptive task planning method under uncertain condition |
Non-Patent Citations (1)
| Title |
|---|
| 易小云: "机器人任务重规划机制的研究与实现", 中国优秀硕士学位论文全文数据库 信息科技辑 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN113591966B (en) | 2024-06-21 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Tang et al. | Perception and navigation in autonomous systems in the era of learning: A survey | |
| JP2023156308A (en) | Systems and methods to enable continual, memory-bounded learning in artificial intelligence and deep learning continuously operating applications across networked computing edges | |
| Mokssit et al. | Deep learning techniques for visual slam: A survey | |
| CN110363816A (en) | A kind of mobile robot environment semanteme based on deep learning builds drawing method | |
| US20220301173A1 (en) | Method and system for graph-based panoptic segmentation | |
| Lee et al. | Deep AI military staff: Cooperative battlefield situation awareness for commander’s decision making | |
| CN114600130A (en) | Process for learning new image classes without labels | |
| US20230134531A1 (en) | Method and system for rapid retrieval of target images based on artificial intelligence | |
| Xu et al. | Vision-based multi-level synthetical evaluation of seismic damage for RC structural components: a multi-task learning approach | |
| Pedro et al. | Colanet: A uav collision avoidance dataset | |
| CN110275919A (en) | Data integrating method and device | |
| Sun et al. | RobNet: real-time road-object 3D point cloud segmentation based on SqueezeNet and cyclic CRF | |
| Li et al. | Automated deep learning system for power line inspection image analysis and processing: Architecture and design issues | |
| CN114782752A (en) | Small sample image grouping classification method and device based on self-training | |
| Ouyang et al. | Semantic slam for mobile robot with human-in-the-loop | |
| CN113591966A (en) | Multi-unmanned system task decomposition method combining graph processing and logical description | |
| Li et al. | A new data fusion framework of business intelligence and analytics in economy, finance and management | |
| Liu et al. | Generating Pixel Enhancement for Road Extraction in High-Resolution Aerial Images | |
| CN111496784A (en) | Space environment identification method and system for robot intelligent service | |
| US20230244985A1 (en) | Optimized active learning using integer programming | |
| Nazar et al. | ENWAR: A RAG-empowered Multi-Modal LLM Framework for Wireless Environment Perception | |
| Yeap et al. | Performance Comparison between Deep Learning and Machine Learning Algorithms in Vehicle Detection | |
| CN114241290B (en) | Indoor scene understanding method, equipment, medium and robot for edge calculation | |
| Buffum et al. | Autonomous drone behavior via MCDM of UFOMap layers | |
| CN117951310A (en) | Traffic scene risk assessment method, device, equipment and medium |
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 |