CN116744368A - Intelligent collaborative heterogeneous air-ground unmanned system based on cloud side end architecture and implementation method - Google Patents
Intelligent collaborative heterogeneous air-ground unmanned system based on cloud side end architecture and implementation method Download PDFInfo
- Publication number
- CN116744368A CN116744368A CN202310800372.1A CN202310800372A CN116744368A CN 116744368 A CN116744368 A CN 116744368A CN 202310800372 A CN202310800372 A CN 202310800372A CN 116744368 A CN116744368 A CN 116744368A
- Authority
- CN
- China
- Prior art keywords
- data
- cloud
- task
- edge computing
- edge
- 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 33
- 238000004364 calculation method Methods 0.000 claims abstract description 27
- 238000013507 mapping Methods 0.000 claims abstract description 24
- 230000005012 migration Effects 0.000 claims abstract description 8
- 238000013508 migration Methods 0.000 claims abstract description 8
- 230000007613 environmental effect Effects 0.000 claims abstract description 6
- 230000008447 perception Effects 0.000 claims abstract description 6
- 238000013500 data storage Methods 0.000 claims description 26
- 238000007781 pre-processing Methods 0.000 claims description 22
- 238000004891 communication Methods 0.000 claims description 17
- 230000003993 interaction Effects 0.000 claims description 16
- 238000012549 training Methods 0.000 claims description 16
- 230000006870 function Effects 0.000 claims description 15
- 238000007726 management method Methods 0.000 claims description 11
- 230000008569 process Effects 0.000 claims description 11
- 230000005540 biological transmission Effects 0.000 claims description 6
- 238000006243 chemical reaction Methods 0.000 claims description 6
- 230000007246 mechanism Effects 0.000 claims description 6
- 238000005457 optimization Methods 0.000 claims description 6
- 238000013480 data collection Methods 0.000 claims description 4
- 230000004927 fusion Effects 0.000 claims description 3
- 229920003087 methylethyl cellulose Polymers 0.000 claims 2
- 238000013528 artificial neural network Methods 0.000 description 9
- 230000033001 locomotion Effects 0.000 description 5
- 230000010354 integration Effects 0.000 description 4
- 238000000605 extraction Methods 0.000 description 3
- 208000018777 Vulvar intraepithelial neoplasia Diseases 0.000 description 2
- 230000015572 biosynthetic process Effects 0.000 description 2
- 230000008859 change Effects 0.000 description 2
- 230000003247 decreasing effect Effects 0.000 description 2
- 230000007547 defect Effects 0.000 description 2
- 238000004458 analytical method Methods 0.000 description 1
- 230000000712 assembly Effects 0.000 description 1
- 238000000429 assembly Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 239000000306 component Substances 0.000 description 1
- 239000002131 composite material Substances 0.000 description 1
- 239000008358 core component Substances 0.000 description 1
- 230000008878 coupling Effects 0.000 description 1
- 238000010168 coupling process Methods 0.000 description 1
- 238000005859 coupling reaction Methods 0.000 description 1
- 230000001419 dependent effect Effects 0.000 description 1
- 238000005265 energy consumption Methods 0.000 description 1
- 230000002068 genetic effect Effects 0.000 description 1
- 238000011478 gradient descent method Methods 0.000 description 1
- 238000003064 k means clustering Methods 0.000 description 1
- 238000012423 maintenance Methods 0.000 description 1
- 238000004519 manufacturing process Methods 0.000 description 1
- 239000011159 matrix material Substances 0.000 description 1
- 239000002071 nanotube Substances 0.000 description 1
- 239000002245 particle Substances 0.000 description 1
- 238000012545 processing Methods 0.000 description 1
- 238000000926 separation method Methods 0.000 description 1
- 230000001360 synchronised effect Effects 0.000 description 1
- 238000012546 transfer Methods 0.000 description 1
- 230000009466 transformation Effects 0.000 description 1
- 230000000007 visual effect Effects 0.000 description 1
Classifications
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04W—WIRELESS COMMUNICATION NETWORKS
- H04W28/00—Network traffic management; Network resource management
- H04W28/02—Traffic management, e.g. flow control or congestion control
- H04W28/08—Load balancing or load distribution
- H04W28/09—Management thereof
- H04W28/0958—Management thereof based on metrics or performance parameters
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04B—TRANSMISSION
- H04B7/00—Radio transmission systems, i.e. using radiation field
- H04B7/14—Relay systems
- H04B7/15—Active relay systems
- H04B7/185—Space-based or airborne stations; Stations for satellite systems
- H04B7/18502—Airborne stations
- H04B7/18506—Communications with or from aircraft, i.e. aeronautical mobile service
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L67/00—Network arrangements or protocols for supporting network services or applications
- H04L67/01—Protocols
- H04L67/10—Protocols in which an application is distributed across nodes in the network
- H04L67/1001—Protocols in which an application is distributed across nodes in the network for accessing one among a plurality of replicated servers
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L67/00—Network arrangements or protocols for supporting network services or applications
- H04L67/50—Network services
- H04L67/51—Discovery or management thereof, e.g. service location protocol [SLP] or web services
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L67/00—Network arrangements or protocols for supporting network services or applications
- H04L67/50—Network services
- H04L67/60—Scheduling or organising the servicing of application requests, e.g. requests for application data transmissions using the analysis and optimisation of the required network resources
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04W—WIRELESS COMMUNICATION NETWORKS
- H04W4/00—Services specially adapted for wireless communication networks; Facilities therefor
- H04W4/30—Services specially adapted for particular environments, situations or purposes
- H04W4/40—Services specially adapted for particular environments, situations or purposes for vehicles, e.g. vehicle-to-pedestrians [V2P]
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02D—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN INFORMATION AND COMMUNICATION TECHNOLOGIES [ICT], I.E. INFORMATION AND COMMUNICATION TECHNOLOGIES AIMING AT THE REDUCTION OF THEIR OWN ENERGY USE
- Y02D10/00—Energy efficient computing, e.g. low power processors, power management or thermal management
Abstract
The invention provides an intelligent collaborative heterogeneous air-ground unmanned system based on a cloud end architecture and an implementation method thereof, wherein the cloud end architecture is used as a basis, and tasks are deployed on a cloud server, edge computing equipment and an executing end intelligent body according to different calculation force requirements of the tasks; the cloud server is responsible for task scheduling and high-precision mapping work with the largest calculation force demand, and the edge computing equipment is responsible for general online repositioning of the calculation force demand and path planning; the execution end agent is responsible for track tracking control and perception of environmental information; cloud edge coordination and edge coordination are achieved between the cloud edge server and the edge computing equipment through task horizontal migration. The tasks borne by the cloud end, the edge and the executing end intelligent agent are independent, no repetition exists, the tasks are matched with each other, the running efficiency of the system is improved, the cost of hardware is reduced, and the adaptability in unfamiliar environments is improved.
Description
Technical Field
The invention provides an intelligent collaborative heterogeneous air-ground unmanned system based on a cloud side architecture and an implementation method thereof, and belongs to the technical field of unmanned aerial vehicles.
Background
The heterogeneous air-ground unmanned system is composed of a plurality of unmanned aerial vehicles and unmanned vehicles, and has the advantages of sensing, positioning, decision making, control and heterogeneous cross-domain collaborative unmanned systems. For the heterogeneous unmanned air space system, the intelligent operation depends on task scheduling of a decision layer, path planning of a planning layer and perception and control of an execution layer. The existing space unmanned system is characterized in that scheduling, planning, sensing and controlling are all centralized on individuals of an unmanned aerial vehicle or an unmanned aerial vehicle, sensing of environment is achieved through a single individual, each individual makes path planning on the unmanned aerial vehicle on the basis of sensing of environment, then tracking of paths is achieved through track tracking control to complete individual tasks, and in the process of executing the individual tasks, the unmanned aerial vehicle and the unmanned aerial vehicle are coordinated in a formation mode.
The existing unmanned air-ground system has a plurality of limitations, for example, the perceived information of the environment is not communicated with each other by individuals, so that the environment is perceived repeatedly, and the perceived efficiency of the environment is reduced; the functions of the individual are highly repeated, and the functions of planning, scheduling and the like which require huge calculation force are realized, so that the hardware cost of the unmanned aerial vehicle and the individual of the unmanned aerial vehicle are increased, the energy consumption is increased, and the endurance time is reduced; when an individual makes a path planning for the individual, the planning information of other individuals is not considered, so that collision is easy to cause; the coordination is realized by the tight coupling mode of formation, and global coordination is lacking, so that the unmanned aerial vehicle and the unmanned aerial vehicle cannot fully exert respective advantages, the coordination efficiency is reduced, and the like. The efficient operation of the heterogeneous unmanned air-ground system is greatly dependent on the efficient and organic combination of four links of decision making, planning, perception and control. It can be seen that the system taking the individual as the independent unit lacks scheduling and coordination of the system level, and the combination between the links is hard. The defects brought by the architecture are difficult to make up in algorithm, so that the existing system functions are split, the functions are respectively deployed at the cloud end, the edge and the execution end through different calculation force demands, and the scheduling and the coordination of the system level are realized through unified collection and processing of data, so that the operation efficiency of the heterogeneous unmanned air-ground system is improved.
CN115314850a proposes an intelligent motion system based on cloud-edge cooperative control, which consists of a central data cloud module, a wireless network transmission module, an edge computing device and an intelligent mobile device.
As can be seen from the implementation description of this technique, this solution has several drawbacks:
1. the cloud module, the edge equipment module and the intelligent mobile equipment all have instruction control functions, and the functions are cross redundant and can easily cause the repetition or coverage of instructions.
2. The intelligent mobile devices bear decision-making functions and are decision-making for individuals, and there is no system-level scheduling coordination between the intelligent mobile devices.
3. The intelligent mobile device has the advantages that a large number of functions requiring computing power in the system are borne, the hardware requirements of the mobile device are high, each mobile device individual needs to be configured with an independent system, and the cost for building the mobile device is high.
4. The whole system does not relate to the mapping perception work of the unknown environment, and has poor adaptability to the unfamiliar environment.
5. The system aims at controlling the movement of the individual under cloud edge coordination, and the method is not applicable to systems of a plurality of individuals.
6. The cloud module, the edge equipment module and the task execution position hosted by the intelligent mobile equipment are solidified, and can not be vertically unloaded between cloud edges and horizontally migrated between edges according to calculation force requirements.
7. The synergy between intelligent mobile devices is large in overall scale.
Disclosure of Invention
Aiming at the problems in the three prior heterogeneous air-ground unmanned control systems, the invention provides an intelligent collaborative heterogeneous air-ground unmanned system based on a cloud end architecture, which is based on the cloud end architecture, and the tasks are deployed in a cloud server, edge computing equipment and an execution end intelligent body according to different calculation force requirements of the tasks; the cloud server is responsible for task scheduling and high-precision mapping work with the largest calculation force demand, and the edge computing equipment is responsible for general online repositioning of the calculation force demand and path planning; the computational power requirements generally refer to image computing power that does not require GPU computing or Jetson TX2 can afford; the execution end agent is responsible for track tracking control and perception of environmental information; cloud edge coordination and edge coordination are achieved between the cloud edge server and the edge computing equipment through task horizontal migration.
Thus, the modularized separation of tasks is cooperated with the task scheduling of a system level. And meanwhile, the task of the execution end is minimized, and the manufacturing and maintenance cost of a single unmanned aerial vehicle and an unmanned aerial vehicle at the execution end is reduced as much as possible. Finally, aiming at the problem of poor adaptability to unfamiliar environments, the execution end intelligent agent realizes distributed collaborative mapping and positioning by taking a laser radar, an IMU and a camera as sensors, and improves the robustness and universality of the unmanned air-ground system in unfamiliar environments.
The specific technical scheme is as follows:
the cloud server consists of a data storage device A and an artificial intelligent big data calculation server. The cloud server firstly receives operation data, original point cloud data, image data and positioning data uploaded by the edge computing device through the firewall security facility through the data access module, and stores and backs up the operation data, the original point cloud data, the image data and the positioning data in the data storage device A. After the original data collection is completed, the scheduling service, the mapping service and the model training service respectively ask the data storage device A for corresponding operation data, point cloud data, image data and intelligent agent positioning data. The scheduling service completes cloud-edge workflow scheduling by analyzing the current intelligent agent running state, task demands and current residual computing power, realizes the task scheduling of the space unmanned system considering global system level, and finally realizes the task cooperation of the execution end to complete scheduling tasks; the map building service completes the reconstruction work of the global map through merging the multi-sensor information such as the point cloud data, the image data and the like, and continuously updates the local map through the map information collected in the operation process to obtain the millimeter-level error map information updated in real time; the model training service optimizes the current co-location model parameters through the errors of the current location information and the model predictive location information, minimizes the location errors and obtains the current optimal model parameters. And after the cloud service finishes the task, transmitting the information to the edge computing equipment. Under the condition that cloud computing power resources are tense and edge computing equipment is idle, part of data preprocessing tasks in the high-precision graph building task can be unloaded to the edge computing equipment to be executed, and the data preprocessing tasks are uploaded to a cloud server after the edge computing equipment completes the tasks.
The edge computing device is comprised of a mobile 5G communication base station and several sets of mobile edge computing devices (MECs). The mobile 5G base station is responsible for communication services of the mobile MEC and the cloud, and communication services of the MEC and the executing end agent, and low-delay transmission of large data volume is realized. In MEC, firstly, path planning application, online repositioning application, data fusion application and data interaction application are instantiated into Pod composed of a plurality of containers, and cloud edge coordination, automatic load balancing and offline autonomy of edges are realized through Kubernetes container management of KubEdge and cloud. In the container instantiated by the data interaction application, the MEC receives cloud data and execution-end agent data through the 5G base station, and meanwhile, format conversion and preprocessing are carried out on the execution-end agent data, and then the data are temporarily stored in a data storage device B which can be accessed by other containers. In the container instantiated by the path planning application, the container first accesses the task target and the current position information of a certain agent in the data storage device B, and obtains a feasible path of the certain agent through a path planning algorithm. In the container instantiated by the online repositioning application, the container firstly accesses point cloud data and image data currently obtained by a certain agent in the data storage device B, compares the point cloud data and the image data with millimeter-level error map data, and performs the co-positioning of multiple agents by combining a co-positioning model issued by a cloud, so as to update the current position and posture information of the agent and realize online repositioning. In the container instantiated by the data preprocessing application, the container first accesses the multi-agent map data and the running data in the data storage device B, synchronizes the data time stamps, and converts the data from the ros_message format to a generic data format. All the tasks can be equally realized in a plurality of groups of MECs, and when the MEC responsible for a certain task is too much in calculation load or can not be accessed due to hardware problems, the parallel migration of the task can be realized through a side-by-side coordination mechanism in the edge core.
The execution end intelligent agent consists of a plurality of heterogeneous unmanned aerial vehicles and unmanned vehicles. The unmanned aerial vehicle is provided with an IMU and a camera, acquires map information, and is provided with the IMU, a laser radar and the camera. And the unmanned aerial vehicle are both provided with a computing device with lower computing power and a 5G wireless communication module, so that the uploading of data and the calculation of the track tracking control quantity are realized.
The method for realizing the intelligent collaborative heterogeneous air-ground unmanned system based on the cloud end architecture performs tasks in unfamiliar environments and comprises the following steps:
step 1: the intelligent agent unmanned aerial vehicle at the execution end and the unmanned aerial vehicle head firstly acquire map data, and the map data is uploaded to the edge computing equipment after the acquisition is completed;
step 2: instantiating the data preprocessing task and the data interaction task into a container at the edge computing device, and performing timestamp synchronization and data format conversion on map data in the data preprocessing task container; uploading the processed data to a cloud server data access interface through a data interaction task container;
step 3: the cloud server stores and backs up the data through a data storage function after receiving the data uploaded by the edge computing device; then instantiating the mapping service into a cloud server container, and accessing map data by the container to realize high-precision mapping; and issuing the millimeter-level error map to the edge computing device;
step 4: the cloud server instantiates the scheduling service into a cloud server container, and firstly, the computing power occupation and task completion time of the current cloud server are evaluated; if the occupation is large or the task cannot be completed on time, unloading the part with smaller calculation amount in part of the tasks to edge computing equipment for carrying out; the larger cloud computing power resource accounts for 90% or more; the smaller refers to the image computing capability which does not need GPU computing or can be borne by Jetson TX 2; then, by accessing the operation data, the executing end agent is integrally scheduled, so that the task cooperation of the executing end agent is realized, and the scheduling information is issued to the edge computing equipment;
step 5: the edge computing device firstly instantiates a path planning task and an online repositioning task into an edge computing device container; the path planning task container makes path planning for the intelligent agent by accessing the millimeter-level error map and the scheduling information of the intelligent agent at the execution end, gives out a specific executable tracking path and issues the specific executable tracking path to the intelligent agent at the corresponding execution end;
step 6: after receiving the corresponding planning paths, the executing end agent applies a track tracking algorithm to track the corresponding paths and continuously uploads operation data to the edge computing equipment; the online repositioning task container in the edge computing equipment realizes online repositioning by accessing the millimeter-level error map and the operation data, and transmits positioning information to the corresponding execution end intelligent agent; repeating the steps until the execution end agent completes the issued dispatching task;
and 6, if the problem of network or hardware occurs in the repetition process, when the edge computing equipment cannot continue the task, horizontally transferring the task through a side-to-side cooperative mechanism, so as to realize uninterrupted execution of the task flow. The task refers to specific tasks distributed to the edge computing equipment, such as online repositioning and path planning tasks, and the task flow refers to the execution of all steps according to the sequence.
Step 7: the edge computing device uploads the positioning information to the cloud server, the cloud server instantiates the model training task into the container, the container completes optimization of the model by accessing the positioning information, and optimized parameters are issued to the edge computing device. The unfamiliar environment and the familiar environment are determined whether the map information acquisition and mapping work is performed on the environment, if the map is established, the unfamiliar environment is the familiar environment, otherwise, the unfamiliar environment is the unfamiliar environment.
After the strange environment completes the step 1-3, the strange environment is converted into a familiar environment. When the overall flow is executed again, the process starts in step 4.
The technical scheme of the invention has the beneficial effects that:
1. the cloud end architecture decouples scheduling, planning, sensing and control tasks involved in the unmanned air-to-ground system and deploys the tasks on different ends according to different calculation force requirements. The cloud is mainly responsible for scheduling, and task scheduling is performed after all the intelligent agent operation data are collected, so that the defect that a single intelligent agent can only schedule itself is avoided, overall scheduling of a system level is realized, and cooperation of unmanned systems of heterogeneous air spaces is realized. The edge equipment is mainly responsible for path planning and online repositioning, and the path planning is carried out on the intelligent agent according to cloud overall scheduling, so that the situation of path crossing and collision when the intelligent agent plans the independent path of the intelligent agent is avoided. The executing end agent is only responsible for the collection of environmental data and the motion control of the executing end agent. Tasks borne by the cloud end, the edge and the executing end agent are independent of each other, repetition is avoided, and the operation efficiency of the system is improved.
2. The computing tasks of the cloud server and the edge computing equipment are realized by a containerization method, and the cloud task unloading and the edge task horizontal migration are realized through the Kubernetes container management deployed in the cloud and the edge computing equipment KubeEdge, so that cloud edge coordination and edge-edge coordination tasks are realized;
3. the task of the executing end agent is minimized and only takes charge of data collection and self control, so that the calculation force requirement is greatly reduced, and the hardware cost of the calculating unit is greatly reduced. The hardware of the computing unit for the executing end agent in the general market is an industrial personal computer, and the price of the computing unit is generally more than ten thousand yuan (taking the industrial personal computer commonly used in the industry as an example). The hardware of the execution end agent calculation unit provided by the invention can be realized by using the minimum system board with the same calculation force as Nvidia Jetson Tx2, the price is generally about 2000 yuan, and the cost of the hardware is greatly reduced.
4. The multi-agent collaborative mapping method under the cloud side architecture is provided, and the map of the related environment is successfully established by application, so that the adaptability in unfamiliar environments is improved.
Drawings
FIG. 1 is a cloud end architecture of the heterogeneous air space unmanned control system of the present invention;
FIG. 2 is a track following control framework of the present invention;
fig. 3 is a high-precision map of the actual environment of an embodiment.
Detailed Description
The specific technical scheme of the invention is described with reference to the accompanying drawings.
An intelligent collaborative heterogeneous air-ground unmanned system based on a cloud end architecture is shown in fig. 1, and the system is composed of a cloud server, edge computing equipment and an execution end intelligent body. The cloud server consists of a data storage device A and an artificial intelligent big data calculation server. The cloud server firstly receives operation data, original point cloud data, image data and positioning data uploaded by the edge computing device through the firewall security facility through the data access module, and stores and backs up the operation data, the original point cloud data, the image data and the positioning data in the data storage device A. After the original data collection is completed, the scheduling service, the mapping service and the model training service respectively ask the data storage device A for corresponding operation data, point cloud data, image data and intelligent agent positioning data. The scheduling service completes cloud-edge workflow scheduling by analyzing the current intelligent agent running state, task demands and current residual computing power, realizes the task scheduling of the space unmanned system considering global system level, and finally realizes the task cooperation of the execution end to complete scheduling tasks; the map building service completes the reconstruction work of the global map through merging the multi-sensor information such as the point cloud data, the image data and the like, and continuously updates the local map through the map information collected in the operation process to obtain the millimeter-level error map information updated in real time; the model training service optimizes the current co-location model parameters through the errors of the current location information and the model predictive location information, minimizes the location errors and obtains the current optimal model parameters. And after the cloud service finishes the task, transmitting the information to the edge computing equipment. Under the condition that cloud computing power resources are tense and edge computing equipment is idle, part of data preprocessing tasks in the high-precision graph building task can be unloaded to the edge computing equipment to be executed, and the data preprocessing tasks are uploaded to a cloud server after the edge computing equipment completes the tasks.
The edge computing device is comprised of a mobile 5G communication base station and several sets of mobile edge computing devices (MECs). The mobile 5G base station is responsible for communication services of the mobile MEC and the cloud, and communication services of the MEC and the executing end agent, and low-delay transmission of large data volume is realized. In MEC, firstly, path planning application, online repositioning application, data fusion application and data interaction application are instantiated into Pod composed of a plurality of containers, and cloud edge coordination, automatic load balancing and offline autonomy of edges are realized through Kubernetes container management of KubEdge and cloud. In the container instantiated by the data interaction application, the MEC receives cloud data and execution-end agent data through the 5G base station, and meanwhile, format conversion and preprocessing are carried out on the execution-end agent data, and then the data are temporarily stored in a data storage device B which can be accessed by other containers. In the container instantiated by the path planning application, the container first accesses the task target and the current position information of a certain agent in the data storage device B, and obtains a feasible path of the certain agent through a path planning algorithm. In the container instantiated by the online repositioning application, the container firstly accesses point cloud data and image data currently obtained by a certain agent in the data storage device B, compares the point cloud data and the image data with millimeter-level error map data, and performs the co-positioning of multiple agents by combining a co-positioning model issued by a cloud, so as to update the current position and posture information of the agent and realize online repositioning. In the container instantiated by the data preprocessing application, the container first accesses the multi-agent map data and the running data in the data storage device B, synchronizes the data time stamps, and converts the data from the ros_message format to a generic data format. All the tasks can be equally realized in a plurality of groups of MECs, and when the MEC responsible for a certain task is too much in calculation load or can not be accessed due to hardware problems, the parallel migration of the task can be realized through a side-by-side coordination mechanism in the edge core.
The execution end intelligent agent consists of a plurality of heterogeneous unmanned aerial vehicles and unmanned vehicles. The unmanned aerial vehicle is provided with an IMU and a camera, acquires map information, and is provided with the IMU, a laser radar and the camera. And the unmanned aerial vehicle are both provided with a computing device with lower computing power and a 5G wireless communication module, so that the uploading of data and the calculation of the track tracking control quantity are realized.
The specific algorithm and implementation methods related to the cloud server, the edge computing device and the execution end agent are respectively described below.
1.1 cloud server
1.1.1 data Access and data storage
And receiving the mobile 5G communication base station in the edge computing equipment by the cloud network port in a Transmission Control Protocol (TCP) communication mode through data access. And stored in the data storage device a in the form of DB (database).
1.1.2 scheduling services
The scheduling service comprises scheduling of cloud edge service tasks and task scheduling of an intelligent agent at an execution end of the space unmanned system. Scheduling of cloud service tasks is mainly performed, namely unloading of cloud tasks. Cloud edge clusters built through Kubernetes container management and CloudCore assemblies are used, and cloud servers monitor cloud computing power and task ending time in real time. When the computing power resource of the cloud server is tense and the computing power of the edge computing equipment is remained, the cloud lightweight task is realized through the scheduling service, wherein the cloud lightweight task comprises a pre-integration link, a point cloud data de-distortion link and the like in the earlier stage of the mapping service, and the cloud computing equipment is unloaded to the edge computing equipment to realize cloud edge coordination. And then, aiming at task scheduling of the execution end agent of the space unmanned system, the task coordination of the execution end agent is realized by analyzing the operation data uploaded by the edge computing equipment and applying a multi-objective optimization strategy based on integer programming and queuing theory.
1.1.3 mapping service
And carrying out distributed collaborative mapping according to the original point cloud data, the image data and the IMU data uploaded by the edge computing equipment. The data of the individual agents is first processed. The first step is IMU pre-integration, and the position, posture and speed of the intelligent agent in the time of two adjacent frames of point cloud data are obtained through the pre-integration process. After IMU pre-integration, de-distortion of the point cloud data is performed. And calculating the difference between the point cloud data acquisition time and the frame starting time, and performing Euler interpolation through a uniform motion model to obtain a transformation matrix of the point cloud and the starting time. After correcting the point cloud distortion, performing feature extraction by using a FAST algorithm, extracting key frames and feature points of image data and point cloud data, tracking the feature points and eliminating outliers. And carrying out feature matching on the basis of feature extraction. The VINs-MONO algorithm is applied to a Visual Inertial Odometer (VIO) to complete the motion estimation of the VIO.
And establishing characteristic dictionaries of different agents on the basis of the extracted key frames and the characteristic points. Based on the word bag model, an image database indexing method is applied, and the overlapped public areas are judged by judging the similarity of feature dictionaries among different intelligent agents, so that the splicing of point cloud data and image data of different intelligent agents is completed, and the splicing of a local map is realized to obtain a complete millimeter-level error map. The splicing method comprises the following steps:
firstly, a k-means clustering method is applied, N extracted feature points are generalized into k types, a first layer of nodes are formed, samples belonging to the nodes are gathered into k types, and a next layer is obtained. And so on, finally obtaining the leaf layer. And (5) splicing after judging the similarity degree of the leaf layers and exceeding a certain threshold value.
1.1.4 model training
Model training is based on the positioning data and network communication topological relation in the operation data uploaded by the edge computing equipment, a co-positioning model is obtained by training neural network parameters, and then the co-positioning model is issued to the edge computing equipment, so that a model foundation is provided for online repositioning service of the edge computing equipment. The training process is as follows:
and taking online repositioning data of the intelligent agent i as a true value output by a neural network, taking online repositioning data of the intelligent agent connected with the intelligent agent i in a communication network topology as a neural network input, taking the neural network as an RBF neural network, training hidden layer neural network parameters by applying a steepest gradient descent method, and when the difference between the output value of the neural network and the true value is within a threshold value and smaller than the last error, considering the trained parameters as current optimal parameters, thus completing the training of the model.
1.1.5Kubernetes container management
And a cloud deployment container arrangement system Kubernetes (K8 s) is used for constructing a cloud K8s cluster and is responsible for automatic deployment and scheduling management of cloud containers. And simultaneously deploying CloudCore components of the edge computing framework KubeEdge, building cloud edge clusters, and unifying all nodes of the nano-tube cloud clusters. Multiple containers (atomic schedule units pod) can be created in the KubeEdge, each container runs an application instance, and management, discovery and access of the application instance are realized through a built-in load balancing strategy.
1.2 edge computing device
1.2.1 Equipment containerization and KubeEdge
Edge core components of the Kubeedge are deployed at edges, all edge equipment nodes are managed to form cloud edge clusters, and cloud edge coordination, edge coordination and off-line autonomy of edge equipment are achieved. According to task demands issued by the cloud end, the edge equipment nodes schedule the tasks to the edge in a containerized mode, or cloud edges are unloaded by the KubeEdge framework according to user demands. And the Kubeedge framework realizes the task horizontal migration between edges according to the requirements of users.
1.2.2 Path planning tasks
Firstly, path planning application is instantiated and deployed in a container through K8s management, the container accesses a millimeter-level error map issued by a cloud, and the position of an ith execution end intelligent agent at K moment is combinedTask target point of ith execution end agent at k moment issued by cloud end +.>Ego-planner planning algorithm is applied to unmanned aerial vehicle intelligent agent, and A is applied to unmanned aerial vehicle intelligent agent * Algorithm, obtaining feasible Path of k moment of corresponding agent i 。
1.2.3 Online relocation task
Firstly, an online repositioning application is instantiated and deployed in a container through K8s management, the container accesses a millimeter-level error map issued by a cloud, and according to image data collected by an executing end intelligent agent i, characteristic points of the image data are extracted and are matched with the millimeter-level error map in characteristics, so that positioning information based on the millimeter-level error map is obtained. And the real-time positioning information of the intelligent agent i is updated in the running process by utilizing the positioning information of a plurality of intelligent agents with the connected communication topology with the intelligent agent i and applying a cooperative positioning model issued by a cloud, so as to complete the online repositioning task of the intelligent agent.
1.2.4 data preprocessing and data interaction tasks
Firstly, through K8s management, data preprocessing and data interaction are instantiated and respectively deployed in a plurality of containers capable of realizing data intercommunication. The data interaction instantiation container firstly receives point cloud data, image data and operation data uploaded by an executing end agent. The data preprocessing instantiation container performs preprocessing on its data basis. Wherein the image data is a series of pictures in a standard format, which may be left unprocessed. The point cloud data is a type data packet customized by a laser radar manufacturer, cannot be directly used for high-precision mapping service of the cloud, and the running data is a certain type specified in the ROS_message and can be directly used only by further analysis. And the time stamps are not synchronized because different intelligent systems may not be consistent in time. Firstly, synchronizing point cloud data and operation data time stamps by utilizing a network time protocol (ntp), then establishing a corresponding float array to transfer the point cloud data and the operation data to form a data type in a standard format, and finishing data preprocessing. And the data interaction instantiation container sends the preprocessed data to a data access interface of the cloud server to finish the transmission of the data in the system.
1.3 execution side agent
1.3.1 map data acquisition and operation data upload
The executing terminal intelligent agent collects environmental data through onboard sensors including but not limited to laser radar, IMU, camera and the like, and uploads the environmental data to the edge computing equipment through a 5G wireless communication module onboard.
1.3.2 track following control
The execution end intelligent agent i receives a planning task Path issued by the edge computing equipment i According to the error between the current position of the intelligent agent and the target position, the control quantity is calculated, so that the intelligent agent can realize high-precision track tracking under the condition of interference (such as road condition change and load change). The system model of the executing-end agent can be abstracted as follows:
wherein x is 1 (t) is agent location information, x 2 (t) is agent speed information, b 0 For constant control gain, u (t) is the control input, x 3 And (t) is the interference received by the agent.
The overall control framework is shown in fig. 2.
Wherein v is 0 (t) is Path i The provided path position information, v 1 (t) tracking signal which is path information, v 2 (t) is the differential signal of the path signal, i.e. the speed information of the path. y (t) =x 1 (t) is the location information of the agent,for the estimate of the agent location information, +.>For the estimated value of the speed information of the agent, +.>Is an estimate of the disturbance experienced by the agent.
The form of the variable gain tracking differentiator is as follows:
where k is a fixed positive constant and,for any one in the argument v 1 (t)-v 0 (t) continuous even function monotonically decreasing in E (0, +)
The form of the variable gain extended state observer is as follows:
where ω is a fixed positive constant and,for estimating signal errors +.>For any one in the argument e 1 (t) E (0, +_E) is a continuous even function with monotonically decreasing.
The nonlinear composite anti-interference controller is as follows:
wherein k is 1 ，k 2 Is a fixed positive constant which is used for the control of the power supply,as any positive bounded continuous function, +.>Is a position tracking error.
Executing a specific task in an unknown environment, and realizing the intelligent collaborative heterogeneous unmanned air space system based on the cloud edge end architecture by the following steps:
step 1: the executing end intelligent agent unmanned plane and the unmanned plane collect map data such as point cloud data, image data and the like, and upload the map data to the edge computing equipment after the collection is completed.
Step 2: the data preprocessing task and the data interaction task are instantiated into a container at the edge computing device, and the map data are subjected to time stamp synchronization and data format conversion in the data preprocessing task container. And uploading the processed data to a cloud server data access interface through a data interaction task container.
Step 3: and the cloud server stores and backs up the data through a data storage function after receiving the data uploaded by the edge computing device. And then instantiating the mapping service into a cloud server container, and accessing map data by the container to realize high-precision mapping. And issue the millimeter level error map to the edge computing device.
Step 4: the cloud server instantiates the scheduling service into a cloud server container, and firstly, the computing power occupation and the task completion time of the current cloud server are evaluated. If the occupation is large or the task cannot be completed on time, unloading the part with smaller calculation amount in part of the tasks to the edge computing equipment, wherein the large occupation means that the cloud computing power resource occupies 90% or more; the smaller refers to the image computing capability which does not need GPU computing or can be borne by Jetson TX 2; . And then, by accessing the operation data, carrying out overall scheduling on the executing end intelligent agent, realizing task cooperation of the executing end intelligent agent, and transmitting scheduling information to the edge computing equipment.
Step 5: the edge computing device first instantiates a path planning task and an online relocation task into the edge computing device container. The path planning task container makes path planning for the intelligent agent by accessing the millimeter-level error map and the scheduling information of the intelligent agent at the execution end, gives out a specific executable tracking path and issues the specific executable tracking path to the intelligent agent at the corresponding execution end.
Step 6: and after receiving the corresponding planned path, the executing end agent applies a track tracking algorithm to track the corresponding path and continuously uploads the operation data to the edge computing equipment. The online repositioning task container in the edge computing equipment realizes online repositioning by accessing the millimeter-level error map and the operation data, and transmits positioning information to the corresponding executing-end intelligent agent. The steps are repeated until the execution end agent completes the issued dispatching task. If the problem of network or hardware occurs in the process of repeating the steps, when the edge computing equipment cannot continue the task, the task can be horizontally migrated through a side-to-side cooperative mechanism, so that uninterrupted execution of the task flow is realized. The task refers to specific tasks distributed to the edge computing equipment, such as online repositioning and path planning tasks, and the task flow refers to the execution of all steps according to the sequence.
Step 7: the edge computing device uploads the positioning information to the cloud server, the cloud server instantiates the model training task into the container, the container completes optimization of the model by accessing the positioning information, and optimized parameters are issued to the edge computing device. The unfamiliar environment and the familiar environment are determined whether the map information acquisition and mapping work is performed on the environment, if the map is established, the unfamiliar environment is the familiar environment, otherwise, the unfamiliar environment is the unfamiliar environment.
The present invention successfully builds a map of the relevant environment as shown in fig. 3.
After the strange environment completes the step 1-3, the strange environment is converted into a familiar environment. When the whole flow is executed again, only the step 4 is needed to start.
In this embodiment, the task offloading and cloud edge cluster establishment method in the scheduling task may be replaced by a scheme provided by a cloud service provider. The multi-objective optimization method based on integer programming and queuing theory can be replaced by a multi-objective optimization method based on genetic algorithm or a multi-objective particle swarm algorithm.
The feature extraction method in the mapping service can be replaced by a SIFT operator method, and the VINs-Mono method can be replaced by any VIO method.
The RBF neural network in model training can be replaced by any neural network which can approximate a nonlinear function.
Ego-planner and A in Path planning task * The algorithm may be replaced by any path planning algorithm.
Claims (6)
1. The intelligent collaborative heterogeneous air-ground unmanned system based on the cloud end architecture is characterized by comprising a cloud server, edge computing equipment and an execution end intelligent body;
based on a cloud side end architecture, deploying tasks on a cloud server, edge computing equipment and an execution end intelligent agent according to different calculation force requirements of each task; the cloud server is responsible for task scheduling and high-precision mapping work with the largest calculation force demand, and the edge computing equipment is responsible for general online repositioning of the calculation force demand and path planning; the computational power requirements generally refer to image computing power that does not require GPU computing or Jetson TX2 can afford; the execution end agent is responsible for track tracking control and perception of environmental information; the cloud server and the edge computing equipment are vertically unloaded through cloud edge tasks, and cloud edge coordination and edge coordination are realized among the edge computing equipment through task horizontal migration;
the cloud server consists of a data storage device A and an artificial intelligent big data calculation server; the cloud server firstly receives operation data, original point cloud data, image data and positioning data uploaded by the edge computing equipment through a firewall security facility through a data access module, and stores and backs up the operation data, the original point cloud data, the image data and the positioning data in a data storage device A; after the original data collection is completed, the scheduling service, the mapping service and the model training service respectively ask the data storage equipment A for corresponding operation data, point cloud data, image data and intelligent agent positioning data; the scheduling service completes cloud-edge workflow scheduling by analyzing the current intelligent agent running state, task demands and current residual computing power, realizes the task scheduling of the space unmanned system considering global system level, and finally realizes the task cooperation of the execution end to complete scheduling tasks; the map building service completes the reconstruction work of the global map by fusing the multi-sensor information, and continuously updates the local map by the map information collected in the running process to obtain the millimeter-level error map information updated in real time; the model training service optimizes the current co-location model parameters through the errors of the current location information and the model prediction location information, minimizes the location error and obtains the current optimal model parameters; after the cloud service finishes the task, transmitting the information to the edge computing equipment;
the edge computing equipment consists of a mobile 5G communication base station and a plurality of groups of mobile edge computing equipment MECs; the mobile 5G base station is responsible for communication services of mobile MEC and cloud end, and MEC and an executing end agent, so that low-delay transmission of large data volume is realized; firstly, a path planning application, an online repositioning application, a data fusion application and a data interaction application are instantiated into a Pod composed of a plurality of containers in MEC, and cloud edge coordination, automatic load balancing and off-line autonomy of edges are realized through Kubernetes container management of a KubEdge and a cloud end;
the execution end intelligent body consists of a plurality of heterogeneous unmanned aerial vehicles and unmanned vehicles; the unmanned aerial vehicle is provided with an IMU and a camera, acquires map information, and is provided with the IMU, a laser radar and the camera to acquire map information; and the unmanned aerial vehicle are both provided with a computing device and a 5G wireless communication module, so that the uploading of data and the calculation of the track tracking control quantity are realized.
2. The cloud-edge-architecture-based intelligent collaborative heterogeneous air-ground unmanned system is characterized in that in a cloud server, partial data preprocessing tasks in a high-precision mapping task are unloaded to edge computing equipment for execution under the condition that cloud computing resources are tense and edge computing equipment is idle, and the cloud server is uploaded after the edge computing equipment completes the tasks.
3. The cloud-edge architecture-based intelligent collaborative heterogeneous air-ground unmanned system of claim 2, wherein in the edge computing device, in the container instantiated by the data interaction application:
the MEC receives cloud data and execution-end agent data through a 5G base station, performs format conversion and preprocessing on the execution-end agent data, and then temporarily stores the data in a data storage device B which can be accessed by other containers;
in a container instantiated by a path planning application, the container firstly accesses the task target and the current position information of a certain agent in the data storage equipment B, and a feasible path of the certain agent is obtained through a path planning algorithm;
in a container instantiated by an online repositioning application, the container firstly accesses point cloud data and image data currently obtained by a certain intelligent agent in the data storage device B, compares the point cloud data and the image data with millimeter-level error map data, and performs the co-positioning of multiple intelligent agents by combining a co-positioning model issued by a cloud, and updates the current position and posture information of the intelligent agent to realize online repositioning;
in a container instantiated by a data preprocessing application, the container firstly accesses map data and operation data of multiple agents in the data storage device B, synchronizes data time stamps and converts the data from an ROS_message format to a universal data format;
all the tasks are equally realized in a plurality of groups of MECs, and when the MEC responsible for a certain task cannot be accessed due to excessive computational load or hardware problems, the parallel migration of the task is realized through a side-to-side collaboration mechanism in the edge core.
4. A method for implementing an intelligent collaborative heterogeneous air-ground unmanned system based on a cloud end architecture according to any one of claims 1 to 3, wherein the task is performed in a strange environment, comprising the steps of:
step 1: the intelligent agent unmanned aerial vehicle at the execution end and the unmanned aerial vehicle head firstly acquire map data, and the map data is uploaded to the edge computing equipment after the acquisition is completed;
step 2: instantiating the data preprocessing task and the data interaction task into a container at the edge computing device, and performing timestamp synchronization and data format conversion on map data in the data preprocessing task container; uploading the processed data to a cloud server data access interface through a data interaction task container;
step 3: the cloud server stores and backs up the data through a data storage function after receiving the data uploaded by the edge computing device; then instantiating the mapping service into a cloud server container, and accessing map data by the container to realize high-precision mapping; and issuing the millimeter-level error map to the edge computing device;
step 4: the cloud server instantiates the scheduling service into a cloud server container, and firstly, the computing power occupation and task completion time of the current cloud server are evaluated; if the occupation is large or the task cannot be completed on time, unloading the part with smaller calculation amount in part of the tasks to edge computing equipment for carrying out; the larger cloud computing power resource accounts for 90% or more; the smaller refers to the image computing capability which does not need GPU computing or can be borne by Jetson TX 2; then, by accessing the operation data, the executing end agent is integrally scheduled, so that the task cooperation of the executing end agent is realized, and the scheduling information is issued to the edge computing equipment;
step 5: the edge computing device firstly instantiates a path planning task and an online repositioning task into an edge computing device container; the path planning task container makes path planning for the intelligent agent by accessing the millimeter-level error map and the scheduling information of the intelligent agent at the execution end, gives out a specific executable tracking path and issues the specific executable tracking path to the intelligent agent at the corresponding execution end;
step 6: after receiving the corresponding planning paths, the executing end agent applies a track tracking algorithm to track the corresponding paths and continuously uploads operation data to the edge computing equipment; the online repositioning task container in the edge computing equipment realizes online repositioning by accessing the millimeter-level error map and the operation data, and transmits positioning information to the corresponding execution end intelligent agent; repeating the steps until the execution end agent completes the issued dispatching task;
step 7: the edge computing device uploads the positioning information to the cloud server, the cloud server instantiates the model training task into the container, the container completes optimization of the model by accessing the positioning information, and optimized parameters are issued to the edge computing device.
5. The method for implementing the intelligent collaborative heterogeneous air-ground unmanned system based on the cloud end architecture according to claim 4, wherein in step 6, if a problem of network or hardware occurs in the repetition process, when the edge computing device cannot continue the task, the task is horizontally migrated through a side-to-side collaboration mechanism, so that uninterrupted execution of the task flow is implemented; the task refers to specific tasks distributed to the edge computing equipment, such as online repositioning and path planning tasks, and the task flow refers to the execution of all steps according to the sequence.
6. The method for realizing the intelligent collaborative heterogeneous air-ground unmanned system based on the cloud end architecture, which is disclosed in claim 4, is characterized in that after the step 1-3 is completed in the strange environment, the strange environment is converted into a familiar environment; when the whole flow is executed again, starting from the step 4; the unfamiliar environment and the familiar environment are determined whether the map information acquisition and mapping work is performed on the environment, if the map is established, the unfamiliar environment is the familiar environment, otherwise, the unfamiliar environment is the unfamiliar environment.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310800372.1A CN116744368B (en) | 2023-07-03 | 2023-07-03 | Intelligent collaborative heterogeneous air-ground unmanned system based on cloud side end architecture and implementation method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310800372.1A CN116744368B (en) | 2023-07-03 | 2023-07-03 | Intelligent collaborative heterogeneous air-ground unmanned system based on cloud side end architecture and implementation method |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN116744368A true CN116744368A (en) | 2023-09-12 |
| CN116744368B CN116744368B (en) | 2024-01-23 |
Family
ID=87914980
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202310800372.1A Active CN116744368B (en) | 2023-07-03 | 2023-07-03 | Intelligent collaborative heterogeneous air-ground unmanned system based on cloud side end architecture and implementation method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN116744368B (en) |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110210806A (en) * | 2019-05-27 | 2019-09-06 | 大连理工大学 | A kind of the cloud base unmanned vehicle framework and its control evaluation method of 5G edge calculations |
| CN112000130A (en) * | 2020-09-07 | 2020-11-27 | 哈尔滨工业大学 | Unmanned aerial vehicle's multimachine cooperation high accuracy is built and is drawn positioning system |
| CN112099510A (en) * | 2020-09-25 | 2020-12-18 | 东南大学 | Intelligent agent control method based on end edge cloud cooperation |
| WO2021067140A1 (en) * | 2019-10-04 | 2021-04-08 | Intel Corporation | Edge computing technologies for transport layer congestion control and point-of-presence optimizations based on extended in-advance quality of service notifications |
| CN112948088A (en) * | 2021-03-17 | 2021-06-11 | 北京理工大学 | Cloud workflow intelligent management and scheduling system in cloud computing platform |
-
2023
- 2023-07-03 CN CN202310800372.1A patent/CN116744368B/en active Active
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110210806A (en) * | 2019-05-27 | 2019-09-06 | 大连理工大学 | A kind of the cloud base unmanned vehicle framework and its control evaluation method of 5G edge calculations |
| WO2021067140A1 (en) * | 2019-10-04 | 2021-04-08 | Intel Corporation | Edge computing technologies for transport layer congestion control and point-of-presence optimizations based on extended in-advance quality of service notifications |
| CN112000130A (en) * | 2020-09-07 | 2020-11-27 | 哈尔滨工业大学 | Unmanned aerial vehicle's multimachine cooperation high accuracy is built and is drawn positioning system |
| CN112099510A (en) * | 2020-09-25 | 2020-12-18 | 东南大学 | Intelligent agent control method based on end edge cloud cooperation |
| CN112948088A (en) * | 2021-03-17 | 2021-06-11 | 北京理工大学 | Cloud workflow intelligent management and scheduling system in cloud computing platform |
Non-Patent Citations (1)
| Title |
|---|
| 毋涛;徐雷;刘畅;: "5G MEC边缘智能架构研究", 信息通信技术, no. 02, pages 48 - 51 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN116744368B (en) | 2024-01-23 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Erdelj et al. | UAVs that fly forever: Uninterrupted structural inspection through automatic UAV replacement | |
| Hu et al. | UAV-assisted vehicular edge computing for the 6G internet of vehicles: Architecture, intelligence, and challenges | |
| Giyenko et al. | Intelligent UAV in smart cities using IoT | |
| Wang et al. | AI-based cloud-edge-device collaboration in 6G space-air-ground integrated power IoT | |
| CN109669477A (en) | A kind of cooperative control system and control method towards unmanned plane cluster | |
| CN108966286B (en) | Unmanned aerial vehicle-assisted mobile edge computing system and information bit distribution method thereof | |
| WO2022001120A1 (en) | Multi-agent system and control method therefor | |
| CN111765924A (en) | Atmospheric environment monitoring method and system based on multiple unmanned aerial vehicles | |
| CN105189237A (en) | Apparatus for controlling a land vehicle which is self-driving or partially self-driving | |
| Cao et al. | MEC-driven UAV-enabled routine inspection scheme in wind farm under wind influence | |
| Maier | Research challenges for systems-of-systems | |
| CN113612820A (en) | Cloud edge device platform control architecture and method based on Superedge and edgeXFioutry | |
| CN114063474A (en) | Semi-physical simulation system of unmanned aerial vehicle cluster | |
| CN111836409A (en) | Multi-agent system structure and control method thereof | |
| CN115801774A (en) | Space-based system edge cloud architecture and self-organizing system and method | |
| CN114884998B (en) | Cooperative software defined vehicle-mounted network system, scheduling method and CACC | |
| KR20210063990A (en) | Method of machine learning based unmanned aerial vehicle mobile edge server collabrative task matching and offloading | |
| CN114337783A (en) | Space distributed edge computing architecture and service processing method | |
| ElSayed et al. | Autonomous drone charging station planning through solar energy harnessing for zero-emission operations | |
| Lin et al. | Integrating mobile edge computing into unmanned aerial vehicle networks: An SDN-enabled architecture | |
| Wang et al. | Edge intelligence for mission cognitive wireless emergency networks | |
| CN115209379A (en) | Power grid cloud edge cooperative inspection system and method based on 5G intelligent networked unmanned aerial vehicle | |
| CN116744368B (en) | Intelligent collaborative heterogeneous air-ground unmanned system based on cloud side end architecture and implementation method | |
| Chen et al. | A fast coordination approach for large-scale drone swarm | |
| CN109034723A (en) | A kind of controHower electrical steps managing and control system for procedural control |
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 |