CN114385359B - Cloud edge task time sequence cooperation method for Internet of things - Google Patents
Cloud edge task time sequence cooperation method for Internet of things Download PDFInfo
- Publication number
- CN114385359B CN114385359B CN202210016729.2A CN202210016729A CN114385359B CN 114385359 B CN114385359 B CN 114385359B CN 202210016729 A CN202210016729 A CN 202210016729A CN 114385359 B CN114385359 B CN 114385359B
- Authority
- CN
- China
- Prior art keywords
- task
- equipment
- time sequence
- internet
- social
- 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.)
- Active
Links
- 238000000034 method Methods 0.000 title claims abstract description 93
- 238000013508 migration Methods 0.000 claims abstract description 137
- 230000005012 migration Effects 0.000 claims abstract description 136
- 238000004364 calculation method Methods 0.000 claims abstract description 24
- 238000004422 calculation algorithm Methods 0.000 claims abstract description 11
- 239000013598 vector Substances 0.000 claims description 44
- 230000008901 benefit Effects 0.000 claims description 39
- 230000008569 process Effects 0.000 claims description 26
- 230000006870 function Effects 0.000 claims description 19
- 230000007774 longterm Effects 0.000 claims description 19
- 238000012545 processing Methods 0.000 claims description 19
- 230000008859 change Effects 0.000 claims description 9
- 238000010606 normalization Methods 0.000 claims description 7
- 238000012795 verification Methods 0.000 claims description 7
- 238000012163 sequencing technique Methods 0.000 claims description 4
- 230000002123 temporal effect Effects 0.000 claims description 4
- 238000009826 distribution Methods 0.000 abstract description 8
- 238000005516 engineering process Methods 0.000 description 7
- 238000004458 analytical method Methods 0.000 description 6
- 230000008447 perception Effects 0.000 description 6
- 238000012549 training Methods 0.000 description 4
- XLYOFNOQVPJJNP-UHFFFAOYSA-N water Substances O XLYOFNOQVPJJNP-UHFFFAOYSA-N 0.000 description 4
- 230000005540 biological transmission Effects 0.000 description 3
- 230000003993 interaction Effects 0.000 description 3
- 239000011159 matrix material Substances 0.000 description 3
- 230000011273 social behavior Effects 0.000 description 3
- 238000009825 accumulation Methods 0.000 description 2
- 238000013461 design Methods 0.000 description 2
- 238000011161 development Methods 0.000 description 2
- 238000010586 diagram Methods 0.000 description 2
- 238000002474 experimental method Methods 0.000 description 2
- 238000004519 manufacturing process Methods 0.000 description 2
- 238000012544 monitoring process Methods 0.000 description 2
- 230000008520 organization Effects 0.000 description 2
- 230000003068 static effect Effects 0.000 description 2
- 238000003860 storage Methods 0.000 description 2
- 238000006467 substitution reaction Methods 0.000 description 2
- 101100339496 Caenorhabditis elegans hop-1 gene Proteins 0.000 description 1
- 208000036829 Device dislocation Diseases 0.000 description 1
- 230000004075 alteration Effects 0.000 description 1
- 230000003190 augmentative effect Effects 0.000 description 1
- 230000006399 behavior Effects 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000033228 biological regulation Effects 0.000 description 1
- 238000004891 communication Methods 0.000 description 1
- 238000012733 comparative method Methods 0.000 description 1
- 230000007423 decrease Effects 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 230000001934 delay Effects 0.000 description 1
- 238000009792 diffusion process Methods 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 238000005265 energy consumption Methods 0.000 description 1
- 238000011478 gradient descent method Methods 0.000 description 1
- 230000036541 health Effects 0.000 description 1
- 238000010191 image analysis Methods 0.000 description 1
- 238000010801 machine learning Methods 0.000 description 1
- 230000007246 mechanism Effects 0.000 description 1
- 238000010295 mobile communication Methods 0.000 description 1
- 230000006855 networking Effects 0.000 description 1
- 238000005457 optimization Methods 0.000 description 1
- 230000009467 reduction Effects 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 238000010187 selection method Methods 0.000 description 1
- 238000004088 simulation Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F9/00—Arrangements for program control, e.g. control units
- G06F9/06—Arrangements for program control, e.g. control units using stored programs, i.e. using an internal store of processing equipment to receive or retain programs
- G06F9/46—Multiprogramming arrangements
- G06F9/50—Allocation of resources, e.g. of the central processing unit [CPU]
- G06F9/5083—Techniques for rebalancing the load in a distributed system
- G06F9/5088—Techniques for rebalancing the load in a distributed system involving task migration
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/30—Information retrieval; Database structures therefor; File system structures therefor of unstructured textual data
- G06F16/36—Creation of semantic tools, e.g. ontology or thesauri
- G06F16/367—Ontology
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/90—Details of database functions independent of the retrieved data types
- G06F16/901—Indexing; Data structures therefor; Storage structures
- G06F16/9024—Graphs; Linked lists
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F9/00—Arrangements for program control, e.g. control units
- G06F9/06—Arrangements for program control, e.g. control units using stored programs, i.e. using an internal store of processing equipment to receive or retain programs
- G06F9/46—Multiprogramming arrangements
- G06F9/54—Interprogram communication
- G06F9/546—Message passing systems or structures, e.g. queues
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N5/00—Computing arrangements using knowledge-based models
- G06N5/02—Knowledge representation; Symbolic representation
-
- G—PHYSICS
- G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
- G16Y—INFORMATION AND COMMUNICATION TECHNOLOGY SPECIALLY ADAPTED FOR THE INTERNET OF THINGS [IoT]
- G16Y10/00—Economic sectors
- G16Y10/75—Information technology; Communication
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Software Systems (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- Databases & Information Systems (AREA)
- Data Mining & Analysis (AREA)
- Computing Systems (AREA)
- Computational Linguistics (AREA)
- Animal Behavior & Ethology (AREA)
- Artificial Intelligence (AREA)
- Evolutionary Computation (AREA)
- Life Sciences & Earth Sciences (AREA)
- Mathematical Physics (AREA)
- Business, Economics & Management (AREA)
- Accounting & Taxation (AREA)
- Development Economics (AREA)
- Economics (AREA)
- General Business, Economics & Management (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention belongs to the field of task coordination of the Internet of things, and particularly relates to a cloud edge end task time sequence coordination method of the Internet of things; the method comprises the following steps: acquiring data of Internet of things equipment, wherein the Internet of things equipment comprises task migration equipment, an MEC server and a cloud server; according to the equipment data of the Internet of things, a candidate computing resource queue is obtained by adopting a computing resource discovery method based on time sequence interestingness; according to the computing resource demand of the task migration equipment, adopting a computing resource selection algorithm based on time sequence social similarity to make an optimal computing resource selection result among the candidate computing resource queues, the MEC server and the available computing resources in the cloud server; the task migration equipment carries out task migration according to the calculation resource selection result; according to the cloud edge end task cooperation method assisted by the end equipment designed according to the time sequence characteristics of the Internet of things equipment, the task cooperation efficiency is improved, the distribution of computing resources in the network is balanced, and the cloud edge end task cooperation method has a wide application prospect.
Description
Technical Field
The invention belongs to the field of task coordination of the Internet of things, and particularly relates to a cloud edge task time sequence coordination method of the Internet of things.
Background
With the development of internet applications and informatization technologies, the internet of things (Internet of Things, ioT) is also widely used worldwide as a product of internet technology. According to the analysis of international data companies and united states mountain medical companies, the total networking number of IoT devices exceeds 260 billions by the end of 2020. IoT devices have certain computing, storage, and communication capabilities, and in traffic, logistics, security, home, retail, etc. scenarios, inter-device collaboration serves IoT applications including data acquisition, image analysis, device monitoring, production orchestration, etc. With the development of machine learning, big data processing, augmented reality and other technologies in recent years, the latency requirement and the computational demand of IoT applications are continuously increasing. However, ioT devices are limited in their own storage, computing, etc., such that the devices cannot match the ever-increasing computing demands, resulting in reduced real-time computing task processing. How to alleviate the huge IoT application computing demands to enable users to obtain higher quality service experiences is a current research hotspot.
At present, thanks to the wide application of the fifth generation mobile communication technology, the edge Computing technology and the cloud Computing technology, when the IoT device has insufficient Computing resources, the device can migrate the Computing task to an edge Computing (MEC) server or a cloud server, and the Computing task of the IoT device is cooperatively processed by utilizing the Computing resources sufficient by the MEC server and the cloud server to meet the huge Computing requirements of IoT application, so that the cloud side cooperative migration processing of the Computing task becomes a feasible scheme for solving the problem of lack of Computing capability of the original IoT device.
In the traditional cloud-edge collaborative task migration technical scheme, a cloud-edge collaborative multi-user computing task collaborative method based on sub-mode optimization designs a greedy algorithm based on sub-mode theory, comprehensively optimizes computing resources of a cloud server and an MEC server, and improves multi-user task migration efficiency; the task migration problem of the cloud edge server is converted into a game problem by a two-stage task cooperation method based on a game theory in a cloud edge environment, and a two-stage migration algorithm is designed for solving; according to the cloud edge collaborative task collaboration method in the service organization type, migration decisions are differentially made on computing tasks with different resource requirements and time delays by introducing an SDN-based service data organization mechanism. However, in the cloud-edge task collaboration solution, the computation resource limitation of the MEC server is ignored, and in addition, because the cloud server has the characteristic of long transmission distance, if a large amount of computation tasks are uploaded to the cloud server to be processed, the total time delay of task processing is increased, so that the low time delay requirement of the IoT application cannot be met.
Aiming at the problems in the traditional cloud-edge collaborative task migration technical scheme, a series of collaborative task collaborative schemes among multiple MEC servers are provided at home and abroad to relieve the computing pressure of the MEC servers and shorten the task processing time delay. For example, in consideration of the load pressure of a single server, when the service quality of a user is obviously deteriorated, computing resources on other MEC servers in a cooperation area are used for relieving computing pressure, and an optimal task cooperation scheme is generated by jointly optimizing the total time delay and the energy consumption of an edge computing network; a task cooperation scheme taking a user as a center realizes the task migration delay reduction by sharing resources on a plurality of MEC servers around the user; however, the above technical solution of cooperative task ignores the limitation of computing resources on the adjacent MEC servers, and when computing resources on the adjacent MEC servers are short, the device can only migrate the computing task to the MEC server with a longer transmission distance or a longer waiting time.
Furthermore, ioT devices have a timing, which tends to work automatically, periodically, in a daily operating environment, to process computing tasks at fixed time nodes or in a specific temporal order. Because of the time sequence of task execution and task migration of the IoT device, the available computing resources in the edge network also change time sequence, if the time sequence of operation of the IoT device is ignored in the process of computing task migration, the time sequence of task migration needs and random device migration needs conflict with each other, and the phenomena of shortage of computing resources in the coverage area of individual edge nodes and increase of waiting time delay of task migration occur, which causes the problems of uneven computing resource distribution and low task cooperation efficiency. Because of the difference between the computing capacity and the running state, a certain amount of idle end equipment with sufficient computing resources exists in the network, and the computing resources of the idle end equipment are reasonably distributed, so that huge local computing pressure can be relieved, and unnecessary waiting time in the task migration process is reduced.
In summary, how to design an end-device assisted cloud-edge task collaboration method according to the timing characteristics of IoT devices, so as to improve the efficiency of cross-domain task migration and balance the distribution of computing resources in the network is a problem to be solved.
Disclosure of Invention
Aiming at the defects existing in the prior art, the invention provides a cloud edge task time sequence cooperation method of the Internet of things, which comprises the following steps: acquiring data of Internet of things equipment, wherein the Internet of things equipment comprises task migration equipment, an MEC server and a cloud server; according to the equipment data of the Internet of things, a candidate computing resource queue is obtained by adopting a computing resource discovery method based on time sequence interestingness; according to the computing resource demand of the task migration equipment, adopting a computing resource selection algorithm based on time sequence social similarity to make an optimal computing resource selection result among the candidate computing resource queues, the MEC server and the available computing resources in the cloud server; the task migration equipment carries out task migration according to the calculation resource selection result;
the process for obtaining the computing resource selection result by adopting the computing resource selection algorithm based on the time sequence social similarity comprises the following steps:
S1: selecting MEC servers with available computing resources larger than computing resources required by task migration equipment from all MEC servers, and calculating system benefits of task migration of the MEC servers;
s2: calculating system benefits of cloud server task migration according to cloud server data;
S3: creating a time sequence social corrugated network by taking task migration equipment as a center;
S4: calculating the time sequence social similarity of the candidate terminal equipment and the task migration equipment according to the time sequence corrugated network; the candidate terminal equipment is the Internet of things equipment in the candidate computing resource queue;
s5: selecting candidate terminal equipment with the greatest social similarity and calculating system benefits of task migration of the candidate terminal equipment;
s6: and selecting equipment corresponding to the maximum value in the system benefit of the MEC server task migration, the task migration system benefit of the cloud server and the system benefit of the candidate terminal equipment task migration as a task migration destination of the task migration equipment.
Preferably, the process of obtaining the candidate computing resource queue by adopting the computing resource discovery method based on the time sequence interestingness comprises the following steps:
calculating the long-term interest and the short-term interest of the Internet of things equipment according to the long-term task processing habit and the short-term task processing change of the Internet of things equipment;
Calculating the time sequence interestingness of the Internet of things equipment according to the long-term interestingness and the short-term interestingness of the Internet of things equipment;
And sequencing the time sequence interestingness according to the sequence from big to small, selecting the first H pieces of equipment of the Internet of things corresponding to the time sequence interestingness to construct a candidate computing resource queue, wherein H is an integer smaller than the total number of the equipment of the Internet of things.
Further, formulas for calculating the long-term interest level and the short-term interest level of the internet of things equipment are respectively as follows:
Wherein, Representing long-term interestingness,/>Representing short-term interestingness; /(I)Representing the total calculated task processing number within the device long time range DeltaT LA,/>Representing the number of times of processing the T w type computing task within the long-time range delta T LA of the device,/>Representing the total computational task throughput of the device over a short time horizon deltat s,Representing the number of times the device o n processes a calculation task of type t w.
Further, a formula for calculating the time sequence interestingness of the internet of things equipment is as follows:
Wherein, Representing time sequence interestingness,/>Representing normalized parameters,/>Representing timing preference,/>Represents a time-series perceptual weight, gamma l represents a first influence coefficient, gamma s represents a second influence coefficient,/>Representing the long-term interest level of the device in task type t w at time slice ts l,/>Representing the short-term interest of the device in task type t w at time slice ts l.
Preferably, creating the time-sequential social corrugated network includes: setting 5 types of social relations, namely a homogeneous object relation, a co-located object relation, a cooperative object relation, an ownership object relation and a social object relation; creating a time sequence knowledge graph according to social relations among the devices of the Internet of things; setting the number of ripple layers, and constructing a social ripple network according to the time sequence knowledge graph. .
Preferably, calculating the time sequence social similarity between the task migration device and the candidate end device according to the time sequence corrugated network includes:
Obtaining an embedded vector of a triplet and an embedded vector of candidate terminal equipment in a time sequence ripple network by adopting HyTE model;
calculating the association probability of the candidate terminal equipment and the task migration equipment according to the embedded vector of the triplet and the embedded vector of the candidate terminal equipment;
calculating social relevance of each ripple level representation in the time sequence ripple network according to the association probability;
obtaining a total social relevance according to the social relevance represented by each ripple level;
and calculating the time sequence social similarity of the candidate terminal equipment and the task migration equipment according to the total social relevance.
Further, obtaining the embedded vector of the triplet using the HyTE model includes:
S31: obtaining a quadruple according to the time sequence knowledge graph, wherein the quadruple comprises a head entity, a tail entity, a social relationship and duration time of the social relationship, and obtaining a triple in each time slice according to the quadruple;
s32: setting a time hyperplane on each time slice, and respectively projecting the triples into embedded vectors in the hyperplane;
S33: and constructing a verification function according to the embedded vector, constructing a loss function according to the verification function, calculating the loss function, returning to the step S32, and obtaining the embedded vector of the triplet when the loss function is minimum.
Further, the formula for calculating the association probability is:
Wherein pos i represents the probability of association of the candidate end device with the task migration device established by analysis of the ith triplet data, Embedded vector representing candidate end device,/>Representing the embedding of a vector by social relationship in the ith triplet data,/>Representing the index of the head entity in the data of the ith triplet by embedding the vector in the data of the ith tripletRepresents the/>A triple data set of a layer sequential ripple network.
Further, the formula for calculating the time-series social similarity is as follows:
wherein sim n,h represents the temporal social similarity of the candidate end device and the task migration device, sigmoid represents the normalization function, Representing the total social relevance of candidate end devices and task migration devices,/>Representing the embedded vector of the candidate end device,
Preferably, formulas for calculating the system benefit of the MEC server task migration, the system benefit of the cloud server task migration and the system benefit of the candidate end device task migration are respectively:
Wherein, System benefit representing MEC server task migration,/>System benefit representing cloud server task migration,/>System benefit representing candidate end device task migration, w being a normalization parameter, ZB m representing computing resource load balancing degree for completing task migration using MEC server, ZB Cd representing computing resource load balancing degree for completing task migration using cloud server, ZB n representing computing resource load balancing degree for completing task migration using candidate end device,/>Representing the total delay of the system.
The beneficial effects of the invention are as follows: firstly, establishing time sequence interestingness by analyzing interest preference of the IoT devices on different time slices, constructing candidate computing resource queues by the IoT terminal devices with similar perceived time sequence interests, and improving accuracy of computing resource perception by introducing the time sequence interestingness; then, using a time sequence knowledge graph to store time sequence social relations among the IoT devices, designing a time sequence social corrugated network based on the time sequence knowledge graph, analyzing the propagation process of social attributes among corrugated levels in the social corrugated network, establishing time sequence social similarity, and selecting candidate terminal devices in a computing resource queue through comparing the time sequence social similarity so as to assist task migration; compared with a local MEC task cooperation method, a cloud edge task cooperation method and a multi-MEC server task cooperation method, the cloud edge end task time sequence cooperation method can reduce waiting time delay by about 26%, 18% and 10%, and effectively improves task cooperation efficiency. In addition, the cloud end task time sequence cooperative method can effectively relieve the calculation pressure on the MEC server so as to balance the distribution of calculation resources; according to the cloud edge end task cooperation method assisted by the end equipment designed according to the time sequence characteristics of the IoT equipment, the efficiency of cross-domain task migration is improved, the distribution of computing resources in the network is balanced, and the cloud edge end task cooperation method has a wide application prospect.
Drawings
FIG. 1 is a flowchart of a computing resource selection algorithm based on temporal social similarity in the present invention;
FIG. 2 is a schematic diagram of a time-series social corrugated network according to the present invention;
FIG. 3 is a schematic diagram of the propagation process of social attributes in a corrugated network in the present invention;
FIG. 4 is a graph showing average latency versus time for four synergy methods of the present invention;
FIG. 5 is a graph showing average computing resource load balance for four collaborative methods according to the present invention;
Fig. 6 is a graph showing average latency versus time for four cooperative methods of the present invention for different IoT device numbers;
Fig. 7 is a graph comparing average computing resource load balances of four cooperative methods under different IoT device numbers.
Detailed Description
The following description of the embodiments of the present invention will be made clearly and completely with reference to the accompanying drawings, in which it is apparent that the embodiments described are only some embodiments of the present invention, but not all embodiments. All other embodiments, which can be made by those skilled in the art based on the embodiments of the invention without making any inventive effort, are intended to be within the scope of the invention.
The invention provides a cloud edge task time sequence cooperation method of the Internet of things, which comprises the following steps: acquiring IoT device data, wherein the IoT device comprises a task migration device, an MEC server and a cloud server; sensing end equipment by adopting a time sequence interestingness-based computing resource discovery method according to the data of the IoT equipment so as to establish a candidate computing resource queue; according to the computing resource demand of the task migration equipment, adopting a computing resource selection algorithm based on time sequence social similarity to make an optimal computing resource selection result among the candidate computing resource queues, the MEC server and the available computing resources in the cloud server; the task migration equipment carries out task migration according to the calculation resource selection result;
as shown in fig. 1, the process of obtaining a computing resource selection result by adopting a computing resource selection algorithm based on time sequence social similarity includes:
S1: selecting MEC servers with available computing resources larger than computing resources required by task migration equipment from all MEC servers, and calculating system benefits of task migration of the MEC servers;
s2: calculating system benefits of cloud server task migration according to cloud server data;
S3: creating a time sequence social corrugated network by taking task migration equipment as a center;
S4: calculating the time sequence social similarity of the candidate terminal equipment and the task migration equipment according to the time sequence corrugated network; the candidate terminal equipment is the Internet of things equipment in the candidate computing resource queue;
s5: selecting candidate terminal equipment with the greatest social similarity and calculating system benefits of task migration of the candidate terminal equipment;
s6: and selecting equipment corresponding to the maximum value in the system benefit of MEC server task migration, the system benefit of cloud server task migration and the system benefit of candidate terminal equipment task migration as a task migration destination of the task migration equipment.
In the running process of the IoT device, the IoT device has a high data interaction frequency for a specific type of computing task due to different usage scenarios, device uses and usage habits, i.e., the IoT device artificially generates interests for the computing task; however, the interest of IoT devices may be changed due to the time sequence of the execution of the computing task, and in order to more accurately perceive the available computing resources, the present invention constructs a candidate computing resource queue by using computing resource discovery based on the time sequence interest, which specifically comprises the following steps:
The interest of IoT devices is affected by both their long-term task processing habits and short-term task processing variations. Over a long period of time, the interest of IoT devices in different types of computing tasks may accumulate due to the behavior of their task processing, yielding a relatively stable long-term interest. Definition of Long-term interestingness Representing the interest degree of the equipment in a certain type of calculation task within a long time range; the calculation tasks can be divided into 28 types of traffic, city calculation, intelligent home, health monitoring and the like according to the regulations of the open source dataset FIWARE, and the long-time range is considered to be set according to specific conditions and can be taken for 24 hours; the long-term interestingness formula for calculating a certain type of calculation task is:
Wherein, Representing long-term interestingness,/>Representing the total calculated task processing number within the device long time range DeltaT LA,/>The number of times a calculation task of the type T w is processed in the long time range Δt LA is represented.
The habit of IoT devices to handle computing tasks may change over time. Within a short time frame, ioT devices exhibit interest drift phenomena to change the interest level in different types of computing tasks. Defining short-term interestingnessDescribing the interest of the device in a certain type of computing task in a short time frame; wherein, the short time range is considered to be set according to specific conditions, and can be taken for 2 hours; the short-term interestingness formula for calculating a certain type of calculation task is:
Wherein, Representing short-term interestingness,/>Representing the total computational task throughput of the device over a short time horizon DeltaT s,/>The number of times the task is processed is calculated for the device o n for the t w type.
Comprehensively considering the influence of two time sequence factors of long-term interest accumulation and short-term interest change on the interest of the IoT device, and defining time sequence preference to represent the time sequence interest of the IoT device in a certain type of calculation task in a time slice; defining a timing preference calculation formula as:
Wherein, Represents timing preference, gamma l represents a first influence coefficient, gamma s represents a second influence coefficient,/>Representing the long-term interest level of the device in task type t w at time slice ts l,/>Representing the short-term interest of the device in task type t w at time slice ts l.
Building a timing preference matrix of task migration device o n on time slice ts l for all task types within a set of task types TEstablishing a set according to the time preference matrix of each time slice in the time period TSThe time period TS is mainly determined according to the computing power of the internet of things device in the scene, and is generally set to 24 hours.
Extracting timing preferences of IoT devices on each time sliceThus, the time sequence interest is established, and the time sequence information on different time slices has different effects on the time sequence interest establishment due to the difference of the time slices in which the equipment is positioned. Setting time sequence perception weight/>Representing the/>, on different time slicesInfluence on time sequence interest, and constructing a time sequence perception weight matrix/>, in a time period TSAnd meet/>And/>Binding/>And establishing time sequence interestingness in a time period TS with sigma TS, wherein the formula is as follows:
Wherein, Representing time sequence interestingness,/>Representing normalized parameters,/>Representing the time-series perceptual weights.
Cloud servers establish devices at the end of each period TS by analyzing in real-time the historical information of each IoT device's performance of tasksAnd calculating time sequence interestingness for each calculation task type of all the IoT devices, sequencing the time sequence interestingness according to the sequence from big to small, and constructing a candidate calculation resource queue, namely a candidate end device queue, according to the IoT devices corresponding to the sequencing of the time sequence interestingness. When a device o n for processing a task of a t w type sends a task migration request, selecting first H pieces of IoT end devices (the H size is not fixed and is generally 6) which can meet the resource requirement of the device o n in a candidate end device queue corresponding to t w, and constructing a candidate computing resource queue/>, according to the first H pieces of IoT end devicesAs input to the computing resource selection process.
In societies, humans create complex social relationships with social information through social attributes. Similar to the social relationships between people, ioT devices may personify establish social relationships between devices according to their collaborative associations, geographic locations, operational states, etc., with the social relationships between devices changing dynamically as computing tasks are performed chronologically.
Setting 5 types of social relations, namely a homogeneous object relation, a co-located object relation, a cooperative object relation, an ownership object relation and a social object relation, wherein the homogeneous object relation represents that the IoT devices have the same manufacturer or production lot, the co-located object relation represents that the IoT devices have similar geographic positions, the cooperative object relation represents that the IoT devices have historical interaction data, the ownership object relation represents that the owners of the IoT devices are the same, and the social object relation represents that the owners of the IoT devices have data interaction data; according to social relation, a time sequence social corrugated network is created by taking task migration equipment as a center, and the specific process is as follows:
Setting a timing knowledge graph to store social relationships of timing variations among IoT devices, the timing knowledge graph G being defined as:
G＝{O,RE,D,RT}
Where O represents a set of IoT devices, RE represents a set of social relationships between the devices, each relationship connects two IoT devices, D represents a set of social relationship start times and end times, and RT is a set of social relationship categories.
The data in G takes a four-element as a minimum management unit, the four-element comprises a head entity, a tail entity, a social relationship and duration time of the social relationship, the head entity and the tail entity belong to the data in O, the social relationship comprises the data of RE and RT types, and the duration time belongs to the data in D. In order to perceive the time sequence of the social relationship, the time sequence knowledge graph G TS in the time period TS is extracted, the time sequence knowledge graph G TS is divided into L static graph slices, and the slices are expressed as follows:
Wherein, Representing a set of four-tuple elements on the time slice ts l, the four-tuple comprising a head entity, a tail entity, a social relationship, and a duration of the social relationship; defining four-tuple elements (h, r, t, [ ts start,tsend ]) to represent forward triples (h, r, t) that can be correctly embedded over the validity period [ ts start,tsend ], h in (h, r, t) represents head entity, r represents social relationship, t represents tail entity, and sign/>Representing the set of forward triples within each time slice, symbol/>Representing the set of negative going triples within each time slice.
Modeling the time sequence knowledge graph G TS by adopting a knowledge graph embedding technology (HYPERPLANE-based Temporally aware Knowledge Graph Embedding, hyTE) based on hyperplane time perception, analyzing the change of social relations in static graph slices, and obtaining entity embedding representation and relation embedding representation of comprehensive time sequence social factors.
Setting a time hyperplane on each time slice, and defining a projection vector array of the hyperplane in a period TS asProjection vector/>Satisfy/>Associating the head entity and the social relation (h, r) with the tail entity t under different time slices, respectively projecting the triples as embedded vectors on corresponding hyperplanes, using bold (h, r, t) to represent the vector form of the triples (h, r, t), and the projection process on the time hyperplanes is as follows:
wherein, h ⊥、r⊥ and t ⊥ are respectively triplet vectors (h, r, t) in the hyperplane An up-projected embedded vector.
In order to judge the correctness of the projection of the triplet vector, defining a verification function to calculate the projection score of the triplet vector on the corresponding hyperplane; the projection score of the triplet vector in D l - is higher, if the projected vector belongs toThe score is lower and the projection is correct; if the projection is incorrect, the score is higher; the verification function is:
Using a verification function Constructing HyTE a model, and training HyTE the model by adopting an edge-ordering-based loss function to optimize the projection score of the triplet vector; the calculation is based on the edge ordering loss function formula:
Wherein, Representing a set of forward triples within a time slice ts l,/>Represents the set of negative triples within the time slice ts l, μ represents the edge spacing parameter for distinguishing positive triples from negative triples.
At the end of each period TS, a forward set of triplet embedded vectors is constructedAnd negative set/>And inputting the model parameters into HyTE training models, training the model parameters by a gradient descent method, and completing training the model when the loss function is minimum to obtain the embedded vector containing the time sequence social factors.
As shown in fig. 2, a ripple layer number is set, equipment for performing task migration is set as central equipment and is placed in the center of a ripple network, equipment on adjacent ripple layers are connected with each other through a time sequence social relationship, and equipment on the same ripple layer is connected with the central equipment through the time sequence social relationship with the same number; computing task migration devices and methods by analyzing the propagation process of social attributes between wavelayersTime sequence social similarity of the candidate terminal equipment.
As shown in fig. 3, when IoT device o n executing a task of type t w selects available computing resources at time ts l for task migration, analysis of propagation process in the corrugated network according to social attributesSocial similarity of the candidate terminal device can h and the central device o n; the propagation process of the social attribute of the task migration device o n between the corrugated layers is similar to water wave diffusion, and as the corrugations are gradually diffused, the transmission of the social attribute of o n is gradually weakened due to the increase of the number of connections. When calculating the social similarity between o n and the candidate terminal device can h, ioT devices with social behaviors need to be selected for analysis, and similar to the water ripple propagation process after two stones are put into water, the ripples of o n and can h are continuously diffused and intersected, as the ripple level increases, the water waves are gradually weakened, and the social similarity of IoT devices on the ripples also continuously decreases. If the number of candidate end devices on a high ripple level is large, the level attenuation is excessive, and in order to prevent the calculation amount from increasing due to a large number of irrelevant candidate end devices, the ripple level/>Is 3.
Computing task migration deviceThe specific process of the time sequence social similarity of the candidate terminal equipment is as follows:
Definition is located at IoT device set/>, on a layer ripple hierarchyThe method comprises the following steps:
Setting up Data set of layer triples, triples being/>Header entity of layer,/>The tail entity of the layer and the social relationship composition of the device are expressed as:
during propagation in hop-1, select a first ripple level The triplet in the set is adopted to acquire an embedding vector h ⊥、r⊥、t⊥ of the triplet and an embedding vector/>, of candidate terminal equipment can h, of the triplet by adopting HyTE modelCalculate at/>The association probability pos i of the candidate terminal device can h and the task migration device o n under the relationship is as follows:
Wherein pos i represents the probability of association of the candidate end device with the task migration device established by analysis of the ith triplet data, Embedded vector representing candidate end device,/>Embedded vector representing social relationship in ith triplet data,/>Representing the embedded vector of the header entity in the ith triplet data.
TraversingTo establish a set of associated probabilities { pos 1,pos2,...,posi,...,posI }, by analyzing the central device o n with the set/>The social relevance between o n and candidate device can h can be expressed as:
Wherein, Representing social relevance between candidate end devices can h and center device o n of the first tier established by analysis of triplet data in the first ripple tier,/>Representing the embedded vector of the tail entity in the ith triplet data.
Since the social properties of o n spread out with the ripple hierarchy, they are locatedAnd/>Social behavior of devices in the collection can embody potential social attributes of o n by calculating/>, in the associated probability pos i formulaReplaced by/>Calculation/>Will/>Substitution/>Thereby calculating/>
Comprehensively considering social behaviors on different ripple levels, the total social relevance of all can h and o n is as follows:
Calculating social similarity of all candidate terminal devices and task migration devices according to the total social relevance, wherein the formula is as follows:
Wherein sim n,h represents social similarity between the candidate terminal device and the task migration device, and sigmoid represents a normalization function, so that a value interval of the time sequence social similarity is [0,1].
Selecting a task migration destination for task migration equipment by adopting a computing resource selection method based on a time sequence social corrugated network, wherein the process is as follows:
The CPU computing resources available on each MEC server in the network are represented as Load center value/>, of computing resourceThe method comprises the following steps:
where K represents the number of MEC servers.
According to the load center value and the use state of the computing resources on each MEC server, the computing resource load balancing degree ZB m;ZBm is a computing resource load estimated value on the MEC server in the network after the task migration equipment completes migration by using the MEC server; the computing resource load balancing ZB m is:
Similarly, after the task migration equipment uses the cloud server to complete migration, the computing resource load balancing degree ZB Cd is a computing resource load estimated value on the MEC server in the network, and can be obtained by calculating the load center value and the total amount of available computing resources on the MEC server when the task migration equipment uses the cloud server to complete task migration; the computing resource load balancing degree ZB n is a computing resource load estimated value on the MEC server in the network after the task migration equipment uses the candidate end equipment to complete migration, and can be obtained by computing the load center value and the total amount of available computing resources on the MEC server when the task migration equipment uses the candidate end equipment to complete task migration. ZB increases as the difference in the use of computing resources on the MEC server increases, zb=0 if the available computing resources on each node are equal.
The system benefits of defining MEC server task migration are:
the system benefits for defining cloud server task migration are:
wherein w represents a normalization parameter, Representing the total delay of the system.
The smaller the task processing delay and the load balance, the higher the system benefit of task migration. If IoT devices upload computing tasks to the MEC server, causing additional latency, ZB m values may change. If the IoT device uploads the computing task to the idle candidate end device or cloud server, ZB n and ZB Cd do not change.
Ordering the social similarity of all candidate terminal devices and task migration devices, selecting the candidate terminal device with the largest social similarity, and calculating the system income of task migration of the candidate terminal device, wherein the calculation formula is as follows:
Comparing the system benefit of the MEC server task migration, the system benefit of the cloud server task migration and the system benefit of the candidate terminal equipment task migration, and selecting the equipment corresponding to the maximum value in the three system benefits as equipment for providing computing resources for the task migration equipment in the task migration process.
And the computing resource selection is completed, and the task migration equipment carries out task migration according to a computing resource selection result obtained by adopting a computing resource selection algorithm based on time sequence social similarity.
Evaluating the performance of the invention, selecting a local MEC task cooperative method (each IoT device can only use the computing resources of the local MEC server to perform task migration, if the local MEC server does not have available computing resources, waiting), a cloud edge task cooperative method (a cloud edge cooperative task cooperative migration scheme is used, an edge server or a cloud server is selected to provide computing resources) and a multi-MEC server task cooperative method (a task cooperative migration scheme in which the edge servers cooperate with each other is used, and computing resources between a plurality of edge servers and the cloud server are adopted) to compare with the invention (the task cooperative process is realized by analyzing the time sequence attribute of the IoT device and selecting idle end equipment to assist); setting average waiting time delayAnd average computing resource load degree/>As a performance index measure, the average waiting time delay represents the total time delay/>, of task migrationThe average computing resource loading level represents the average of computing resource loading levels ZB in the edge network.
Experiments are carried out on the social Internet of things dataset FIWARE, and the results are shown in fig. 4 and 5, so that a cloud edge end task time sequence cooperation method is adopted in an initial time rangeAnd/>Slightly higher than other comparison methods, the curve of the cloud edge task time sequence collaborative method scheme is reduced to the minimum value and gradually tends to be stable along with the increase of time, and the/>And/>All lower than the other three comparison methods. The reason for generating the simulation results is that:
1) The IoT devices have sufficient computing resources in the initial time range, and the computing tasks are processed by the IoT devices because the IoT devices have weak computing power and the MEC servers have sufficient computing resources, so that four comparison schemes in the initial time range Higher,/>Lower.
2) Because the time sequence interest information cannot be fully extracted in the initial time range, the embedded vector cannot fully fuse the time sequence social characteristics due to the problem of cold start of the time sequence knowledge graph, so that the accuracy of computing resource perception and computing resource selection is low.
3) Along with the continuous accumulation of time sequence interest information and time sequence social information, the cloud edge task time sequence cooperation method generates a decision conforming to the time sequence state of equipment according to the computing resource with the maximum benefit of the time sequence information sharing system, so thatAnd/>Lower than other comparative methods.
When the operation time of the IoT device is short, the cloud-side task timing coordination method is poor in performance because the timing information cannot be fully extracted, but compared with the local MEC task coordination method, the cloud-side task coordination method and the multi-MEC server task coordination method, the cloud-side task timing coordination method can reduce the latency by about 26%, 18% and 10% and effectively improve the task coordination efficiency as the operation time is increased. In addition, in the task cooperation process, the cloud edge task time sequence cooperation method can enable the computing resource distribution to be continuously optimized towards the balanced direction.
In order to further verify the performance of the cloud-edge task timing collaborative method, experiments are conducted by setting the number of different IoT devices. The results are shown in fig. 6 and 7, and as the number of devices increases, the cloud edge end task time sequence cooperation method and the other three comparison methods are adoptedAnd/>Also gradually increases, and/>, of cloud edge task timing coordination methodAnd/>Minimum. When the total number of devices grows to 640, the/>, of the four methodsReach maximum and/>In a reduced trend, the cloud edge task time sequence cooperative method still keeps the lowest/>And/>
Firstly, establishing time sequence interestingness by analyzing interest preference of the IoT devices on different time slices, constructing candidate computing resource queues by the IoT terminal devices with similar perceived time sequence interests, and improving accuracy of computing resource perception by introducing the time sequence interestingness; then, the time sequence social relation among the IoT devices is stored by utilizing the time sequence knowledge graph, a time sequence social corrugated network is designed based on the time sequence knowledge graph, the propagation process of social attributes among corrugated levels in the social corrugated network is analyzed, time sequence social similarity is established, and candidate terminal devices in a computing resource queue are selected according to the time sequence social similarity so as to assist task migration; compared with a local MEC task cooperation method, a cloud edge task cooperation method and a multi-MEC server task cooperation method, the cloud edge end task time sequence cooperation method can reduce waiting time delay by about 26%, 18% and 10%, and effectively improves task cooperation efficiency. In addition, the cloud end task time sequence cooperative method can effectively relieve the calculation pressure on the MEC server so as to balance the distribution of calculation resources; according to the cloud edge end task cooperation method assisted by the end equipment designed according to the time sequence characteristics of the IoT equipment, the efficiency of cross-domain task migration is improved, the distribution of computing resources in the network is balanced, and the cloud edge end task cooperation method has a wide application prospect.
While the foregoing is directed to embodiments, aspects and advantages of the present invention, other and further details of the invention may be had by the foregoing description, it will be understood that the foregoing embodiments are merely exemplary of the invention, and that any changes, substitutions, alterations, etc. which may be made herein without departing from the spirit and principles of the invention.
Claims (7)
1. The cloud edge task time sequence cooperation method for the Internet of things is characterized by comprising the following steps of: acquiring data of Internet of things equipment, wherein the Internet of things equipment comprises task migration equipment, an MEC server and a cloud server; according to the equipment data of the Internet of things, a candidate computing resource queue is obtained by adopting a computing resource discovery method based on time sequence interestingness; according to the computing resource demand of the task migration equipment, adopting a computing resource selection algorithm based on time sequence social similarity to make an optimal computing resource selection result among the candidate computing resource queues, the MEC server and the available computing resources in the cloud server; the task migration equipment carries out task migration according to the calculation resource selection result;
The process of obtaining candidate computing resource queues by adopting a computing resource discovery method based on time sequence interestingness comprises the following steps:
calculating the long-term interest and the short-term interest of the Internet of things equipment according to the long-term task processing habit and the short-term task processing change of the Internet of things equipment;
Calculating the time sequence interestingness of the Internet of things equipment according to the long-term interestingness and the short-term interestingness of the Internet of things equipment;
Sequencing the time sequence interestingness according to the sequence from big to small, selecting the first H pieces of equipment of the Internet of things corresponding to the time sequence interestingness to construct a candidate computing resource queue, wherein H is an integer smaller than the total number of the equipment of the Internet of things;
the process for obtaining the computing resource selection result by adopting the computing resource selection algorithm based on the time sequence social similarity comprises the following steps:
S1: selecting MEC servers with available computing resources larger than computing resources required by task migration equipment from all MEC servers, and calculating system benefits of task migration of the MEC servers;
s2: calculating system benefits of cloud server task migration according to cloud server data;
S3: creating a time sequence social corrugated network by taking task migration equipment as a center; creating a time-sequential social corrugated network includes: setting 5 types of social relations, namely a homogeneous object relation, a co-located object relation, a cooperative object relation, an ownership object relation and a social object relation; creating a time sequence knowledge graph according to social relations among the devices of the Internet of things; setting a ripple layer number, and constructing a time sequence social ripple network according to a time sequence knowledge graph;
S4: calculating the time sequence social similarity of the candidate terminal equipment and the task migration equipment according to the time sequence corrugated network; the candidate terminal equipment is the Internet of things equipment in the candidate computing resource queue; calculating the time sequence social similarity between the task migration equipment and the candidate terminal equipment according to the time sequence corrugated network comprises the following steps:
Obtaining an embedded vector of a triplet and an embedded vector of candidate terminal equipment in a time sequence ripple network by adopting HyTE model;
calculating the association probability of the candidate terminal equipment and the task migration equipment according to the embedded vector of the triplet and the embedded vector of the candidate terminal equipment;
calculating social relevance of each ripple level representation in the time sequence ripple network according to the association probability;
obtaining a total social relevance according to the social relevance represented by each ripple level;
calculating time sequence social similarity of the candidate terminal equipment and the task migration equipment according to the total social relevance;
s5: selecting candidate terminal equipment with the greatest social similarity and calculating system benefits of task migration of the candidate terminal equipment;
s6: and selecting equipment corresponding to the maximum value in the system benefit of MEC server task migration, the system benefit of cloud server task migration and the system benefit of candidate terminal equipment task migration as a task migration destination of the task migration equipment.
2. The internet of things cloud end task timing coordination method of claim 1, wherein formulas for calculating long-term interest and short-term interest of internet of things equipment are respectively:
Wherein, Representing long-term interestingness,/>Representing short-term interestingness; /(I)Representing the total calculated task processing number within the device long time range DeltaT LA,/>Representing the number of times of processing the T w type computing task within the long-time range delta T LA of the device,/>Representing the total computational task throughput of the device over a short time horizon DeltaT s,/>Representing the number of times the device o n processes a calculation task of type t w.
3. The internet of things cloud end task time sequence coordination method of claim 1, wherein a formula for calculating time sequence interestingness of internet of things equipment is as follows:
Wherein, Represents time sequence interestingness, w tsl represents normalization parameters,/>Representing timing preference,/>Represents a time-series perceptual weight, gamma l represents a first influence coefficient, gamma s represents a second influence coefficient,/>Representing the long-term interest level of the device in task type t w at time slice ts l,/>Representing the short-term interest of the device in task type t w at time slice ts l.
4. The method for task timing coordination at cloud edge end of internet of things according to claim 1, wherein obtaining the embedded vector of the triplet by adopting HyTE model comprises:
S31: obtaining a quadruple according to the time sequence knowledge graph, wherein the quadruple comprises a head entity, a tail entity, a social relationship and duration time of the social relationship, and obtaining a triple in each time slice according to the quadruple;
s32: setting a time hyperplane on each time slice, and respectively projecting the triples into embedded vectors in the hyperplane;
S33: and constructing a verification function according to the embedded vector, constructing a loss function according to the verification function, calculating the loss function, returning to the step S32, and obtaining the embedded vector of the triplet when the loss function is minimum.
5. The internet of things cloud end task timing coordination method of claim 1, wherein the formula for calculating the association probability is:
Wherein pos i represents the probability of association of the candidate end device with the task migration device, Embedded vector representing candidate end device,/>Embedded vector representing social relationship in ith triplet data,/>Embedded vector representing head entity in ith triplet data,/>Represents the/>A triple data set of a layer sequential ripple network.
6. The internet of things cloud end task timing coordination method of claim 1, wherein the formula for calculating the timing social similarity is as follows:
wherein sim n,h represents the temporal social similarity of the candidate end device and the task migration device, sigmoid represents the normalization function, Representing the total social relevance of candidate end devices and task migration devices,/>Representing the embedded vector of the candidate end device.
7. The method for collaborative task timing at cloud end of internet of things according to claim 1, wherein formulas for calculating system benefit of MEC server task migration, system benefit of cloud server task migration and system benefit of candidate end device task migration are respectively:
Wherein, System benefit representing MEC server task migration,/>System benefit representing cloud server task migration,/>Representing system benefits of task migration of candidate end equipment, w is a normalization parameter, ZB m represents computing resource load balancing degree when task migration is completed by using MEC server, ZB Cd represents computing resource load balancing degree when task migration is completed by using cloud server, ZB n represents computing resource load balancing degree when task migration is completed by using candidate end equipment, and I/ORepresenting the total delay of the system.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210016729.2A CN114385359B (en) | 2022-01-07 | 2022-01-07 | Cloud edge task time sequence cooperation method for Internet of things |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210016729.2A CN114385359B (en) | 2022-01-07 | 2022-01-07 | Cloud edge task time sequence cooperation method for Internet of things |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN114385359A CN114385359A (en) | 2022-04-22 |
| CN114385359B true CN114385359B (en) | 2024-05-14 |
Family
ID=81199590
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210016729.2A Active CN114385359B (en) | 2022-01-07 | 2022-01-07 | Cloud edge task time sequence cooperation method for Internet of things |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114385359B (en) |
Families Citing this family (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN118410173A (en) * | 2024-03-08 | 2024-07-30 | 暨南大学 | System user behavior prediction analysis method, system and equipment based on time sequence knowledge graph |
Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109684520A (en) * | 2019-01-07 | 2019-04-26 | 东北大学 | Large-scale time sequence diagram vertex similarity calculation method |
| CN110851713A (en) * | 2019-11-06 | 2020-02-28 | 腾讯科技（北京）有限公司 | Information processing method, recommendation method and related equipment |
| CN111586146A (en) * | 2020-04-30 | 2020-08-25 | 贵州电网有限责任公司 | Wireless internet of things resource allocation method based on probability transfer deep reinforcement learning |
| CN111930518A (en) * | 2020-09-22 | 2020-11-13 | 北京东方通科技股份有限公司 | Knowledge graph representation learning-oriented distributed framework construction method |
| CN112231489A (en) * | 2020-10-19 | 2021-01-15 | 中国科学技术大学 | Knowledge learning and transferring method and system for epidemic prevention robot |
| CN112860407A (en) * | 2021-03-02 | 2021-05-28 | 东南大学 | Cloud-edge cooperative sequence-dependent flow shop task scheduling execution optimization method |
| CN113553165A (en) * | 2020-04-23 | 2021-10-26 | 东北大学秦皇岛分校 | Game theory-based mobile edge computing task unloading and resource scheduling method |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| EP3262807A4 (en) * | 2015-02-25 | 2018-10-10 | Nokia Solutions and Networks Oy | Mechanism to support operator assisted parental control |
-
2022
- 2022-01-07 CN CN202210016729.2A patent/CN114385359B/en active Active
Patent Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109684520A (en) * | 2019-01-07 | 2019-04-26 | 东北大学 | Large-scale time sequence diagram vertex similarity calculation method |
| CN110851713A (en) * | 2019-11-06 | 2020-02-28 | 腾讯科技（北京）有限公司 | Information processing method, recommendation method and related equipment |
| CN113553165A (en) * | 2020-04-23 | 2021-10-26 | 东北大学秦皇岛分校 | Game theory-based mobile edge computing task unloading and resource scheduling method |
| CN111586146A (en) * | 2020-04-30 | 2020-08-25 | 贵州电网有限责任公司 | Wireless internet of things resource allocation method based on probability transfer deep reinforcement learning |
| CN111930518A (en) * | 2020-09-22 | 2020-11-13 | 北京东方通科技股份有限公司 | Knowledge graph representation learning-oriented distributed framework construction method |
| CN112231489A (en) * | 2020-10-19 | 2021-01-15 | 中国科学技术大学 | Knowledge learning and transferring method and system for epidemic prevention robot |
| CN112860407A (en) * | 2021-03-02 | 2021-05-28 | 东南大学 | Cloud-edge cooperative sequence-dependent flow shop task scheduling execution optimization method |
Non-Patent Citations (3)
| Title |
|---|
| "Modelling and Implementing Social Community Clouds";Ioan Petri;《IEEE Transactions on Services Computing》;20170630;第10卷(第3期);第410-422页 * |
| "移动云计算多媒体服务技术研究";周挺;《中国优秀硕士学位论文全文数据库 信息科技辑》;20160315(第2016年03期);第I138-7936页 * |
| "面向物联网的任务迁移优化研究";薛彦聪;《中国优秀硕士学位论文全文数据库 信息科技辑》;20230615(第2023年06期);第I136-184页 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN114385359A (en) | 2022-04-22 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Lu et al. | Optimization of lightweight task offloading strategy for mobile edge computing based on deep reinforcement learning | |
| Elgendy et al. | Joint computation offloading and task caching for multi-user and multi-task MEC systems: reinforcement learning-based algorithms | |
| CN107948083B (en) | SDN data center congestion control method based on reinforcement learning | |
| Ali et al. | Smart computational offloading for mobile edge computing in next-generation Internet of Things networks | |
| Li et al. | Method of resource estimation based on QoS in edge computing | |
| Li et al. | Multi-edge collaborative offloading and energy threshold-based task migration in mobile edge computing environment | |
| Li et al. | DQN-enabled content caching and quantum ant colony-based computation offloading in MEC | |
| Shukla et al. | FAT-ETO: Fuzzy-AHP-TOPSIS-Based efficient task offloading algorithm for scientific workflows in heterogeneous fog–cloud environment | |
| CN114385359B (en) | Cloud edge task time sequence cooperation method for Internet of things | |
| Mahan et al. | A novel resource productivity based on granular neural network in cloud computing | |
| CN113672372B (en) | Multi-edge collaborative load balancing task scheduling method based on reinforcement learning | |
| Németh et al. | A snapshot of the frontiers of client selection in federated learning | |
| Gu et al. | On Intelligent Placement Decision-Making Algorithms for Wireless Digital Twin Networks Via Bandit Learning | |
| Zhang et al. | Adaptive digital twin server deployment for dynamic edge networks in IoT system | |
| Lockhart et al. | Scission: Context-aware and performance-driven edge-based distributed deep neural networks | |
| Huang et al. | Temporal computing resource allocation scheme with end device assistance | |
| Liu | Animation art design online system based on mobile edge computing and user perception | |
| CN114615705B (en) | Single-user resource allocation strategy method based on 5G network | |
| CN106209978B (en) | Alliance relation service combination selection system and method | |
| Wakgra et al. | Multi-Objective Offloading Optimization in MEC and Vehicular-Fog Systems: A Distributed-TD3 Approach | |
| Zhou | A Heuristic Task Scheduling Strategy for Intelligent Manufacturing in the Big Data‐Driven Fog Computing Environment | |
| CN112579246B (en) | Virtual machine migration processing method and device | |
| Ren et al. | Balanced allocation method of physical education distance education resources based on linear prediction | |
| CN114462506A (en) | Communication network aided planning method supporting preference strategy and application thereof | |
| CN113568704A (en) | Container cloud task scheduling optimization method based on triple elite incubation algorithm |
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 |