CN114035671A - Cloud + multi-terminal cooperative energy-saving task unloading method for multi-working-mode wearable computing system - Google Patents

Cloud + multi-terminal cooperative energy-saving task unloading method for multi-working-mode wearable computing system Download PDF

Info

Publication number
CN114035671A
CN114035671A CN202111316851.3A CN202111316851A CN114035671A CN 114035671 A CN114035671 A CN 114035671A CN 202111316851 A CN202111316851 A CN 202111316851A CN 114035671 A CN114035671 A CN 114035671A
Authority
CN
China
Prior art keywords
wearable
task
control device
communication unit
computing system
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Pending
Application number
CN202111316851.3A
Other languages
Chinese (zh)
Inventor
赵蕴龙
王继梦
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Nanjing University of Aeronautics and Astronautics
Original Assignee
Nanjing University of Aeronautics and Astronautics
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by Nanjing University of Aeronautics and Astronautics filed Critical Nanjing University of Aeronautics and Astronautics
Priority to CN202111316851.3A priority Critical patent/CN114035671A/en
Publication of CN114035671A publication Critical patent/CN114035671A/en
Pending legal-status Critical Current

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F1/00—Details not covered by groups G06F3/00 - G06F13/00 and G06F21/00
    • G06F1/26—Power supply means, e.g. regulation thereof
    • G06F1/32—Means for saving power
    • G06F1/3203—Power management, i.e. event-based initiation of a power-saving mode
    • G06F1/3234—Power saving characterised by the action undertaken
    • G06F1/329—Power saving characterised by the action undertaken by task scheduling
    • 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
    • G06F9/44—Arrangements for executing specific programs
    • G06F9/445—Program loading or initiating
    • G06F9/44594—Unloading
    • 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]
    • G06F9/5061—Partitioning or combining of resources
    • G06F9/5072—Grid computing
    • 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

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Software Systems (AREA)
  • Physics & Mathematics (AREA)
  • General Engineering & Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Mathematical Physics (AREA)
  • Management, Administration, Business Operations System, And Electronic Commerce (AREA)

Abstract

The invention provides a cloud + multi-terminal collaborative energy-saving task unloading method for a multi-working-mode wearable computing system, which specifically comprises the following steps: modeling a system, and constructing a wearable computing system framework of cloud + multi-terminal cooperative computing; switching system working modes in real time according to working conditions of the wearable communication unit, determining available intelligent wearable equipment computing nodes, constructing an unloading decision model, and providing a task unloading optimization problem aiming at reducing energy consumption of wearable control equipment; converting the minimum problem of the average energy consumption of the wearable control equipment into the minimum problem of the Lyapunov drift wig of each time slot by utilizing the Lyapunov optimization theory; and finally, obtaining the optimal unloading strategy through a binary particle swarm algorithm. The task unloading algorithm provided by the invention can determine the execution position of the task in the time slot, and can reduce the energy consumption of the wearable control equipment and ensure the good use experience of the user while ensuring the stability of the system.

Description

Cloud + multi-terminal cooperative energy-saving task unloading method for multi-working-mode wearable computing system
Technical Field
The invention relates to the field of task scheduling, in particular to a cloud + multi-terminal cooperative energy-saving task unloading method for a multi-working-mode wearable computing system.
Background
The rapid development of microelectronic technology and computer technology has promoted the production and development of wearable computing devices, and wearable computing systems are a wearable, personalized, new-form personal mobile computing system, and can achieve natural and continuous assistance and enhancement for people. In the architecture of wearing computing systems, the computing system should be wearable like glasses, watches, or clothing and interact with the user as the context changes. The wearable computing system plays the role of an intelligent assistant for users through displays with different shapes, convenient and easy-to-use input equipment, a large number of environment perception components and wireless local area networks.
The computing unloading technology unloads the computing tasks to the sensor nodes, the edge server or the cloud server, so that the computing capacity of the wearable computing system can be improved, the response speed of the wearable computing system is improved to a certain extent, the average energy consumption of the wearable computing system is reduced, and the cruising ability of the system is improved. The wearable computer terminal can independently complete wearable application, and the computing task can be unloaded to other computing nodes for execution through the computing unloading technology only when the network resources and the computing resources of other computing nodes are good, so that the application has two modes of local execution and task unloading execution, and the computing unloading technology can be flexibly selected between the two modes. The indication of successful application uninstall is that the response time to the execution result of the application is completed within the time limit, because uninstalling the application beyond the time limit is meaningless and affects the user experience. The user satisfaction is related to the successful execution rate of the tasks in unit time (namely, the proportion of the tasks which are completed on time to the total tasks), and when the successful execution rate of the tasks is higher, the user satisfaction is higher.
The cloud server has abundant computing resources, and the mobile device can unload the tasks to the cloud server through a task unloading technology, so that the purpose of reducing computing delay or reducing computing energy consumption of the mobile terminal is achieved.
According to analysis of application scenes of wearing calculation, a wearing calculation system needs stronger cruising ability, and the wearing calculation system needs to support various applications of different types besides data acquisition work of various sensor devices, wherein the applications have different characteristics (data volume, expected completion time and calculation complexity), can be divided into three types according to the application calculation complexity and data transmission quantity, and are high in calculation complexity and relatively high in data transmission quantity, such as target detection and face detection; applications with high computational complexity and relatively low data transmission volume such as motion recognition; applications with low computational complexity and relatively large data transmission volume, such as data fusion. Different offloading decisions are made for different types of applications.
In recent years, the communication technology in China enters a rapid development stage, the 4G technology is mature, the 5G technology is in a leading position in the technical field of wireless communication, other communication technologies such as Tiantong and WiFi exist, and due to the deployment problem of different communication technologies, the service quality provided by different communication modes may be different at a certain position, so that the communication modes need to be adjusted in real time, the excellent information service requirements of individuals are met, and the personalized requirements in the Internet age are met.
Therefore, the invention provides a cloud + multi-terminal collaborative computing task computing framework for a multi-working mode wearable computing system, and the problem that the task needs to be considered is that due to the difference of computing resources and working conditions of all nodes and the diversified requirements of users, a proper computing node is selected to execute the task and the good use experience of the users is guaranteed.
Disclosure of Invention
According to the defects of the prior art and the characteristics of a wearable computing system, the invention aims to provide a cloud + multi-terminal cooperation energy-saving task unloading method for the wearable computing system with multiple working modes. In order to achieve the purpose of the invention, the invention adopts the following technical scheme: a cloud + multi-terminal cooperative energy-saving task unloading method for a multi-working mode wearable computing system.
Drawings
FIG. 1 is a diagram of a system model of the present invention;
FIG. 2 is a flow chart of a cloud + multi-end collaborative energy-saving task offloading method for a multi-working mode wearable computing system according to the present invention;
FIG. 3 is a No _ WCU mode of a wearable computing system designed in accordance with the present invention;
FIG. 4 is a diagram of the Width _ WCU mode of a wearable computing system designed in accordance With the present invention;
Detailed Description
In order that the present invention may be more readily and clearly understood, reference is now made to the following detailed description of the invention taken in conjunction with the accompanying drawings, in which:
the technical scheme provided by the invention is a cloud + multi-terminal cooperative energy-saving task unloading method for a multi-working mode wearable computing system, which specifically comprises the following steps:
the method comprises the following steps that firstly, a wearable computing system is modeled, the wearable computing system comprises intelligent wearable equipment, control unit nodes, a wearable communication unit and remote server nodes, and communication links exist among the nodes;
step two, constructing a task model, constructing a wearable control device, an intelligent wearable device, a wearable communication unit, a remote server calculation model and a communication model, generating calculation tasks on the wearable control device, and executing the tasks in series according to a task generation sequence;
determining the working mode of the wearable computing system according to the working condition of the wearable communication unit;
determining a working node which can be used for task unloading according to the working condition and the resource condition of the intelligent wearable equipment, and constructing an unloading decision model;
constructing an optimization target to obtain the execution energy consumption of the wearable control equipment, and constructing the minimum problem of the average execution energy consumption of the wearable control equipment;
step six, eliminating the user satisfaction limit through Lyapunov optimization, and converting the average energy consumption minimization problem of the wearable control equipment into a Lyapunov drift and penalty minimization problem;
and seventhly, solving the optimal calculation unloading decision of the wearable control equipment through a binary particle swarm algorithm.
Further, the step one is a specific process for modeling the wearable computing system as follows:
(1.1) the Wearable computing system is divided into four layers as shown in FIG. 1, including a Smart Wearable Device (WSD), a Wearable Control Device (WCD), a Wearable Communication Unit (WCU), and a Remote Server (RS). The main features of these four layers are summarized below:
(1.2) Intelligent wearable device
The intelligent wearable equipment mainly comprises sensor equipment worn by a user, such as an intelligent camera, an intelligent bracelet, an intelligent vest and a heart rate detector. The intelligent wearable device is connected with the wearable control device through WiFi or Bluetooth, data transmission is achieved, and the working condition and the resource condition of the current device are fed back to the wearable control device at regular time.
(1.3) wearable control device
Wearing controlgear includes mobile devices such as user's smart mobile phone or panel computer, these equipment link to each other with intelligent wearing equipment, collect each wearing equipment's information, including humiture, video recording, each item physical characteristics etc. simultaneously, each intelligent wearing equipment transmits its resource condition and the behavior in current time slot for wearing controlgear, this is just on a point to wearing controlgear in each item information centralization, consequently, this paper will wear controlgear as task uninstallation decision node, wear controlgear local also can carry out certain task simultaneously.
Wearing controlgear and wearing the communication unit through wiFi and establish and be connected, wearing controlgear can pass through cellular mobile network technique and cloud server communication simultaneously, so use the reduction to wear controlgear energy consumption as the optimization target here, neglect and wear the communication unit energy consumption, under the condition that wears the communication unit and become invalid, wear controlgear still can be through 4G or 5G and remote server communication.
(1.4) wearable communication Unit
The wearable communication unit mainly comprises a radio station, a router or a base station, the wearable communication unit mainly serves to provide a communication medium for the wearable control equipment, can receive data of the wearable control equipment and forward the data to a remote server, and meanwhile, the node also has data processing capacity and can be used as a selectable node for task unloading. The wearable computing system can work in a No _ WCU mode and a Width _ WCU mode according to the working condition of the wearable communication device, the wearable computing system can establish connection With a remote server through a wired network (optical fibers and the like), and the wearable communication unit forwards the wearable computing unit to a cloud server With rich resources for complex computing tasks.
(1.5) remote Server
The remote server is a traditional centralized cloud server, which can provide rich computing resources and applications for mobile end users, and when a computing task needs more computing resources and the amount of data is small, such as target detection, action recognition, and the like, the task can be optionally offloaded to the remote server for execution.
Further, the specific process of constructing the task model and the calculation model of each calculation node in the second step is as follows:
(2.1) construction of task model
In a wearable computing system, a time slot model with running time set as equal-length time slots is set, the time slot length is set as tau, a time index is set as T ═ 0, 1, 2, 3.., N }, and a task set T containing N tasks is generated by wearable control equipment in the time slot Tt＝{t1，t2，t3，t4，…，tnN tasks are independent and have no relation, and the tasks tiIs set to (C)i，Di，Fi) In which C isiAs task tiCalculated amount of (D)iAs task tiAmount of data transmission of FiAs task tiThe delay limit of (2).
(2.2) wearable control equipment calculation model
At the wearable control device, some basic tasks with small calculation amount are supported and processed due to the limit of the battery capacity. Meanwhile, the wearable control device is used as a task generation node, the data to be processed and the task request are both located at the wearable control device, and the wearable control device is required to make a real-time decision by using limited computing resources. The execution time spent by the task when the wearable control device normally processes is represented as:
Figure BSA0000257095760000031
wherein f isWCDRepresenting the calculated frequency of the wearing control device, so that the task consumes energy to perform at the wearing control device
Figure BSA0000257095760000032
Comprises the following steps:
Figure BSA0000257095760000033
wherein P isWCDRepresents the energy consumption of the wearable control device in unit time for executing the task.
(2.3) remote Server computing model
The remote server can provide richer computing resources compared with other nodes, but the data transmission delay is higher than that of other nodes, and when the task is unloaded to the remote server node, the execution delay of the task is as follows:
Figure BSA0000257095760000041
wherein f isRSExpressed as the computing power of the remote server node, DRSRepresenting a transmission delay of data from the wearable control device to the remote server node, the transmission delay being represented as follows:
Figure BSA0000257095760000042
wherein B is1Expressed as the bandwidth between the wearable control device and the wearable communication unit, B2Expressed as the bandwidth between the worn communication unit and the remote server, we generally consider the transmission speed within the local area network to be greater than the transmission speed of the wide area network.
When the task selection is unloaded to the remote server node, the energy consumption generated by the wearable control device is mainly generated by data communication between the wearable control device and the wearable communication unit, and is specifically represented as follows:
Figure BSA0000257095760000043
wherein P isrRepresents the energy consumption of the communication between the wearable control device and the wearable communication unit in unit time.
(2.4) wearable communication unit calculation model
Compared with a remote server, the wearable communication unit has the advantages that the computing resources are reduced, the wearable control device is close to the wearable communication unit, the data transmission delay is low, and the time spent in unloading the task to the wearable communication unit is as follows:
Figure BSA0000257095760000044
wherein f isWCURepresenting the computational power of the wearing communication unit, DWCURepresenting a transmission delay of data transmitted from the wearable control device to the wearable communication unit, the transmission delay being represented as follows:
Figure BSA0000257095760000045
wherein B is1Expressed as the bandwidth between the wearable control device and the wearable communication unit.
The energy consumption generated by unloading the task to the wearable communication unit mainly comprises the energy consumption generated by uploading data to the wearable communication unit by the wearable control device, and is represented as follows:
Figure BSA0000257095760000046
wherein P isrRepresents the energy consumption of the communication between the wearable control device and the wearable communication unit in unit time.
(2.5) intelligent wearable equipment calculation model
In intelligent wearing equipment, some sensor terminals have the computing power, can carry out simple data preprocessing work as the calculation node, can help wearing controlgear to carry out partly task when necessary, and when the task was uninstalled to intelligent wearing equipment, the execution delay of task was:
Figure BSA0000257095760000047
wherein f isWSDRepresenting the computing power of the smart wearable device, DWSDThe transmission delay of data transmission from the wearable control device to the intelligent wearable device is represented as follows:
Figure BSA0000257095760000048
wherein B is0Expressed as the bandwidth between the wearable control device and the intelligent wearable device.
The energy consumption generated by unloading the task to the intelligent wearable device mainly comprises the energy consumption generated by transmitting data to the intelligent wearable device by the wearable control device, and is represented as follows:
Figure BSA0000257095760000051
wherein P issShow wearing control deviceEnergy consumption of communication between the equipment and the intelligent wearable equipment in unit time.
Further, the three specific processes of selecting the communication mode according to the working condition of the node of the wearable communication unit are as follows:
(3.1) the wearable control device determines the working mode of the whole wearable computing system by detecting the working condition of the wearable communication unit in the time slot t in real time, and when the wearable communication unit cannot work normally, the working mode of the wearable computing system is set to be the No _ WCU mode, as shown in fig. 3, and when the wearable communication unit works normally, the working mode of the wearable computing system is set to be the Width _ WCU mode, as shown in fig. 4;
further, the fourth step of determining a working node which can be used for task unloading according to the working condition and the resource condition of the intelligent wearable device, and the specific process of constructing the unloading decision model is as follows:
(4.1) the wearable control device determines an available task unloading node set R at the current time slot according to the working condition and the available resource condition of the intelligent wearable device in the time slot t:
R＝{WSD1，WSD2，…，WSDm}
(4.2) when the working mode of the wearable computing system is in a No _ WCU mode and the set of task unloading nodes available for the intelligent wearable device is R, the unloading decision model is as follows:
task tiThe offloading decision is represented as follows:
Figure BSA0000257095760000052
wherein
Figure BSA0000257095760000057
Representing a task tiWhether the operation is executed at the node S, wherein S is an available computing node in the current working mode, and in the case of the mode one, S belongs to the group of R U { WCD, RS };
Figure BSA0000257095760000053
so in time slot t, the offload decision matrix for n tasks is I (t) ═ I1(t)，I2(t)，…，Ii(t)，…，In(t) }. An offload decision matrix I (t) is shown below:
Figure BSA0000257095760000054
(4.3) when the working mode of the wearable computing system is in a With _ WCU mode and the task unloading node available for the intelligent wearable device is R, the unloading decision model is as follows:
task tiThe offloading decision is represented as follows:
Figure BSA0000257095760000055
wherein
Figure BSA0000257095760000058
Indicating whether the task ti is executed at the S node, wherein S belongs to R U { WSD, WCD, WCU, RS };
Figure BSA0000257095760000056
so in time slot t, the offload decision matrix for n tasks is I (t) ═ I1(t)，I2(t)，…，Ii(t)，…，In(t) }. An offload decision matrix I (t) is shown below:
Figure BSA0000257095760000061
further, the concrete process of constructing the optimization target in the fifth step is as follows:
(5.1) first, an execution task t is obtainediThe energy consumption generated is as follows:
Figure BSA0000257095760000062
the energy consumption for processing all tasks in a time slot t is expressed as follows:
Figure BSA0000257095760000063
(5.2) in order to ensure the user's good use experience, the present invention defines an indicator variable σ (i (t)) to indicate whether the user is satisfied with the on-time completion rate of the task in the time slot t, i.e. whether the user's experience is good:
Figure BSA0000257095760000064
wherein P (i) (t) is task _ finish/task _ total, which is a task on-time completion rate and indicates a task probability of completing within a deadline time within a time interval t, P is a preset task on-time completion rate, σ (i) (t)) is 0, which indicates that the task on-time completion rate meets a requirement, user experience is good, and σ (i (t)) is 1, which indicates that the task on-time completion rate does not meet the requirement, and user experience is not good.
(5.3) the present invention defines a user satisfaction factor S to represent the proportion of time slots that a user is satisfied within the so-used time slots, expressed as follows:
Figure BSA0000257095760000065
(5.4) in conclusion, aiming at the characteristics of the wearable computing system, in order to reduce the energy consumption of the wearable control device and ensure the good use experience of the user, the invention sets out an optimization problem:
Figure BSA0000257095760000066
where ω represents the user's dissatisfaction with the slot coefficient.
Further, the sixth step eliminates the user satisfaction limit through lyapunov optimization, and converts the average energy consumption minimization problem of the wearable control device into a lyapunov drift and penalty minimization problem, wherein the specific process is as follows:
(6.1) in order to satisfy the long-term constraint condition of user satisfaction, the operation satisfying the constraint condition is decomposed into each time slice to be completed, and a virtual queue Q (t) is defined:
Q(t+1)＝max[Q(t)-ω，0]+σ(I(t)) (23)
wherein Q (t) represents the queue backlog of the system at that time, and Q (t) is initially set to 0, and when Q (t) is greater,
meaning that the longer the response time of the task, the lower the user satisfaction.
(6.2) according to Lyapunov optimization, in order to guarantee the queue Q (t) stability, the following Lyapunov equation is defined:
Figure BSA0000257095760000067
(6.3) the change per execution of the operation is the Lyapunov drift, expressed as:
Figure BSA0000257095760000071
(6.4) defining the conditional lyapunov drift Δ (q (t)) in view of the current system state q (t) as:
Figure BSA0000257095760000072
(6.5) the drift-plus-penalty equation for each interval is:
Δ(Q(t))+VE{E(I(t))|Q(t)} (27)
(6.6) where V is a trade-off parameter between queue stability and objective function optimality. Substituting formula (26) for formula (27) to obtain:
Figure BSA0000257095760000073
(6.7) therefore, the original optimization objective is herein converted into:
minI(t)Δ(Q(t))+VE{E(I(t))|Q(t)} (29)
(6.8) since the equation contains quadratic terms, it is difficult to solve, and it is converted into the problem of solving the upper limit of the objective function, as shown in equation (28), because omega is a constant,
Figure BSA0000257095760000074
since the control variable is not included, the upper limit problem of equation (28) is converted to
minI(t)[VE(I(t))+Q(t)σ(t)] (30)
(6.9) therefore the offload decision function D (Q (t), I (t)) is defined as
D(Q(t)，I(t))＝VE(I(t))+Q(t)σ(t) (31)
The decision function D (Q (t), I (t)) is minimized by selecting a reasonable unloading decision I (t), which transforms the original NP-difficult problem into the current simpler problem.
Further, the specific process of solving the optimal calculation unloading decision of the wearable control device through the binary particle swarm algorithm in the seventh step is as follows:
(7.1) generating an initial population, wherein the initial population of the invention adopts a random generation mode, and the mode has the advantages of uniform distribution and capability of ensuring particle richness.
And (7.2) setting a fitness function as the decision function D (Q) (t), I (t)) provided by the previous step, and recording the individual historical optimal position of each particle and the historical optimal position of the group.
(7.3) the speed and position are updated as follows:
Figure BSA0000257095760000081
sigm(Vt+1)＝1/[1+exp(-Vt+1)] (33)
and (7.4) iterating for 100 times, checking whether the iteration requirement is met or the maximum iteration time is reached, and outputting an optimal unloading decision.

Claims (8)

1. A cloud + multi-end cooperative energy-saving task unloading method for a multi-working mode wearable computing system is characterized by comprising the following steps: the method comprises the following specific steps:
the method comprises the following steps that firstly, a wearable computing system is modeled, the wearable computing system comprises intelligent wearable equipment, control unit nodes, a wearable communication unit and remote server nodes, and communication links exist among the nodes;
step two, constructing a task model, constructing a wearable control device, an intelligent wearable device, a wearable communication unit, a remote server calculation model and a communication model, generating calculation tasks on the wearable control device, and executing the tasks in series according to a task generation sequence;
determining the working mode of the wearable computing system according to the working condition of the wearable communication unit;
determining a working node which can be used for task unloading according to the working condition and the resource condition of the intelligent wearable equipment, and constructing an unloading decision model;
constructing an optimization target to obtain the execution energy consumption of the wearable control equipment, and constructing the minimum problem of the average execution energy consumption of the wearable control equipment;
step six, eliminating the user satisfaction limit through Lyapunov optimization, and converting the average energy consumption minimization problem of the wearable control equipment into a Lyapunov drift and penalty minimization problem;
and seventhly, solving the optimal calculation unloading decision of the wearable control equipment through a binary particle swarm algorithm.
2. The energy-saving task unloading method oriented to cloud + multi-terminal collaboration of the multi-working-mode wearable computing system according to claim 1, characterized in that: and in the first step, a task unloading model of the wearable computing system is constructed according to the characteristics of the wearable computing system. The specific process is as follows:
1) the Wearable computing system is divided into four layers, including a Smart Wearable Device (WSD), a Wearable Control Device (WCD), a Wearable Communication Unit (WCU), and a Remote Server (RS). The main features of these four layers are summarized below:
a) intelligent wearable device
The intelligent wearable equipment mainly comprises sensor equipment worn by a user, such as an intelligent camera, an intelligent bracelet, an intelligent vest and a heart rate detector. The intelligent wearable device is connected with the wearable control device through WiFi or Bluetooth, data transmission is achieved, and the working condition and the resource condition of the current device are fed back to the wearable control device at regular time.
b) Wearable control device
Wearing controlgear includes mobile devices such as user's smart mobile phone or panel computer, these equipment link to each other with intelligent wearing equipment, collect each wearing equipment's information, including humiture, video recording, each item physical characteristics etc. simultaneously, each intelligent wearing equipment transmits its resource condition and the behavior in current time slot for wearing controlgear, this is just on a point to wearing controlgear in each item information centralization, consequently, this paper will wear controlgear as task uninstallation decision node, wear controlgear local also can carry out certain task simultaneously.
Wearing controlgear and wearing the communication unit through wiFi and establish and be connected, wearing controlgear can pass through cellular mobile network technique and cloud server communication simultaneously, so use the reduction to wear controlgear energy consumption as the optimization target here, neglect and wear the communication unit energy consumption, under the condition that wears the communication unit and become invalid, wear controlgear still can be through 4G or 5G and remote server communication.
c) Wearable communication unit
The wearable communication unit mainly comprises a radio station, a router or a base station, the wearable communication unit mainly serves to provide a communication medium for the wearable control equipment, can receive data of the wearable control equipment and forward the data to a remote server, and meanwhile, the node also has data processing capacity and can be used as a selectable node for task unloading. The wearable computing system can work in a No _ WCU mode and a Width _ WCU mode according to the working condition of the wearable communication device, the wearable computing system can establish connection With a remote server through a wired network (optical fibers and the like), and the wearable communication unit forwards the wearable computing unit to a cloud server With rich resources for complex computing tasks.
d) Remote server
The remote server is a traditional centralized cloud server, which can provide rich computing resources and applications for mobile end users, and when a computing task needs more computing resources and the amount of data is small, such as target detection, action recognition, and the like, the task can be optionally offloaded to the remote server for execution.
3. The energy-saving task unloading method oriented to cloud + multi-terminal collaboration of the multi-working-mode wearable computing system according to claim 1, characterized in that: and step two, constructing a task model, and constructing a calculation model and a communication model of each node, wherein the specific process is as follows:
1) constructing task models
In a wearable computing system, a time slot model with running time set as equal-length time slots is set, the time slot length is set as tau, a time index is set as T ═ 0, 1, 2, 3.., N }, and a task set T containing N tasks is generated by wearable control equipment in the time slot Tt＝{t1，t2，t3，t4，…，tnN tasks are independent and have no relation, and the tasks tiIs set to (C)i，Di，Fi) In which C isiAs task tiCalculated amount of (D)iAs task tiAmount of data transmission of FiAs task tiThe delay limit of (2).
2) Wearable control equipment calculation model
At the wearable control device, only some basic tasks are supported that are computationally small due to the limitation of battery capacity. Meanwhile, the wearable control device is used as a task generation node, the data to be processed and the task request are both located at the wearable control device, and the wearable control device is required to make a real-time decision by using limited computing resources. The execution time spent by the task when the wearable control device normally processes is represented as:
Figure FSA0000257095750000021
wherein f isWCDRepresenting the calculated frequency of the wearing control device, so that the task consumes energy to perform at the wearing control device
Figure FSA0000257095750000022
Comprises the following steps:
Figure FSA0000257095750000023
wherein P isWCDRepresents the energy consumption of the wearable control device in unit time for executing the task.
3) Remote server computing model
The remote server can provide richer computing resources compared with other nodes, but the data transmission delay is higher than that of other nodes, and when the task is unloaded to the remote server node, the execution delay of the task is as follows:
Figure FSA0000257095750000024
wherein f isRSExpressed as the computing power of the remote server node, DRSRepresenting a transmission delay of data from the wearable control device to the remote server node, the transmission delay being represented as follows:
Figure FSA0000257095750000025
wherein B is1Expressed as the bandwidth between the wearable control device and the wearable communication unit, B2Expressed as the bandwidth between the worn communication unit and the remote server, we generally consider the transmission speed within the local area network to be greater than the transmission speed of the wide area network.
When the task selection is unloaded to the remote server node, the energy consumption generated by the wearable control device is mainly generated by data communication between the wearable control device and the wearable communication unit, and is specifically represented as follows:
Figure FSA0000257095750000031
wherein P isrRepresents the energy consumption of the communication between the wearable control device and the wearable communication unit in unit time.
4) Wearable communication unit calculation model
Compared with a remote server, the wearable communication unit has the advantages that the computing resources are reduced, the wearable control device is close to the wearable communication unit, the data transmission delay is low, and the time spent in unloading the task to the wearable communication unit is as follows:
Figure FSA0000257095750000032
wherein f isWCURepresenting the computational power of the wearing communication unit, DWCURepresenting a transmission delay of data transmitted from the wearable control device to the wearable communication unit, the transmission delay being represented as follows:
Figure FSA0000257095750000033
wherein B is1Expressed as the bandwidth between the wearable control device and the wearable communication unit.
The energy consumption generated by unloading the task to the wearable communication unit mainly comprises the energy consumption generated by uploading data to the wearable communication unit by the wearable control device, and is represented as follows:
Figure FSA0000257095750000034
wherein P isrRepresents the energy consumption of the communication between the wearable control device and the wearable communication unit in unit time.
5) Constructing a computing model of an intelligent wearable device
In intelligent wearing equipment, some sensor terminals have the computing power, can carry out simple data preprocessing work as the calculation node, can help wearing controlgear to carry out partly task when necessary, and when the task was uninstalled to intelligent wearing equipment, the execution delay of task was:
Figure FSA0000257095750000035
wherein f isWSDRepresenting the computing power of the smart wearable device, DWSDThe transmission delay of data transmission from the wearable control device to the intelligent wearable device is represented as follows:
Figure FSA0000257095750000036
wherein B is0Expressed as the bandwidth between the wearable control device and the intelligent wearable device.
The energy consumption generated by unloading the task to the intelligent wearable device mainly comprises the energy consumption generated by transmitting data to the intelligent wearable device by the wearable control device, and is represented as follows:
Figure FSA0000257095750000037
wherein P issThe energy consumption of communication between the wearable control device and the intelligent wearable device in unit time is represented.
4. The energy-saving task unloading method oriented to cloud + multi-terminal collaboration of the multi-working-mode wearable computing system according to claim 1, characterized in that: selecting a communication mode according to the working condition of the nodes of the wearable communication unit in the third step, wherein the specific process comprises the following steps:
1) the wearable control device determines the working mode of the whole wearable computing system by detecting the working condition of the wearable communication unit in the time slot t in real time, the working mode of the wearable computing system is set to be the No _ WCU mode under the condition that the wearable communication unit cannot work normally, and the working mode of the wearable computing system is set to be the Path _ WCU mode under the condition that the wearable communication unit works normally.
5. The energy-saving task unloading method oriented to cloud + multi-terminal collaboration of the multi-working-mode wearable computing system according to claim 1, characterized in that: in the fourth step, according to the working condition and the resource condition of the intelligent wearable device, a working node which can be used for task unloading is determined, and an unloading decision model is constructed, wherein the specific process is as follows:
1) the wearable control equipment determines an available task unloading node set R under the current time slot according to the working condition and the available resource condition of the intelligent wearable equipment in the time slot t:
R＝{WSD1，WSD2，…，WSDm}
2) when the working mode of the wearable computing system is in a No _ WCU mode and the set of task unloading nodes available for the intelligent wearable device is R, the unloading decision model is as follows:
task tiThe offloading decision is represented as follows:
Figure FSA0000257095750000041
wherein
Figure FSA0000257095750000047
Representing a task tiWhether the operation is executed at the node S, wherein S is an available computing node in the current working mode, and in the case of the mode one, S belongs to the group of R U { WCD, RS };
Figure FSA0000257095750000042
so in time slot t, the offload decision matrix for n tasks is I (t) ═ I1(t)，I2(t)，…，Ii(t)，…，In(t) }. An offload decision matrix I (t) is shown below:
Figure FSA0000257095750000043
3) when the working mode of the wearable computing system is in a Width _ WCU mode and the available task unloading node of the intelligent wearable device is R, the unloading decision model is as follows:
task tiThe offloading decision is represented as follows:
Figure FSA0000257095750000044
wherein
Figure FSA0000257095750000048
Representing a task tiWhether to execute in S node, where S ∈ R { WSD, WCD, WCU, RS };
Figure FSA0000257095750000045
so in time slot t, the offload decision matrix for n tasks is I (t) ═ I1(t)，I2(t)，…，Ii(t)，…，In(t) }. An offload decision matrix I (t) is shown below:
Figure FSA0000257095750000046
6. the energy-saving task unloading method oriented to cloud + multi-terminal collaboration of the multi-working-mode wearable computing system according to claim 1, characterized in that: and in the fifth step, an optimization target is constructed to obtain the execution energy consumption of the wearable control equipment, and the problem of minimizing the average execution energy consumption of the wearable control equipment is constructed, wherein the concrete process is as follows:
1) first, an execution task t is obtainediThe energy consumption generated is as follows:
Figure FSA0000257095750000051
the energy consumption for processing all tasks in a time slot t is expressed as follows:
Figure FSA0000257095750000052
2) in order to ensure a good user experience, the present invention defines an indicator variable σ (i (t)) to indicate whether the user is satisfied with the on-time completion rate of the task in the time slot t, i.e. whether the user experience is good:
Figure FSA0000257095750000053
wherein P (i) (t) is task _ finish/task _ total, which indicates the probability of task completion within the deadline time t, ρ is a preset task on-time completion rate, σ (i (t)) is 0, which indicates that the task on-time completion rate meets the requirement, user experience is good, and σ (i (t)) is 1, which indicates that the task on-time completion rate does not meet the requirement, and user experience is poor.
3) The invention defines a user satisfaction coefficient S to represent the satisfactory time slot proportion of the user in the used time slot, and the user satisfaction coefficient S is represented as follows:
Figure FSA0000257095750000054
4) in summary, aiming at the characteristics of the wearable computing system, in order to reduce the energy consumption of the wearable control device and ensure the good use experience of the user, the invention solves the optimization problem:
Figure FSA0000257095750000055
Figure FSA0000257095750000056
where ω represents the user's dissatisfaction with the slot coefficient.
7. The energy-saving task unloading method oriented to cloud + multi-terminal collaboration of the multi-working-mode wearable computing system according to claim 1, characterized in that: in the sixth step, the user satisfaction limit is eliminated through Lyapunov optimization, the average energy consumption minimization problem of the wearable control equipment is converted into a Lyapunov drift and penalty minimization problem, and the specific process is as follows:
1) in order to satisfy the long-term constraint condition of user satisfaction, the operation satisfying the constraint condition is decomposed into each time slice to be completed, and a virtual queue Q (t) is defined:
Q(t+1)＝max[Q(t)-ω，0]+σ(I(t)) (23)
wherein q (t) represents the queue backlog of the system at the moment, the initial value of q (t) is defined as 0, and when q (t) is larger, the longer the response time of the task is, the lower the satisfaction degree of the user is.
2) According to lyapunov optimization, to ensure queue q (t) stability, the following lyapunov equation is defined:
Figure FSA0000257095750000057
3) the change produced by each execution of the operation is lyapunov drift, which is expressed as:
Figure FSA0000257095750000058
4) considering the current system state q (t), the conditional lyapunov drift Δ (q (t)) is defined as:
Figure FSA0000257095750000059
Figure FSA0000257095750000061
5) the drift-plus-penalty equation for each interval is:
Δ(Q(t))+VE{E(I(t))|Q(t)} (27)
6) where V is a trade-off parameter between queue stability and objective function optimality. Substituting formula (26) for formula (27) to obtain:
Figure FSA0000257095750000062
7) therefore, the original optimization objective is converted into:
minI(t)Δ(Q(t))+VE{E(I(t))|Q(t)} (29)
8) because the equation contains quadratic terms, which are difficult to solve, the quadratic terms are converted into the problem of solving the upper limit of the objective function, as shown in equation (28), because omega is a constant,
Figure FSA0000257095750000063
since the control variable is not included, the upper limit problem of equation (28) is converted to
minI(t)[VE(I(t))+Q(t)σ(t)] (30)
9) The offload decision function D (Q (t), I (t)) is thus defined as
D(Q(t)，I(t))＝VE(I(t))+Q(t)σ(t) (31)
The decision function D (Q (t), I (t)) is minimized by selecting a reasonable unloading decision I (t), which transforms the original NP-difficult problem into the current simpler problem.
8. The energy-saving task unloading method oriented to cloud + multi-terminal collaboration of the multi-working-mode wearable computing system according to claim 1, characterized in that: in the seventh step, the optimal calculation unloading decision of the wearable control device is solved through a binary particle swarm algorithm, and the specific process is as follows:
1) the initial population is generated, and the initial population of the invention adopts a random generation mode, which has the advantages of uniform distribution and particle richness guarantee.
2) And setting a fitness function as the decision function D (Q (t), I (t)) provided by the previous step, and recording the individual historical optimal position of each particle and the historical optimal position of the group.
3) The speed and position are updated as follows:
Figure FSA0000257095750000064
sigm(Vt+1)＝1/[1+exp(-Vt+1)] (33)
4) and (5) iterating for 100 times, checking whether the iteration requirement is met or the maximum iteration number is reached, and outputting an optimal unloading decision.
CN202111316851.3A 2021-11-08 2021-11-08 Cloud + multi-terminal cooperative energy-saving task unloading method for multi-working-mode wearable computing system Pending CN114035671A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202111316851.3A CN114035671A (en) 2021-11-08 2021-11-08 Cloud + multi-terminal cooperative energy-saving task unloading method for multi-working-mode wearable computing system

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202111316851.3A CN114035671A (en) 2021-11-08 2021-11-08 Cloud + multi-terminal cooperative energy-saving task unloading method for multi-working-mode wearable computing system

Publications (1)

Publication Number Publication Date
CN114035671A true CN114035671A (en) 2022-02-11

Family

ID=80136782

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202111316851.3A Pending CN114035671A (en) 2021-11-08 2021-11-08 Cloud + multi-terminal cooperative energy-saving task unloading method for multi-working-mode wearable computing system

Country Status (1)

Country Link
CN (1) CN114035671A (en)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN115209514A (en) * 2022-09-16 2022-10-18 荣耀终端有限公司 Method for closing cellular communication function and related electronic equipment

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN115209514A (en) * 2022-09-16 2022-10-18 荣耀终端有限公司 Method for closing cellular communication function and related electronic equipment

Similar Documents

Publication Publication Date Title
CN110856134B (en) Large-scale wireless sensor network data collection method based on unmanned aerial vehicle
CN109829332B (en) Joint calculation unloading method and device based on energy collection technology
CN112888002B (en) Game theory-based mobile edge computing task unloading and resource allocation method
CN110267292B (en) Cellular network flow prediction method based on three-dimensional convolutional neural network
Lu et al. Optimization of task offloading strategy for mobile edge computing based on multi-agent deep reinforcement learning
CN112633316B (en) Load prediction method and device based on boundary estimation theory
CN113230658A (en) Resource allocation method and device, computer readable medium and electronic equipment
CN113238847B (en) Distribution and scheduling method based on distributed network environment and capable of distributing tasks
CN115175217A (en) Resource allocation and task unloading optimization method based on multiple intelligent agents
CN116489708B (en) Meta universe oriented cloud edge end collaborative mobile edge computing task unloading method
Chen et al. Multi-user edge-assisted video analytics task offloading game based on deep reinforcement learning
CN114035671A (en) Cloud + multi-terminal cooperative energy-saving task unloading method for multi-working-mode wearable computing system
CN115081946A (en) Power consumption regulation and control method, system, device, electronic equipment and computer medium
Tang et al. Layer dependency-aware learning scheduling algorithms for containers in mobile edge computing
CN113709385A (en) Video processing method and device, computer equipment and storage medium
Jin et al. Optimal policy characterization enhanced proximal policy optimization for multitask scheduling in cloud computing
CN114116061B (en) Workflow task unloading method and system in mobile edge computing environment
Ling et al. Dynamic age minimization with real-time information preprocessing for edge-assisted IoT devices with energy harvesting
CN115408072A (en) Rapid adaptation model construction method based on deep reinforcement learning and related device
Li et al. Task computation offloading for multi-access edge computing via attention communication deep reinforcement learning
Huang et al. Service reservation and pricing for green metaverses: A stackelberg game approach
CN113382066A (en) Vehicle user selection method and system based on federal edge platform
CN116185523A (en) Task unloading and deployment method
CN115577805A (en) Dynamic resource cooperation method and system for data training
CN115051374A (en) Control method and device for electric heating equipment to participate in electric power peak regulation and storage medium

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication