Wireless energy supply mobile edge computing unloading method and system based on deep learning

By using deep learning to optimize link selection and transmission power in wireless power-supply mobile edge computing networks, the delay problem caused by the "far and near effects" is solved, and more efficient energy collection and data offloading is achieved, meeting the immediate needs.

CN119071852APending Publication Date: 2024-12-03SHAANXI NORMAL UNIV
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202411015257.4
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-07-26
Publication Date
2024-12-03

AI Technical Summary

Technical Problem

There is a "fast and near effect" in existing wireless power-supply mobile edge computing networks, resulting in users far away from access points to get less energy and longer communication distances, resulting in long task transmission time, unable to meet immediate needs, and no deep learning is considered Impact on task unloading.

Method used

Using deep learning-based methods, transmission constraints, energy constraints, link capacity constraints and offload time constraints are constructed, and the link selection and transmission are optimized through segmented linearization and mixed integer linear planning models. Power to minimize system delay.

Benefits of technology

Through joint optimization of node collaborative transmission and deep learning framework, data offload delay is reduced, data offload delay of "far" nodes is improved, deployment costs in relay collaboration mode are reduced, and energy collection and data offloading is minimized. Time overhead.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119071852A_ABST
    Figure CN119071852A_ABST
Patent Text Reader

Abstract

The invention provides a wireless energy supply mobile edge computing unloading method and a wireless energy supply mobile edge computing unloading system based on deep learning. The method comprises the following steps: collecting energy from a hybrid access point (AP), and then directly unloading tasks needing to be unloaded to the AP or unloading the tasks needing to be unloaded to the AP through equipment; constructing a transmission constraint, an energy constraint, a link capacity constraint, a flow constraint, a quantity conservation constraint and an unloading time constraint in an unloading process; constructing an objective function under all constraints, converting a problem into a mixed integer linear programming model MILP problem by utilizing piecewise linearization PWL, and solving according to a BCMD approximation algorithm to obtain a problem approximate solution; transmitting power is predicted according to a DLOO algorithm, a problem is converted into an LP problem, a feasible solution is obtained, and the minimum time expenditure needed by energy collection and data unloading is calculated. An approximation algorithm based on BCMD and a power prediction algorithm based on DL are provided, and link selection and transmission power optimization in the task unloading process are achieved so as to achieve the purpose of minimizing system time delay.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention belongs to the field of wireless power supply network optimization, and in particular to a wireless power supply mobile edge computing unloading method and system based on deep learning. Background Art

[0002] A wireless power supply network is a technical system that transmits energy in the air through electromagnetic fields. This technology eliminates the physical wires or cables required in traditional power transmission, thereby providing more flexibility and convenience. Its core principle is to use electromagnetic induction, magnetic resonance or microwave transmission technology to transmit electrical energy from the power source to the receiving end. It is to add an energy supply module to the mobile edge computing network to meet the problem of limited battery life of user devices. The purpose is to use wireless energy to charge the device or support device communication. The application scenarios of wireless power supply networks are wide, including consumer electronic devices, electric vehicles, medical equipment, industrial automation, etc.

[0003] Deep learning is a branch of machine learning that automatically extracts features and recognizes patterns based on a multi-layer neural network structure. It simulates the neuronal connections of the human brain and can achieve high-precision prediction and classification in the fields of images, speech, and text through training with large amounts of data. It has been applied in the field of wireless networks, such as routing, location estimation, link selection, and load balancing. Due to the high efficiency of DL technology in processing complex data, more and more DL technology will be used to process IoT tasks in the future.

[0004] Existing research works focus on both task offloading and transmission power control. However, the research combining power control and task offloading does not consider the impact of channel link status. Integrating wireless power transmission technology into edge servers will produce a "near-far effect". Users far away from the AP not only get less energy, but also have a longer communication distance than those who are closer to the AP. "Far" users require more time to transmit data than "near" users. In addition, research involving link status has not considered the impact of DL on task offloading. Previous studies usually have long calculation times and cannot meet the user's demand for instant task processing. In addition, the addition of collaborative devices to the system as auxiliary user devices will directly affect the rate of task offloading and the requirements of data integrity. In summary, optimizing the latency performance of WPMEC plays an important role. Summary of the invention

[0005] The purpose of the present invention is to provide a wireless powered mobile edge computing offloading method and system based on deep learning to solve the problem of "near-far effect" in WPMEC and to enable all tasks to reach the AP as quickly as possible under energy-constrained conditions.

[0006] To achieve the above object, the present invention adopts the following technical solutions:

[0007] In a first aspect, the present invention provides a wireless power supply mobile edge computing unloading method based on deep learning, comprising:

[0008] Collect energy from hybrid access points (APs) and then offload tasks that need to be offloaded directly to the APs or to the APs through devices;

[0009] Constructing transmission constraints, energy constraints, link capacity constraints, flow constraints, quantity conservation constraints and offloading time constraints during the offloading process;

[0010] The objective function is constructed under all constraints, and the problem is transformed into a mixed integer linear programming model MILP problem using piecewise linearization PWL. The approximate solution of the problem is obtained by solving it using the BCMD approximation algorithm.

[0011] The transmission power is predicted according to the DLOO algorithm, the problem is converted into an LP problem, a feasible solution is obtained, and the minimum time overhead required for energy collection and data offloading is calculated.

[0012] Optionally, the unidirectional transmission constraint for establishing a link specifically includes: a binary variable B ij Indicates the status of the feasible link during uplink transmission; if L i,j If the information transmission is activated in the uplink, its value is 1; otherwise, the value is 0. The link constraints are as follows:

[0013]

[0014] Among them B ij , B ji ∈{0,1}, user i and user j can only perform one-way transmission in a certain time block.

[0015] Optionally, create an energy constraint:

[0016] The total transmission energy consumption of UN i is as follows:

[0017]

[0018] Where P c Indicates the power loss caused by additional equipment during transmission, P i represents the transmission power of user i

[0019] The total energy consumption of UN i in receiving information is as follows:

[0020]

[0021] where r u,i Indicates that user node i passes through link L u,i The amount of information received, E elecIndicates the energy required to receive 1 bit of data;

[0022] To ensure that the transmission energy consumption and receiving energy consumption of user i do not exceed the energy obtained by the user, the following energy constraints should be met:

[0023] E ri +E ti ≤E i .

[0024] Where E i It represents the energy collected by user i from AP in the downlink.

[0025] Optionally, the user can use link L i,j The amount of information transmitted satisfies the link capacity constraint as follows:

[0026] C ij =Wlog2(1+SNR ij )

[0027] r ij ≤C ij t ij

[0028] in W represents bandwidth, η represents noise power; r ij Indicates the amount of information transmitted.

[0029] Optionally, after receiving the task data to be offloaded, each user must transmit all generated and forwarded data through its own output link, and the flow constraints are as follows:

[0030]

[0031] Among them I i represents the amount of data generated by UN i;

[0032] After all users have completed information transmission, the AP needs to receive all information before the end of the time block. The AP meets the following traffic conservation constraints:

[0033] ∑ i∈N I i =I A

[0034] The total data offloading time of all users must not be less than the minimum time consumed to offload the user's tasks to the AP. The constraints are as follows:

[0035]

[0036] Optionally, the objective function is:

[0037] MD:min t0+Toff

[0038]

[0039] E ri +E ti ≤E i .

[0040] r ij ≤C ij t ij

[0041]

[0042] B ij ∈{0,1},r ij ≥0,P i ≥0,t0≥0,t ij ≥0

[0043] Where t0 represents the energy harvesting time, T off represents the minimum time required to offload the task of user i to AP, B ij Indicates the state of the feasible link during uplink transmission, E ti represents the total transmission energy consumption of user i, E ri represents the total energy consumption of user i receiving information, E i represents the energy obtained by the user, C ij Indicates that the user is on link L i,j The transmission rate on ij represents the amount of information transmitted, t ij Indicates that user i’s task is on link L i,j Unloading time on I i Represents the amount of data generated by user i.

[0044] Optionally, the nonlinear function is converted into a piecewise linear function using the PWL method, and an error threshold is introduced to ensure that the approximate error between the piecewise function and the logarithmic function is not too large; C ij Transformed into the following formula:

[0045]

[0046] Then split the link L i,j The logarithmic function, S ij represents the number of segments, where the qth interval is represented by q∈S ij , The slope of the qth segment is expressed as follows:

[0047]

[0048] Based on the deep learning framework DLOO algorithm, the transmission power of each user is predicted and the problem is converted into an LP problem. First, the deep belief network DBN is trained offline, and the trained DBN is used to predict the transmission power of each user online. When the size of the data set increases to the set state, the DBN module will be updated with the new data set. As the new channel state is observed, these iterations are repeated and the DBN module is continuously updated to obtain the minimum time overhead required for energy collection and data offloading.

[0049] In a second aspect, the present invention provides a wireless power supply mobile edge computing unloading system based on deep learning, comprising:

[0050] A task offloading module is used to collect energy from a hybrid access point AP and then offload the tasks to be offloaded directly to the AP or to the AP through a device;

[0051] A constraint building module, used to build transmission constraints, energy constraints, link capacity constraints, flow constraints, quantity conservation constraints and offloading time constraints during the offloading process;

[0052] The objective function construction module is used to construct the objective function under all constraints, transform the problem into a mixed integer linear programming model MILP problem using piecewise linearization PWL, and obtain the approximate solution of the problem according to the BCMD approximation algorithm;

[0053] The calculation output module is used to predict the transmission power according to the DLOO algorithm, transform the problem into an LP problem, obtain a feasible solution, and calculate the minimum time overhead required for energy collection and data offloading.

[0054] In a third aspect, the present invention provides a computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the computer program, the processor implements the steps of a wireless power supply mobile edge computing offloading method based on deep learning.

[0055] In a fourth aspect, the present invention provides a computer-readable storage medium, which stores a computer program. When the computer program is executed by a processor, it implements the steps of a wireless power supply mobile edge computing offloading method based on deep learning.

[0056] Compared with the prior art, the present invention has the following technical effects:

[0057] The present invention aims at the problem of excessive system delay caused by the "near-far effect" in the WPMEC network and the unreasonable link selection during task unloading, and introduces a node cooperative transmission method. By forwarding data between nodes, the data unloading delay of the "far" node is improved, and the deployment cost of the relay node in the relay cooperation mode is eliminated. Under the condition of energy constraints, the problem is solved using a piecewise linear method to reduce the data unloading delay. The proposed DLOO framework is used to automatically predict the transmission power of each user to ensure the accuracy of the calculation while reducing the calculation delay. By combining cooperative devices and transmission power control to jointly optimize the communication link, an approximate algorithm based on BCMD and a power prediction algorithm based on DL are proposed to achieve the selection of links and optimization of transmission power during task unloading to achieve the goal of minimizing system delay. BRIEF DESCRIPTION OF THE DRAWINGS

[0058] Figure 1 It is a flow chart of the present invention.

[0059] Figure 2 It is the overall architecture of the WPMEC network.

[0060] Figure 3 This is the DLOO framework structure diagram.

[0061] Figure 4 This is the connection diagram between various algorithms. DETAILED DESCRIPTION

[0062] The specific implementation of the present invention will be described in more detail below in conjunction with the schematic diagram. The advantages and features of the present invention will become clearer from the following description and claims. It should be noted that the drawings are all in a very simplified form and are not in exact proportions, and are only used to facilitate and clearly assist in explaining the purpose of the embodiments of the present invention.

[0063] Example 1, please refer to Figure 1 ,

[0064] Embodiment 2, as shown in FIG. 2 , the overall architecture of the WPMEC network of this patent is composed of user equipment, equipment and hybrid access point (AP). First, the user equipment collects energy from the AP and then directly offloads the task to be offloaded to the AP or offloads it to the AP through the equipment; then, the energy constraint, flow constraint, channel capacity and other constraints are met during the transmission process. The problem is transformed into a MILP problem using the PWL method, and an approximate solution to the problem is obtained by solving it according to the BCMD approximation algorithm. Finally, the transmission power is predicted according to the DLOO algorithm, and the problem is transformed into an LP problem to obtain a feasible solution, thereby reducing the complexity of the model.

[0065] The wireless power supply mobile edge computing unloading method based on deep learning implemented by the present invention comprises the following steps:

[0066] Step 1: User i and user j can only perform one-way transmission in a certain time block. The link constraints are as follows:

[0067]

[0068] Step 2: During the task transmission process, energy consumption is generated in two main ways: transmitting and receiving information. The total transmission energy consumption of UN i is as follows:

[0069]

[0070] The receiving energy consumption of UN i is as follows:

[0071]

[0072] To ensure that the transmission energy consumption and receiving energy consumption of user i do not exceed the energy obtained by the user, the energy constraint is as follows:

[0073] E ri +E ti ≤E i .

[0074] Step 3: The user passes through link L i,j The amount of data transmitted is within the link capacity C ij Within the range of , the link capacity constraints are as follows:

[0075] C ij =Wlog2(1+SNR ij )

[0076] r ij ≤C ij t ij

[0077] Step 4: The amount of data transmitted by the user is not less than the amount of data generated and forwarded by the user. The traffic constraints are as follows:

[0078]

[0079] Step 5: After all users complete the task transmission, the AP needs to receive all the information within the time range. The flow conservation constraint is as follows:

[0080] ∑ i∈N I i =I A

[0081] Step 6: The total data offloading time of the user is not less than the minimum consumption time of offloading to the AP. The offloading time constraints are as follows:

[0082]

[0083] Step 7: Construct an optimization model for minimizing system latency in heterogeneous wireless powered mobile edge computing networks:

[0084] MD:min t0+T off

[0085]

[0086] E ri +E ti ≤E i .

[0087] r ij ≤C ij t ij

[0088]

[0089] B ij ∈{0,1},r ij ≥0,P i ≥0,t0≥0,t ij ≥0

[0090] Where: t0 represents the energy harvesting time, T off represents the minimum time required to offload the task of user i to AP, B ij Indicates the state of the feasible link during uplink transmission, E ti represents the total transmission energy consumption of user i, E ri represents the total energy consumption of user i receiving information, E i represents the energy obtained by the user, C ij Indicates that the user is on link L i,j The transmission rate on ij represents the amount of information transmitted, t ij Indicates that user i’s task is on link L i,j where I is the amount of data generated by user i.

[0091] Step 8: Use the PWL method to convert the nonlinear function into a piecewise linear function and introduce an error threshold to ensure that the approximate error between the piecewise function and the logarithmic function is not too large. ij Transformed into the following formula:

[0092]

[0093] Then split the link L i,j The logarithmic function, S ij represents the number of segments, where the qth interval is represented by q∈S ij , The slope of the qth segment is expressed as follows:

[0094]

[0095] Algorithm 1 is as follows:

[0096]

[0097] Among them, r ij Transformed into the following formula:

[0098]

[0099] Furthermore, the decision variable of the transmission power Pi is eliminated, the problem is transformed into a MILP problem, and the approximate solution of the MILP problem is obtained using the proposed BCND approximation algorithm. Algorithm 2 is shown below:

[0100]

[0101] Step 9: Design a deep learning framework (DLOO algorithm) to predict the transmission power of each user, convert the problem into an LP problem, and reduce the complexity of the model. It mainly includes three aspects. First, the deep belief network (DBN) is trained offline to ensure that the neural network performs well using previous data. Algorithm 3 is shown below:

[0102]

[0103] Then, the trained DBN is used to predict the transmission power of each user online. Finally, when the size of the dataset increases to a certain extent, the DBN module is updated with the new dataset. As new channel states are observed, these iterations are repeated and the DBN module is continuously updated. Algorithm 4 is shown below:

[0104]

[0105] In another embodiment of the present invention, a wireless power supply mobile edge computing unloading system based on deep learning is provided, which can be used to implement the above-mentioned wireless power supply mobile edge computing unloading method based on deep learning. Specifically, the system includes:

[0106] A task offloading module is used to collect energy from a hybrid access point AP and then offload the tasks to be offloaded directly to the AP or to the AP through a device;

[0107] A constraint building module, used to build transmission constraints, energy constraints, link capacity constraints, flow constraints, quantity conservation constraints and offloading time constraints during the offloading process;

[0108] The objective function construction module is used to construct the objective function under all constraints, transform the problem into a mixed integer linear programming model MILP problem using piecewise linearization PWL, and obtain the approximate solution of the problem according to the BCMD approximation algorithm;

[0109] The calculation output module is used to predict the transmission power according to the DLOO algorithm, transform the problem into an LP problem, obtain a feasible solution, and calculate the minimum time overhead required for energy collection and data offloading.

[0110] The division of modules in the embodiments of the present invention is schematic and is only a logical function division. There may be other division methods in actual implementation. In addition, each functional module in each embodiment of the present invention may be integrated into one processor, or may exist physically separately, or two or more modules may be integrated into one module. The above-mentioned integrated modules may be implemented in the form of hardware or in the form of software functional modules.

[0111] In another embodiment of the present invention, a computer device is provided, the computer device including a processor and a memory, the memory is used to store a computer program, the computer program includes program instructions, and the processor is used to execute the program instructions stored in the computer storage medium. The processor may be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gates or transistor logic devices, discrete hardware components, etc., which are the computing core and control core of the terminal, which are suitable for implementing one or more instructions, and are specifically suitable for loading and executing one or more instructions in a computer storage medium to implement the corresponding method flow or corresponding function; the processor described in the embodiment of the present invention can be used for the operation of the wireless power supply mobile edge computing unloading method based on deep learning.

[0112] In another embodiment of the present invention, the present invention also provides a storage medium, specifically a computer-readable storage medium (Memory), which is a memory device in a computer device for storing programs and data. It is understandable that the computer-readable storage medium here may include both a built-in storage medium in a computer device and an extended storage medium supported by the computer device. The computer-readable storage medium provides a storage space that stores the operating system of the terminal. In addition, one or more instructions suitable for being loaded and executed by a processor are also stored in the storage space, and these instructions may be one or more computer programs (including program codes). It should be noted that the computer-readable storage medium here may be a high-speed RAM memory or a non-volatile memory, such as at least one disk memory. The processor may load and execute one or more instructions stored in the computer-readable storage medium to implement the corresponding steps of the wireless power supply mobile edge computing unloading method based on deep learning in the above-mentioned embodiment.

[0113] Those skilled in the art will appreciate that embodiments of the present invention may be provided as methods, systems, or computer program products. Therefore, the present invention may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0114] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0115] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0116] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0117] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the above embodiments, ordinary technicians in the relevant field should understand that the specific implementation methods of the present invention can still be modified or replaced by equivalents. Any modification or equivalent replacement that does not depart from the spirit and scope of the present invention should be covered within the scope of protection of the claims of the present invention.

Claims

1. A wireless power supply mobile edge computing offloading method based on deep learning, characterized in that: include: Collect energy from hybrid access points (APs) and then offload tasks that need to be offloaded directly to the APs or to the APs through devices; Constructing transmission constraints, energy constraints, link capacity constraints, flow constraints, quantity conservation constraints and offloading time constraints during the offloading process; The objective function is constructed under all constraints, and the problem is transformed into a mixed integer linear programming model MILP problem using piecewise linearization PWL. The approximate solution of the problem is obtained by solving it using the BCMD approximation algorithm. The transmission power is predicted according to the DLOO algorithm, the problem is converted into an LP problem, a feasible solution is obtained, and the minimum time overhead required for energy collection and data offloading is calculated.

2. The wireless power supply mobile edge computing unloading method based on deep learning according to claim 1 is characterized in that: The unidirectional transmission constraints for establishing a link specifically include: binary variable B ij Indicates the status of the feasible link during uplink transmission; if L i,j To activate information transmission in the uplink, its value is 1; otherwise, the value is 0; the link constraints are as follows: Among them B ij , B ji ∈{0,1}, user i and user j can only perform one-way transmission in a certain time block.

3. The wireless power supply mobile edge computing unloading method based on deep learning according to claim 1 is characterized in that: Create energy constraints: The total transmission energy consumption of UNi is as follows: Where P c Indicates the power loss caused by additional equipment during transmission, P i represents the transmission power of user i The total energy consumption of UNi when receiving information is as follows: where r u,i Indicates that user node i passes through link L u,i The amount of information received, E elec Indicates the energy required to receive 1 bit of data; To ensure that the transmission energy consumption and receiving energy consumption of user i do not exceed the energy obtained by the user, the following energy constraints should be met: AND ri +E ti ≤E i . Where E i It represents the energy collected by user i from AP in the downlink.

4. The wireless power supply mobile edge computing offloading method based on deep learning according to claim 1 is characterized in that: The user passes through the link L i,j The amount of information transmitted satisfies the link capacity constraint as follows: C ij =Wlog2(1+SNR ij ) r ij ≤C ij t ij in W represents bandwidth, η represents noise power; r ij Indicates the amount of information transmitted.

5. The wireless power supply mobile edge computing unloading method based on deep learning according to claim 4 is characterized in that: After receiving the task data to be offloaded, each user must transmit all generated and forwarded data through its own output link, and the traffic constraints are as follows: Among them I i Indicates the amount of data generated by UNi; After all users have completed information transmission, the AP needs to receive all information before the end of the time block. The AP meets the following traffic conservation constraints: ∑ i∈N I i =I A The total data offloading time of all users must not be less than the minimum time consumed to offload the user's tasks to the AP. The constraints are as follows:

6. The wireless power supply mobile edge computing unloading method based on deep learning according to claim 5 is characterized in that: The objective function is: MD:min t0+T off st: AND ri +E ti ≤E i . r ij ≤C ij t ij B ij ∈{0,1},r ij ≥0,P i ≥0,t0≥0,t ij ≥0 Where t0 represents the energy harvesting time, T off represents the minimum time required to offload the task of user i to AP, B ij Indicates the state of the feasible link during uplink transmission, E ti represents the total transmission energy consumption of user i, E ri represents the total energy consumption of user i receiving information, E i represents the energy obtained by the user, C ij Indicates that the user is on link L i,j The transmission rate on ij represents the amount of information transmitted, t ij Indicates that user i’s task is on link L i,j Unloading time on I i Represents the amount of data generated by user i.

7. The wireless power supply mobile edge computing unloading method based on deep learning according to claim 6 is characterized in that: The PWL method is used to convert the nonlinear function into a piecewise linear function, and an error threshold is introduced to ensure that the approximate error between the piecewise function and the logarithmic function is not too large; C ij Transformed into the following formula: Then split the link L i,j The logarithmic function, S ij represents the number of segments, where the qth interval is represented by The slope of the qth segment is expressed as follows: Based on the deep learning framework DLOO algorithm, the transmission power of each user is predicted and the problem is converted into an LP problem. First, the deep belief network DBN is trained offline, and the trained DBN is used to predict the transmission power of each user online. When the size of the data set increases to the set state, the DBN module will be updated with the new data set. As the new channel state is observed, these iterations are repeated and the DBN module is continuously updated to obtain the minimum time overhead required for energy collection and data offloading.

8. A wireless power supply mobile edge computing unloading system based on deep learning, characterized in that: include: A task offloading module is used to collect energy from a hybrid access point AP and then offload the tasks to be offloaded directly to the AP or to the AP through a device; A constraint building module, used to build transmission constraints, energy constraints, link capacity constraints, flow constraints, quantity conservation constraints and offloading time constraints during the offloading process; The objective function construction module is used to construct the objective function under all constraints, transform the problem into a mixed integer linear programming model MILP problem using piecewise linearization PWL, and obtain the approximate solution of the problem according to the BCMD approximation algorithm; The calculation output module is used to predict the transmission power according to the DLOO algorithm, transform the problem into an LP problem, obtain a feasible solution, and calculate the minimum time overhead required for energy collection and data offloading.

9. A computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the computer program, the steps of the wireless power supply mobile edge computing offloading method based on deep learning are implemented as described in any one of claims 1 to 7.

10. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the steps of the deep learning-based wireless power supply mobile edge computing offloading method as described in any one of claims 1 to 7 are implemented.