Task Offloading Method for TDMA Wireless Energy-Harvesting Edge Computing Networks with Minimized Energy Consumption

Abstract

The present invention discloses a task offloading method for a TDMA wireless power supply edge computing network that minimizes energy consumption, including a hybrid access point and I wireless devices. The task offloading method for the TDMA wireless power supply edge computing network that minimizes energy consumption trains the energy consumption model by constructing an energy consumption model of the system and, through given several offloading decisions, optimizing the duration of the hybrid access point transmitting wireless energy to each wireless device and the transmission duration of all wireless devices offloading tasks to the edge server until the energy consumption model converges, obtaining the optimal offloading decisions for each wireless device, and the energy consumption corresponding to the optimal offloading decisions is the minimum energy consumption. The entire process only optimizes the duration of the hybrid access point transmitting wireless energy to each wireless device and the transmission duration of all wireless devices offloading tasks to the edge server, thereby enabling the model to converge quickly and the process of the offloading decision to be efficient and fast.

Description

Technical Field

[0001] The present invention belongs to the field of task offloading allocation, and specifically relates to a task offloading method for a TDMA wireless power supply edge computing network that minimizes energy consumption. Background Art

[0002] The Internet of Things (IoT) is very important for collecting data from items. It can be used in a wide range of applications, such as environmental monitoring, healthcare, and industrial automation systems. However, the IoT still faces various challenges. One of the main challenges is related to the limited energy availability of IoT nodes. Many IoT nodes are battery-powered, which limits their lifespan and performance. Therefore, extending the lifespan of IoT nodes is a major challenge to be addressed.

[0003] In recent years, due to the development of deep reinforcement learning (DRL), research has shown that DRL has significant advantages in large state and action spaces. It uses deep neural networks (DNNs) to learn from historical samples and generate an optimal mapping between the state and behavior spaces. Compared with traditional optimization methods, the method based on the DRL model has lower complexity and can achieve a near-optimal binary offloading solution. In addition, they are particularly suitable for large-scale network sizes and time-varying wireless channels, which helps to implement updated binary offloading strategies and provides an efficient solution to the binary offloading problem in a wireless power supply edge computing (WP-MEC) network.

[0004] However, in a WP-MEC network using binary offloading, it also involves the allocation of other resources. For example, channel gain, the distance between each wireless device and the hybrid access point, etc. will all affect energy consumption. If only the DRL model optimizes all parameters (including the binary offloading strategy), it is difficult to converge, which may make it difficult for the DRL model to learn a near-optimal solution. Summary of the Invention

[0005] The purpose of the present invention is to propose a task offloading method for a TDMA wireless power supply edge computing network that minimizes energy consumption to solve the problems proposed in the background art.

[0006] To achieve the above purpose, the technical solutions adopted by the present invention are as follows:

[0007] The task offloading method for a TDMA wireless power supply edge computing network that minimizes energy consumption proposed by the present invention is applied to a wireless power supply edge computing network system. The wireless power supply edge computing network system includes a hybrid access point and I wireless devices. The hybrid access point includes an edge server for executing the offloading tasks of each wireless device and a radio frequency energy transmitter for transmitting wireless energy to each wireless device;

[0008] Task offloading method for a TDMA wireless power - supplied edge - computing network that minimizes energy consumption, including:

[0009] Obtain the task volume information of each wireless device and the channel gain between each wireless device and the hybrid access point. Using the duration for which the hybrid access point transmits wireless energy to each wireless device and the power of wireless energy transmission, construct the energy consumption model \(E(m)\) of the wireless power - supplied edge - computing network system based on deep reinforcement learning as follows:

[0010]

[0011]

[0012]

[0013]

[0014] \(\beta\geq0\), \((1 - e)\)

[0015]

[0016] Among them, \(I=\{1,2,3,\cdots,I\}\) represents the set of each wireless device, \(i\) represents the \(i\) - th wireless device, \(t = \{t_1,t_2,t_3,\cdots,t\) I \} represents the set of transmission durations for each wireless device to offload tasks to the edge server in a TDMA manner, \(t\) i represents the transmission duration for the \(i\) - th wireless device to offload tasks to the edge server, \(\beta\) represents the duration for which the hybrid access point transmits wireless energy to each wireless device through the radio - frequency transmitter, \(m=\{m_1,m_2,m_3,\cdots,m\) I \} represents the set of offloading decisions of each wireless device, \(m\) i represents the offloading decision of the \(i\) - th wireless device, and it is a binary offloading decision. \(m\) i = 0 indicates that the wireless device chooses to calculate the task on the local server, \(m\) i = 1 indicates that the wireless device chooses to offload the task to the edge server for calculation. \(k\) represents the computational energy - efficiency coefficient, \(f\) represents the speed of the local processor of the wireless device, and the unit is cycles per second. \(\varphi\) represents the number of cycles required for the local server to calculate one - bit task data of the wireless device. \(W\) represents the transmission bandwidth, \(\mu\) represents the energy - capture efficiency of the wireless device, and \(0\lt\mu\lt1\). \(h\) i represents the channel gain between the hybrid access point and the \(i\) - th wireless device, \(P\) represents the power of transmitting wireless energy, \(\sigma\) 2 represents the background noise, \(S\) i represents the task volume information of the \(i\) - th wireless device itself, \(t\) max represents the transmission - constraint time in the task - offloading decision process, \(K = \{i:m\) i{i|1 ≤ i ≤ I} represents the index set of the task offloading of the wireless device to the edge server for computing, (1 - b) represents the constraint condition that β satisfies, and (1 - c) represents the task UL data transmission constraint of the i-th wireless device, ensuring that the wireless device with the offloading decision of m i = 1 completely offloads the task to the edge server, and (1 - d) represents that the task offloading decision of the wireless device is subject to time constraints;

[0017] By giving a number of offloading decisions, optimize the duration of the wireless energy transmission from the hybrid access point to each wireless device and the transmission duration of all wireless devices' tasks offloaded to the edge server to train the energy consumption model until the energy consumption model converges;

[0018] Input the task volume information and channel gain of each wireless device into the trained energy consumption model to obtain the optimal offloading decision for each wireless device.

[0019] Preferably, the constraint condition (1 - b) of the energy consumption model is obtained through the following steps:

[0020] During the process of the wireless device selecting to calculate the task on the local server, after the wireless device captures energy from the hybrid access point, the local server starts task calculation, and the energy E captured by the wireless device from the hybrid access point i satisfies:

[0021] E i = μPh i β

[0022] kf 3 τ i ≤ E i

[0023] where

[0024]

[0025] is obtained,

[0026] kf 2 S i φ ≤ μPh i β

[0027] Furthermore, further obtain the constraint condition (1 - b);

[0028] where τ i represents the time consumption of the local server computing task of the i-th wireless device.

[0029] Preferably, the constraint condition (1 - c) of the energy consumption model is obtained through the following steps:

[0030] During the process of a wireless device offloading task selection to an edge server, the task transmission rate r of the i-th wireless device i satisfies:

[0031]

[0032] The number of bits B transmitted by the i-th wireless device for the task within t i satisfies: i satisfies:

[0033]

[0034] The i-th wireless device needs to offload the task completely to the edge server. Therefore, the UL data constraint needs to satisfy the following constraint:

[0035]

[0036] Then for the offloading decision of the i-th wireless device m i = 1, it satisfies the following constraint Furthermore, the constraint condition (1 - c) is obtained.

[0037] Preferably, by giving several offloading decisions, the duration for the hybrid access point to transmit wireless energy to each wireless device and the optimal transmission duration for all wireless devices to offload tasks to the edge server are obtained, including:

[0038] Exhaustively generate 2 I offloading decisions to give the offloading decision m;

[0039] Without the constraint of formula (1 - d), when m i = 0, if the energy consumption is minimized, then formula (1 - b) takes the equal sign, and all wireless devices computing at the local server can complete the tasks, then the optimal duration for the hybrid access point to transmit wireless energy to each wireless device is directly deduced and i ∈ Z, where Z = {i: m i = 0, 1 ≤ i ≤ I} represents the index set of tasks computed at the local server;

[0040] When m i = 1, formula (1 - b) ensures that the wireless devices offloading tasks to the edge server complete sending all task data, and the optimal duration for the hybrid access point to transmit wireless energy to each wireless device is obtained through derivation and i ∈ K;

[0041] Therefore, without the constraint of (1 - d), the optimal duration β' for the hybrid access point to transmit wireless energy to each wireless device is β' = max(β1, β2);

[0042] Under the constraint of formula (1-d), for formula (1-d), let where T off represents the transmission constraint time of the task offloading decision-making process, and T off is a convex function of β. According to the properties of convex functions, there exists β that satisfies the solution set under the constraint of formula (1-d), and β ∈ [β low , β up , where β low represents the lower bound of the β solution set, and β up represents the upper bound of the β solution set;

[0043] Therefore, under the constraint of formula (1-d), according to the minimization of energy consumption, select β low , and obtain the optimal continuous duration β * = max(β’, β low ) for the hybrid access point to transmit wireless energy to each wireless device. By classifying and discussing the magnitudes of β’ and β low , the value of β * is obtained, and the optimal continuous duration β * for the hybrid access point to transmit wireless energy to each wireless device is obtained. After that, the fixed-point iteration method is used to find the optimal transmission duration t i * for each wireless device to offload its task to the edge server. Furthermore, the optimal transmission duration t* for all wireless devices to offload their tasks to the edge server is obtained, and the m corresponding to the minimum energy consumption E(m) is the optimal offloading decision of the system.

[0044] Preferably, the value of β low is obtained by classifying and discussing β’ and β * , including:

[0045] When β’ > β low , β’ satisfies formula (1-d), so β * = β’;

[0046] When β’ ≤ β low , the golden section method is used to search for β3 that minimizes T off . And to satisfy formulas (1-b) and (1-c), the lower bound of the search interval is set to β’, and in the interval [β’, β3], T off is a monotonically decreasing function of β. Then, the bisection method is used to find β off that satisfies the condition T max = t low in the interval [β’, β3]. Then β * = β low .

[0047] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0048] The task offloading method for the TDMA wireless power supply edge computing network with minimized energy consumption constructs an energy consumption model of the system, and by giving several offloading decisions, optimizes the duration of continuously transmitting wireless energy from the hybrid access point to each wireless device and the transmission duration of offloading the tasks of all wireless devices to the edge server, so as to train the energy consumption model until the energy consumption model converges, and inputs the task volume information and channel gain of each wireless device into the trained energy consumption model to obtain the optimal offloading decision for each wireless device, and the energy consumption corresponding to the optimal offloading decision is the minimum energy consumption. The whole process only optimizes the duration of continuously transmitting wireless energy from the hybrid access point to each wireless device and the transmission duration of offloading the tasks of all wireless devices to the edge server, thereby enabling the model to converge quickly and the offloading decision-making process to be efficient and fast. Brief Description of the Drawings

[0049] Figure 1 It is a schematic diagram of a wireless device and a hybrid access point in the task offloading method for the TDMA wireless power supply edge computing network with minimized energy consumption of the present invention. Detailed Embodiments

[0050] Next, the technical solutions in the embodiments of the present application will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present application.

[0051] It should be noted that when a component is referred to as being "connected" to another component, it can be directly connected to the other component or there may also be an intermediate component. Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by those skilled in the technical field to which this application belongs. The terms used in the specification of this application herein are only for the purpose of describing specific embodiments and are not intended to limit this application.

[0052] As Figure 1 shown, a task offloading method for a TDMA wireless power supply edge computing network with minimized energy consumption is applied to a wireless power supply edge computing network system. The wireless power supply edge computing network system includes a hybrid access point (Hybrid-Access Point, H-AP) and I wireless devices (Wireless Device, WD, D i ). The hybrid access point includes an edge server for executing the offloading tasks of each wireless device and a radio frequency energy transmitter for transmitting wireless energy to each wireless device;

[0053] Task offloading method for a TDMA wireless power supply edge computing network that minimizes energy consumption, including:

[0054] Step S1: Obtain the task volume information of each wireless device and the channel gain between each wireless device and the hybrid access point. Using the duration of continuous wireless energy transmission from the hybrid access point to each wireless device and the power of wireless energy transmission, construct an energy consumption model E(m) of the wireless power supply edge computing network system based on deep reinforcement learning as follows:

[0055]

[0056]

[0057]

[0058]

[0059] β ≥ 0, (1 - e)

[0060]

[0061] Among them, I = {1, 2, 3, …, I} represents the set of each wireless device, i represents the i-th wireless device, t = {t1, t2, t3, …, t I} represents the set of transmission durations for each wireless device to offload tasks to the edge server using TDMA (Time Division Multiple Access), t i represents the transmission duration for the i-th wireless device to offload tasks to the edge server, β represents the duration of continuous wireless energy transmission from the hybrid access point to each wireless device through the radio frequency transmitter, m = {m1, m2, m3, …, m I} represents the set of offloading decisions for each wireless device, m i represents the offloading decision of the i-th wireless device, and it is a binary offloading decision. m i = 0 means the wireless device selects to calculate the task on the local server, m i = 1 means the wireless device selects to offload the task to the edge server for calculation, k represents the computing energy efficiency coefficient, f represents the speed of the local processor of the wireless device, and the unit is cycles per second, φ represents the number of cycles required for the local server to calculate one bit of task data of the wireless device, W represents the transmission bandwidth, μ represents the energy capture efficiency of the wireless device, and 0 < μ < 1, h i represents the channel gain between the hybrid access point and the i-th wireless device, P represents the power of wireless energy transmission, σ 2 represents the background noise, S i represents the task volume information of the i-th wireless device itself, t maxDenote the transmission constraint time of the task offloading decision process. Let \(K = \{i:m i = 1, 1\leq i\leq I\}\) represent the index set of the tasks offloaded from wireless devices to the edge server for computing. \((1 - b)\) represents the constraint condition that \(\beta\) satisfies, and \((1 - c)\) represents the task UL data transmission constraint of the \(i\)-th wireless device, ensuring that the wireless device with \(m i = 1\) offloads the task completely to the edge server. \((1 - d)\) indicates that the task offloading decision of the wireless device is subject to time constraints.

[0062] It should be noted that the energy consumption of the wireless-powered edge computing network system consists of the energy consumption of edge server computing and the energy consumption of local server computing. The edge server will return the calculation results of the tasks to the wireless devices. The computing power of the edge server is much faster than that of the local server. The computing time of the edge server, and the energy consumption is determined by the computing time and computing power. Therefore, the energy consumption of the computing process of the edge server can be ignored. Thus, the energy consumption model in formula (1 - a) is determined by the energy consumption of local server computing. At the same time, the energy consumption of local server computing is related to the wireless energy transmitted from the hybrid access point to the wireless device. The wireless energy transmitted from the hybrid access point to the wireless device is the energy captured by the wireless device, and the local tasks are executed by the captured energy.

[0063] Specifically, in this embodiment, taking the value of \(I\) as 10 for illustration, then \(I=\{1,2,3,\cdots,10\}\), \(t = \{t_1,t_2,t_3,\cdots,t 10 \}\), \(m=\{m_1,m_2,m_3,\cdots,m 10 \}\), \(K = \{i:m i = 1, 1\leq i\leq 10\}\).

[0064] The constraint condition \((1 - b)\) of the energy consumption model is obtained through the following steps:

[0065] When the wireless device selects to calculate the task on the local server, after the wireless device captures energy from the hybrid access point, the local server starts the task calculation, and the energy \(E i captured by the wireless device from the hybrid access point satisfies:

[0066] E i =\mu P_h i \beta

[0067] k_f 3 \tau i \leq E i

[0068] where

[0069]

[0070] Obtained

[0071] kf 2 S i φ ≤ μPh i β

[0072] Furthermore Furthermore, the constraint condition (1 - b) is obtained;

[0073] where τ i represents the time consumption of the local server computing task of the i-th wireless device.

[0074] The constraint condition (1 - c) of the energy consumption model is obtained through the following steps:

[0075] During the process of the wireless device selecting to offload the task to the edge server for computing, the task transmission rate r of the i-th wireless device i satisfies:

[0076]

[0077] The number of bits B transmitted by the i-th wireless device for the task within t i satisfies: i satisfies:

[0078]

[0079] The i-th wireless device needs to completely offload the task to the edge server, so the UL data constraint needs to satisfy the following constraints:

[0080]

[0081] Then for the offloading decision of the i-th wireless device m i = 1, the following constraints are satisfied Furthermore, the constraint condition (1 - c) is obtained.

[0082] Step S2: By giving a number of offloading decisions, optimize the duration of the hybrid access point transmitting wireless energy to each wireless device and the transmission duration of all wireless devices offloading tasks to the edge server, to train the energy consumption model until the energy consumption model converges.

[0083] It should be noted that the energy consumption model uses a DNN model.

[0084] Specifically, exhaustively generate 2 10 offloading decisions to give offloading decisions: m = {m1, m2, m3, …, m 10}, m1 = {0,0,0,0,0,0,0,0,0,0}, m2 = {1,0,0,0,0,0,0,0,0,0}, m3{0,1,0,0,0,0,0,0,0,0}, m4 = {0,0,1,0,0,0,0,0,0,0}, m5 = {0,0,1,0,0,0,0,0,0,0}, m6 = {0,0,1,0,0,0,0,0,0,0}, ……, m1024 = {1,1,1,1,1,1,1,1,1,1};

[0085] Without the constraint of formula (1 - d), when m i = 0, if the energy consumption is minimized, then formula (1 - b) takes the equal sign, and all wireless devices calculated on the local server can complete the tasks (that is, the slowest local calculation needs to complete its own tasks), then the optimal duration for the hybrid access point to transmit wireless energy to each wireless device can be directly derived and i ∈ Z, where Z = {i: m i = 0, 1 ≤ i ≤ 10} represents the index set of tasks calculated on the local server;

[0086] When m i = 1, formula (1 - b) ensures that the wireless devices that offload tasks to the edge server complete sending all task data, and the optimal duration for the hybrid access point to transmit wireless energy to each wireless device is obtained through derivation and i ∈ K; The specific derivation process is as follows:

[0087] First, let

[0088]

[0089] Lemma: i ∈ K, B i is monotonically increasing with respect to both β and t i Proof: The first - order derivative of B

[0090] with respect to β is: i The first - order derivative of B

[0091]

[0092] The first - order derivative of B i with respect to t i is:

[0093]

[0094] According to the basic inequality, when x > 1, lnx > (x - 1) / x holds. By substituting x with we can obtain (dB i / dt i) > 0, which completes the proof.

[0095] According to Lemma 5-1, as long as t i is infinite, there will be

[0096]

[0097] According to constraint (1-c), it must be satisfied that:

[0098]

[0099] Substituting formula (3) into formula (4) gives:

[0100]

[0101] Then, to minimize the system energy consumption, formula (5) must take the equal sign, and we can directly derive the optimal duration

[0102] Therefore, without the constraint of (1-d), the optimal duration β' for the hybrid access point to transmit wireless energy to each wireless device is β' = max(β1, β2).

[0103] Under the constraint of formula (1-d), from formula (2), it can be observed that t i depends on β. For formula (1-d), let where T off represents the transmission constraint time of the task offloading decision process, and T off is a convex function of β (which has been proven in the literature. Cite the literature: Chi K, Zhu Y H, Li Y, et al. Minimization of transmission completion time in wireless powered communication networks[J]. IEEE Internet of Things Journal, 2017, 4(5): 1671-1683.). According to the properties of convex functions, there exists β that satisfies the solution set under the constraint of formula (1-d), and β ∈ [β low , β up , β low represents the lower bound of the β solution set, and β up represents the upper bound of the β solution set;

[0104] Therefore, under the constraint of formula (1-d), according to the minimization of energy consumption, by choosing β low , the optimal duration β for the hybrid access point to transmit wireless energy to each wireless device is obtained* = max(β’, β low ), and the value of β is obtained by classifying and discussing the magnitudes of β’ and β low : *

[0105] When β’ > β low , β’ satisfies formula (1 - d), so β * = β’;

[0106] When β’ ≤ β low , the golden section method is used to search for β3 that minimizes T off . And to satisfy formulas (1 - b) and (1 - c), the lower bound of the search interval is set to β’, and in the interval [β’, β3], T off is a monotonically decreasing function of β. Then the bisection method is used to find β off that satisfies the condition T max = t low in the interval [β’, β3], and then β * = β low ;

[0107] After obtaining the optimal continuous duration β * for the hybrid access point to transmit wireless energy to each wireless device, the fixed - point iteration method is used to find the optimal transmission duration t i * for each wireless device to offload its tasks to the edge server. Furthermore, the optimal transmission duration t* for all wireless devices to offload their tasks to the edge server is obtained, and the m corresponding to the minimum E(m) is the optimal offloading decision m of the wireless energy - supplied edge - computing network system.

[0108] Step S3: Input the task amount information and channel gain of each wireless device into the trained energy - consumption model to obtain the optimal offloading decision for each wireless device.

[0109] Specifically, the energy consumption E(m) corresponding to the optimal offloading decision m is the minimum energy consumption.

[0110] ​The task offloading method of the TDMA wireless energy - powered edge - computing network for minimizing energy consumption constructs an energy consumption model of the system. By giving several offloading decisions and optimizing the duration of wireless energy transmission from the hybrid access point to each wireless device and the transmission duration of tasks of all wireless devices to the edge server, the energy consumption model is trained until it converges. Then, the task volume information and channel gain of each wireless device are input into the trained energy consumption model to obtain the optimal offloading decision for each wireless device. Moreover, the energy consumption corresponding to the optimal offloading decision is the minimum energy consumption. The whole process only optimizes the duration of wireless energy transmission from the hybrid access point to each wireless device and the transmission duration of tasks of all wireless devices to the edge server, thus enabling the model to converge quickly and the offloading decision - making process to be efficient and fast.

[0111] The technical features of the above - described embodiments can be combined arbitrarily. For the sake of concise description, not all possible combinations of the technical features in the above - described embodiments are described. However, as long as there is no contradiction in the combination of these technical features, it should be considered as the scope recorded in this specification.

[0112] The above - described embodiments only represent the embodiments of the present application that are described more specifically and in detail, but should not be construed as a limitation on the scope of the patent application. It should be noted that for those of ordinary skill in the art, without departing from the concept of the present application, several modifications and improvements can still be made, and these all belong to the protection scope of the present application. Therefore, the protection scope of the patent of the present application should be subject to the appended claims.

Claims

1. A task offloading method for a TDMA wireless power supply edge computing network that minimizes energy consumption, applied to a wireless power supply edge computing network system, characterized in that: The wireless power supply edge computing network system includes a hybrid access point and I wireless devices. The hybrid access point includes an edge server for executing the offloading tasks of each wireless device and a radio frequency energy transmitter for transmitting wireless energy to each wireless device. The task offloading method for the TDMA wireless power supply edge computing network with minimized energy consumption includes: Obtain the task volume information of each wireless device and the channel gain between each wireless device and the hybrid access point. Based on the duration of continuous wireless energy transmission from the hybrid access point to each wireless device and the power of wireless energy transmission, construct the energy consumption model E(m) of the wireless power supply edge computing network system based on deep reinforcement learning as follows: β≥0， (1-e) where \(I = \{1, 2, 3, \ldots, I\}\) represents the set of each wireless device, \(i\) represents the \(i\)-th wireless device, \(t=\{t_1, t_2, t_3, \ldots, t\}\) I represents the set of transmission durations for each wireless device to offload tasks to the edge server in a TDMA manner, and \(t\) i represents the transmission duration for the \(i\)-th wireless device to offload tasks to the edge server. \(\beta\) represents the duration for the hybrid access point to transmit wireless energy to each wireless device through the radio frequency transmitter. \(m = \{m_1, m_2, m_3, \ldots, m\}\) I represents the set of offloading decisions for each wireless device, and \(m\) i represents the offloading decision of the \(i\)-th wireless device, and it is a binary offloading decision. When \(m\) i = 0, it means the wireless device selects to compute the task on the local server. When \(m\) i = 1, it means the wireless device selects to offload the task to the edge server for computing. \(k\) represents the computing energy efficiency coefficient, \(f\) represents the speed of the local processor of the wireless device, and the unit is cycles per second. \(\varphi\) represents the number of cycles required for the local server to compute one bit of task data of the wireless device. \(W\) represents the transmission bandwidth, \(\mu\) represents the energy capture efficiency of the wireless device, and \(0 < \mu < 1\). \(h\) i represents the channel gain between the hybrid access point and the \(i\)-th wireless device. \(P\) represents the power for transmitting wireless energy, and \(\sigma\) 2 represents the background noise. \(S\) i represents the task volume information of the \(i\)-th wireless device itself. \(t\) max represents the transmission constraint time for the offloading decision process. \(K=\{i:m\) i = 1, 1\leq i\leq I\} represents the index set of wireless devices that offload tasks to the edge server for computing. \((1 - b)\) represents the constraint condition that \(\beta\) satisfies, and \((1 - c)\) represents the task UL data transmission constraint of the \(i\)-th wireless device, ensuring that the wireless device with the offloading decision of \(m\) i = 1 completely offloads the task to the edge server. \((1 - d)\) represents that the offloading decision of the wireless device is subject to time constraints; By giving a number of offloading decisions, optimize the duration of continuous wireless energy transmission from the hybrid access point to each wireless device and the transmission duration of offloading the tasks of all wireless devices to the edge server to train the energy consumption model until the energy consumption model converges. Input the task volume information and channel gain of each wireless device into the trained energy consumption model to obtain the optimal offloading decision for each wireless device.

2. The task offloading method for a TDMA wireless power supply edge computing network that minimizes energy consumption according to claim 1, characterized in that: The constraint condition (1-b) of the energy consumption model is obtained through the following steps: During the task selection of the wireless device in the local server computing process, after the wireless device captures energy from the hybrid access point, the local server starts task computing, and the energy E captured by the wireless device from the hybrid access point i satisfies: E i = μPh i β kf 3 τ i ≤E i Where, Obtained, kf 2 S i φ ≤ μPh i β Furthermore, the constraint condition (1-b) is obtained; Among them, τ i represents the time consumption of the local server computing task of the i-th wireless device.

3. The task offloading method for the TDMA wireless energy - powered edge - computing network that minimizes energy consumption according to claim 1, characterized in that: The constraint condition (1-c) of the energy consumption model is obtained through the following steps: During the process of a wireless device offloading task selection to an edge server, the task transmission rate r of the i-th wireless device i satisfies: The number of bits B of the transmission of the i-th wireless device within t i for the in-task satisfies: i ​ The i-th wireless device needs to completely offload the task to the edge server. Therefore, the UL data constraint needs to satisfy the following constraint: For the \(i\) -th wireless device \(m\) i with an offloading decision of \(= 1\), the following constraints are satisfied Furthermore, the constraint condition \((1 - c)\) is obtained.

4. The task offloading method for a TDMA wireless power supply edge computing network that minimizes energy consumption according to claim 1, wherein: The step of obtaining the duration of continuous wireless energy transmission from the hybrid access point to each wireless device and the optimal transmission duration of offloading the tasks of all wireless devices to the edge server by giving a number of offloading decisions includes: Exhaustive generation 2 I unloading decisions for a given unloading decision m; Without the constraint of formula (1 - d), when m i = 0, if the energy consumption is minimized, then the equality sign of formula (1 - b) is achieved, and all wireless devices calculated on the local server can complete the tasks, then the optimal duration for the hybrid access point to transmit wireless energy to each wireless device can be directly derived and i ∈ Z, where Z = {i: m i = 0, 1 ≤ i ≤ I} represents the index set of tasks calculated on the local server; When m i = 1, formula (1 - b) ensures that the wireless devices that offload tasks to the edge server complete sending all task data, and the optimal duration for the hybrid access point to transmit wireless energy to each wireless device is obtained through derivation and i ∈ K; Therefore, without the constraint of (1-d), the optimal duration β' of continuous wireless energy transmission from the hybrid access point to each wireless device is β' = max(β1, β2); Under the constraint of formula (1 - d), for formula (1 - d), let where T off represents the transmission constraint time of the task offloading decision process, and T off is a convex function of β. According to the properties of convex functions, there exists β that satisfies the solution set under the constraint of formula (1 - d), and β ∈ [β low , β up , where β low represents the lower bound of the β solution set, and β up represents the upper bound of the β solution set; Therefore, under the constraint of formula (1-d), β is selected according to the minimization of energy consumption low , and the optimal duration β for the hybrid access point to transmit wireless energy to each wireless device is obtained * = max(β’, β low ), and β is obtained by classifying and discussing the magnitudes of β’ and β low . After obtaining the value of β * , which is the optimal duration for the hybrid access point to transmit wireless energy to each wireless device, the fixed-point iteration method is used to find the optimal transmission duration t * * for offloading the tasks of each wireless device to the edge server. Furthermore, the optimal transmission duration t* for offloading the tasks of all wireless devices to the edge server is obtained, and the m corresponding to the minimum energy consumption E(m) is the optimal offloading decision of the system i .

5. The task offloading method of the TDMA wireless energy - powered edge computing network for minimizing energy consumption according to claim 4, wherein: The β' and β are obtained by low classifying and discussing to obtain β * values, including: When β' > β low at this time, β' satisfies formula (1 - d), so β * = β'; When β' ≤ β low , the golden section method is used to search for β3 that minimizes T off . And to satisfy formulas (1 - b) and (1 - c), the lower bound of the search interval is set to β’, and in the interval [β’, β3], T off is a monotonically decreasing function of β. Then the bisection method is used to find β off that satisfies the condition T max = t low in the interval [β’, β3]. Then β * = β low .

Images