Task offloading method, apparatus and system

Abstract

Embodiments of the present application provide a task offloading method, device and system for wireless powered communication network (WPCN), by inputting the environment parameters of each WD at the beginning of the current time frame into a deterministic policy deep gradient (DDPG) model, the environment parameters including the remaining offloading data amount L and the remaining energy E of each WD at the beginning of the current time frame, to obtain the offloading strategy of each WD in the current time frame output by the DDPG model, wherein the offloading strategy includes backscattering time ta, active transmission time tb and active transmission power Pa, and each WD transmits the offloading data amount according to the determined offloading strategy in each time frame until all the offloading data amount is transmitted, so that each WD in the WPCN can fully utilize the energy and almost simultaneously complete the task offloading regardless of the size of the task amount to be offloaded by each WD, thereby reducing the system task offloading delay and improving the system energy utilization.

Description

Technical Field

[0001] This application relates to the field of Internet of Things (IoT) communication technology, and in particular to a task offloading method, apparatus and system in a Wireless Powered Communication Network (WPCN). Background Technology

[0002] With the rollout of 5G technology, wireless devices (WDs) in the Internet of Things (IoT) will experience explosive growth. In IoT communication, since WDs typically face resource and energy constraints, latency and energy consumption are two crucial factors that need to be considered.

[0003] Mobile Edge Computing (MEC) offloads compute-intensive tasks to nearby MEC servers via Device Drivers (WDs) to execute the tasks on the MEC servers and receive the task processing results returned by the MEC servers, thereby reducing the task processing latency of WDs.

[0004] Wireless Powered Communication Networks (WPCNs) are wireless networks that combine two different wireless communication technologies: backscatter communication and cognitive radio. A typical WPCN consists of one Hybrid Access Point (HAP) and multiple Wildcard Devices (WDs). The HAP broadcasts a primary carrier signal to provide an ambient carrier for energy harvesting and backscatter communication to the WDs in the network, mitigating the energy constraints of the WDs. It also provides downlink communication support to the WDs (such as transmitting the processing results of offload tasks). WDs can transmit their offloaded task data to the HAP using a backscatter transmission mode while the HAP broadcasts the primary carrier signal, or they can harvest energy first while the HAP broadcasts the primary carrier signal, and then transmit the offloaded task data to the HAP using an active transmission mode through spectrum sensing when the HAP is idle.

[0005] Due to resource and energy constraints on WDs (Drivers for Loading and Distributing), only one of backscatter or active transmission can typically be used for transmission at a time. Furthermore, in a WPCN (Work-Wide Network for Loading and Distributing) network, when multiple WDs need to offload tasks simultaneously, only one WD can be offloaded at a time. Related technologies typically employ an equal-sharing strategy to allocate the active and backscatter transmission modes and corresponding working times among multiple WDs in the WPCN. However, even after a WD completes its offloading, it still receives transmission time, wasting its energy resources and impacting the latency of other WDs' offloading, easily leading to low system energy utilization and long latency. Summary of the Invention

[0006] This application provides a task offloading method, apparatus, and system to solve the problems of low system energy utilization and long latency when multiple WDs in a WPCN perform task offloading in the prior art.

[0007] On one hand, embodiments of this application provide a task offloading method for a Power over Wireless Communication Network (WPCN), wherein the WPCN includes N wireless devices (WD), where N is an integer not less than 2, and the method includes:

[0008] Step S101: Obtain each of the N WDs j The environmental parameters at the start of the current time frame, where j = 1, 2, ..., N, the environmental parameters include each WD j The first amount of data to be unloaded, L j and each of the aforementioned WD j Remaining energy E j The first amount of data to be unloaded, L j For each of the WD j The remaining amount of data to be uninstalled;

[0009] Step S102: Place each WD j The environmental parameters at the start of the current time frame are input to the deterministic policy deep gradient DDPG model.

[0010] Step S103: Obtain each WD output by the DDPG model. j The offloading strategy in the current time frame, wherein the offloading strategy includes the backscattering time t b,j Active transmission time t a,j and active transmission power P a,j ;as well as,

[0011] Step S104: Control each WD j According to the backscattering time t in the current time frame b,jPerform reverse transmission and according to the active transmission time t a,j and the active transmission power P a,j Active transmission is performed to transfer the amount of data L to be unloaded. j Transmitted to network-side nodes.

[0012] Secondly, embodiments of this application provide a task unloading device, comprising:

[0013] The memory stores program code, environmental parameters of the wireless device (WD), and offloading strategies. The environmental parameters include the amount of data to be offloaded and the remaining energy of the WD at the start of a Power over Wireless Communication Network (WPCN) time frame. The offloading strategy includes the backscattering time t transmitted by the WD during that WPCN time frame. b,j Active transmission time t a,j and active transmission power P a,j ;

[0014] A deterministic policy deep gradient DDPG model is used to determine the unloading policy based on the environmental parameters; and

[0015] A processor is used to invoke the program code and the DDPG model to execute the task unloading method as described in the first aspect.

[0016] Thirdly, embodiments of this application provide a task offloading system, including a hybrid access node (HAP) and multiple wireless devices (WDs), wherein the HAP includes the task offloading device as described in the second aspect.

[0017] The task offloading method, apparatus, and system provided in this application embodiment input the environmental parameters of each WD at the beginning of the current time frame into a deterministic policy deep gradient DDPG model. The environmental parameters include the remaining amount of data to be offloaded L and the remaining energy E of each WD at the beginning of the current time frame. The offloading strategy of each WD in the current time frame is obtained by the DDPG model output. The offloading strategy includes backscatter time ta, active transmission time tb, and active transmission power Pa. Each WD transmits the amount of data to be offloaded in each time frame according to the determined offloading strategy until all the data to be offloaded is transmitted. This allows each WD in the WPCN, regardless of the size of the task to be offloaded by each WD, to make full use of energy and complete the task offloading almost simultaneously when the system latency is minimized. This reduces the system task offloading latency and improves the system energy utilization rate. Attached Figure Description

[0018] To more clearly illustrate the technical solutions in this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0019] Figure 1 This is a schematic diagram of the system architecture of the WPCN network according to an embodiment of this application;

[0020] Figure 2 This is a schematic diagram of time frames in a WPCN network according to an embodiment of this application;

[0021] Figure 3 This is a flowchart illustrating the task unloading method according to an embodiment of this application;

[0022] Figure 4 This is a flowchart illustrating a task unloading method according to another embodiment of this application;

[0023] Figure 5 This is a flowchart illustrating a task unloading method according to another embodiment of this application;

[0024] Figure 6 A schematic diagram of the structure of a task unloading device according to an embodiment of this application; and

[0025] Figure 7 This is a schematic diagram of the structure of a task unloading device according to another embodiment of this application. Detailed Implementation

[0026] To make the objectives, technical solutions, and advantages of this application clearer, the technical solutions in the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this application. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0027] In this embodiment of the application, "and / or" describes the relationship between associated objects, indicating that there can be three relationships. For example, A and / or B can represent: A existing alone, A and B existing simultaneously, or B existing alone, where A and B can be singular or plural.

[0028] Figure 1 This is a schematic diagram of the system architecture of a WPCN network according to an embodiment of this application. Figure 1As shown, the WPCN network system includes a Hybrid Access Point (HAP) and multiple wireless devices WD1, WD2, ... and WD N Where N is an integer not less than 2. On the downlink, the HAP broadcasts the primary carrier signal, providing an ambient carrier for energy harvesting and backscatter communication for wireless devices in the network, and also providing downlink communication support for wireless devices in the network, such as transmitting the processing results of offloading tasks. On the uplink, as the access point for task offloading, it receives task data offloaded via active transmission or backscatter transmission. In some embodiments, the WD can offload task data to the HAP or other relay devices, or it can offload task data to the MEC server via the HAP or other relay devices. For simplicity, this application uses the example of the WD offloading task data to the MEC server via the HAP, but this method is also applicable to other situations, which will not be described in detail here.

[0029] Since backscattering reflects the primary carrier by adjusting the antenna impedance and cannot adjust the carrier frequency, all backscattering devices in the network are equipped with only one antenna and operate on the same frequency band using time-division multiple access (TDMA). HAPs operate in a duty-cycling manner, such as... Figure 2 As shown, each time frame consists of two stages:

[0030] 1) Wireless Energy Transfer Time (WET), also known as the energy harvesting or backscattering phase: When the HAP occupies the channel for downlink transmission, the WD can only choose to harvest energy or offload the task using backscattering. Only one WD can offload data using backscattering at a time, while other WDs can still harvest energy; and

[0031] 2) Wireless Idle Time (WIT), also known as the active transmission phase: When the HAP is idle, the WD can occupy the channel and use the active transmission mode to offload tasks. Only one WD can perform active transmission at the same time.

[0032] A time frame can be normalized to a unit length "1". Let w represent the WET phase duration, then the WIT phase duration is 1-w. Assume there are N = {1, 2, ..., N} WDs in the system, and L = {L... 1, L2…,L N} is the set of data sizes to be unloaded for each WD. All WDs in the network share WET and WIT times. If the j-th WD, i.e., WD jThe time lengths allocated to WET and WIT are respectively represented by t. b,j and t a,j If expressed, then there are constraints:

[0033]

[0034] In related technologies, when multiple WDs are offloading tasks, a strategy of equal distribution or a greedy algorithm is usually used to allocate the backscatter time or active transmission time of multiple WDs. However, the amount of tasks to be offloaded by each WD is not taken into account, which can easily lead to system delays and low energy utilization.

[0035] This application proposes a task offloading method, apparatus, and system for task offloading among multiple WDs in a WPCN network. The optimal offloading strategy for each WD is determined through the DDPG model, which reduces system offloading latency and improves system energy utilization. The specific implementation process is described below.

[0036] In the scenario where multiple WDs perform task offloading in this application, the following explanation is provided first: Compared with the uplink transmission time of the offloading task, the backhaul delay from the MEC server returning the processing result via HAP to the WD is very small and can be ignored;

[0037] The backscatter receiver, or HAP, is typically a "cooperative receiver" consisting of an Ambient Backscatter Communication (AmBC) receiver and a conventional receiver. Through continuous interference cancellation, it can recover the backscatter signal from the mixed signal (i.e., backscatter + downlink transmission signal). Therefore, the interference of the HAP's main signal on the backscatter signal can be ignored. However, the scattered signals from different WDs are still inseparable; therefore, only one WD can perform backscatter communication in a WPCNs system at a time.

[0038] In backscatter mode, the energy collected by WD's transducers can be directly used to drive the backscatter and codec hardware units without consuming additional energy.

[0039] Figure 3 This is a flowchart illustrating a task offloading method according to an embodiment of this application. This method can be executed by a corresponding task offloading device, which can be a separate network element in the WPCN network or integrated into the HAP; this application does not limit its scope. Figure 3 As shown, the task unloading method may include the following steps S101 to S104.

[0040] Step S101: Obtain each of the N WDs jThe environmental parameters at the start of the current time frame, where j = 1, 2, ..., N, the environmental parameters include each WD j The first amount of data to be unloaded, L j and each of the aforementioned WD j Remaining energy E j The first amount of data to be unloaded, L j For each of the WD j The remaining amount of data to be uninstalled.

[0041] Assuming the current time frame is the k-th time frame, k = 1, 2, ..., each WD j The environmental parameters at the start of the k-th time frame may include the first amount of data to be unloaded, L. j (k), and the remaining energy E j (k). When k=1, the first amount of data to be unloaded, L j (1) WD j The initial amount of data to be unloaded, i.e., WD j Before the task unloading and transfer have been completed, the remaining energy E j (1) Also WD j Initial energy, per WD in one embodiment j The initial energy is 0. The initial amount of data to be unloaded and the initial energy can be obtained directly from the WD or reported actively by the WD. This application does not make specific restrictions on this.

[0042] Step S102: Place each WD j The environmental parameters at the start of the current time frame are input to the deterministic policy deep gradient DDPG model.

[0043] In some embodiments, the method further includes training the DDPG model with a tuple (S, A, R) consisting of states, actions, and rewards, wherein the state S includes each of the WDs. j In the environmental parameters of the current time frame, action A includes each WD. j In the unloading policy of the current time frame, the reward R is based on each WD. j The first amount of data to be unloaded L j and the second amount of data to be unloaded l j Confirmed. The second volume of data to be unloaded is l. j For the WD j The amount of data to be unloaded transmitted in the current time frame.

[0044] In some embodiments, the first amount of data to be unloaded L jOnce the entire data transmission is completed in the current time frame and the task unloading is finished, the second data volume to be unloaded, l, will begin. j The amount of data to be uninstalled is the same as the first batch.

[0045] In some embodiments, it may be based on each WD j The first amount of data to be unloaded L j The second amount of data to be unloaded j The reward R is determined by the Jain Fairness Index, which may be determined based on the cumulative amount of unloaded data transmitted by the N WDs.

[0046] Step S103: Obtain each WD output by the DDPG model. j The offloading strategy for the current time frame, wherein the offloading strategy includes the backscattering time t b,j Active transmission time t a,j and active transmission power P a,j .

[0047] Step S104: Control each WD j According to the backscattering time t in the current time frame b,j Perform reverse transmission and according to the active transmission time t a,j and the active transmission power P a,j Active transmission is performed to transfer the first amount of data to be unloaded, L. j Transmitted to network-side nodes.

[0048] Specifically, the task offloading device can directly send the task offloading policy to each WD. j Alternatively, the task offloading policy can be sent to each WD via HAP. j To make WD j The task is unloaded and transferred according to the corresponding unloading strategy, which is not limited in this application.

[0049] In some embodiments, the first remaining unloaded quantity L j (k) If the transmission of the kth frame is not completed, the method embodiment of this application further includes step S105, referring to Figure 4 Describe it.

[0050] Figure 4 This is a flowchart illustrating a task unloading method according to another embodiment of this application, as shown below. Figure 4 As shown, this embodiment is in Figure 3 In addition to the above, it also includes step S105, based on each WD j The offloading strategy in the current time frame determines each WD. jThe environmental parameters at the start of the next time frame, and according to each WD j The environmental parameters at the start of the next time frame are determined, and steps S101 to S104 are repeated until the N WDs are reached. j All tasks have been completed, including unloading and transferring.

[0051] In some embodiments, the tuple (S, A, R) may also include each of the WDs. j The environmental parameter S' in the next time frame.

[0052] In one embodiment, the specific implementation of step S105 can be found in [reference needed]. Figure 5 See below for details.

[0053] Figure 5 This is a flowchart illustrating a task unloading method according to another embodiment of this application. Figure 4 Further explanation of step S105 in the illustrated embodiment is as follows: Figure 5 As shown, step S105 further includes steps S1051 to S1055.

[0054] Step S1051: According to each WD j active transmission rate r a,j Backscattering rate r b,j The backscattering time t b,j and the active transmission time t a,j Determine each WD j The second amount of data to be unloaded j The second amount of data to be unloaded l j For the WD j The amount of data to be unloaded transmitted in the current time frame.

[0055] Step S1052: Based on the first amount of data to be unloaded L j and the second amount of data to be unloaded l j Determine each WD j The first amount of data to be unloaded at the start of the next time frame.

[0056] Step S1053: According to each WD j The backscattering time t b,j Determine each WD j During the energy harvesting time of the current time frame, and according to each WD j The active transmit power, downlink channel gain, energy conversion efficiency, and energy harvesting time determine each WD j Energy collected in the current time frame.

[0057] Step S1054: According to each WD j The active transmission time t a,j and the active transmission power p a,j Determine each WD j Energy consumed in the current time frame.

[0058] Step S1055: According to each WD j The energy collected in the current time frame and each WD j The energy consumed in the current time frame determines each WD. j The remaining energy at the start of the next time frame.

[0059] In step S1055, the remaining energy at the start of the next time frame does not exceed the value of each WD. j The maximum capacity of the battery. Specifically, the remaining energy at the start of the next frame is taken as the battery's maximum capacity and according to each WD. j The energy collected in the current time frame and each WD j The energy consumed in the current time frame determines each WD. j The smaller of the two remaining energies at the start of the next time frame.

[0060] The specific implementation process of the above method embodiment will be described below. Taking the current time frame as the k-th frame as an example, and taking the case where the remaining unloaded amount has not been transmitted in the k-th frame (that is, k is not less than 2, and the remaining unloaded amount needs to be transmitted in multiple time frames), the implementation process is similar for other cases, and will not be described again.

[0061] First determine WD j At the end of the k-th time frame, the first amount of data to be unloaded, L j (k)=L j (k-1)-l j (k) For a WD, the data offloaded within a WPCN time frame consists of two parts: the data offloaded by backscattering during the WET phase and the data actively transmitted during the WIT phase. Therefore, for a WD j In other words, the total amount of data actually unloaded within a time frame can be expressed as:

[0062] l j =r b,t t b,t +r a,j t a,j (1).

[0063] In some embodiments, it may be based on WD jThe intrinsic backscattering rate R b The capacity C(t) of the backscattered binary symmetric channel is determined by WD. j The actual backscattering rate r b,j .

[0064] WD j The intrinsic backscattering rate R b It depends on WD j The actual decoding circuit. However, decoder noise and uplink channel noise affect the bit error rate at the receiver, i.e.:

[0065]

[0066] in It is the noise power of the additive white Gaussian noise (AWGAN) in the backscatter channel. This represents the noise power of the decoding circuit, ρ∈(0,1) represents the reflection coefficient, and P H g represents the active transmit power of the HAP. j and h j They represent WD j downlink channel gain and uplink channel gain, N s Let erfc(·) represent the sampling rate in the backscattering circuit, and erfc(·) be the complementary error function, i.e.:

[0067]

[0068] Therefore, the capacity of the backscattering binary symmetric channel is:

[0069] C(t)=1+εlogε+(1-ε)log(1-ε) (4).

[0070] Assuming the backscattering rate is fixed at R b If it remains unchanged, then WD j The actual backscattering rate can be expressed as:

[0071] r b，j =C(t)·R b (5).

[0072] In some embodiments, the active transmission rate r can be determined by the active transmission bandwidth and the active transmission power. a，j The active transmission rate can be expressed by the following formula:

[0073]

[0074] Where B is the active transmission bandwidth, p a，jThis indicates the active transmission power. The received signal of the HAP is... Where x j Indicates unit power. Add Gaussian noise.

[0075] In some embodiments, WD j Remaining energy E j It can be done through WD j Energy e collected during the WET phase a，j And the energy loss caused by active transmission during the WIT phase. a，j To determine, the energy loss e caused by active transmission. h，j It can be expressed by the following formula:

[0076] e a，j =p a，j t a，j (7),

[0077] The energy collected during the WET phase can be expressed by the following formula:

[0078] e h,j =ηg j P H (wt b,j (8),

[0079] Where η∈(0,1) represents the energy conversion efficiency, P H g represents the active transmit power of the HAP. j Indicates WD j Downlink channel gain.

[0080] E j (k) represents the energy stored in the battery at the end of the (k-1)th time frame, that is, at the beginning of the kth time frame:

[0081] E j (k)=min{E max E j (k-1)+e h，j (k)-e a,j (k)} (9)

[0082] Where E max This indicates the battery's maximum capacity. It should be noted that WD... j During the WIT phase, the energy consumption for active transmission should be limited by battery energy, namely:

[0083] p a，j (k)·t a，j (k)=e a，j (k)≤E j (k-1)+eh，j (k).

[0084] In some embodiments, for the DDPG model, tuple pairs consisting of state, action, and reward can be used (S k A k R k The neural network is trained. Among them,

[0085] Status: Represents the environmental parameters of all WDs at the start of the k-th time frame.

[0086] S k ={E1(k), ...E N (k), L1(k), ..., L N (k)};

[0087] Action: Represents the offloading strategy adopted by all WDs in the k-th time frame:

[0088] in Indicates WD j The unloading strategy in the k-th time frame.

[0089] Bonus: Can be obtained through each WD j The first amount of data to be unloaded L j The second amount of data to be unloaded j And the Jain Fairness Index is used to determine this, which can be expressed by the following formula (10):

[0090]

[0091] in, Let Jain's fairness index be the value at time frame k. Furthermore, the larger the value, the better the fairness. Indicates WD j The cumulative amount of unloaded data from the start to the current time frame.

[0092] Because the DDPG algorithm incorporates the Jain fairness index, it ensures that all WD unloading tasks are completed simultaneously.

[0093] Using the above triplet Indicates WD j To determine the optimal task unloading strategy Tj in the k-th time frame, the following problem P1 can be solved using the DDPG model, where P1 can be expressed as:

[0094] It is subject to the following constraints:

[0095]

[0096] e a，j (k)≤E j (k-1)+e h，j (k) (C5)

[0097] Where γ∈(0,1) represents the attenuation coefficient. Constraint (C3) states that the sum of the backscatter transmission times of all WDs cannot exceed the WET time; constraint (C4) states that the sum of the active transmission times of all WDs cannot exceed the WIT time; constraint (C5) states that the energy used for active transmission cannot exceed the energy stored in the battery.

[0098] In some embodiments, the optimal unloading strategy can be determined by inputting the above-mentioned state, action, and reward triplet into DDPG to solve the P1 problem.

[0099] There are four networks in DDPG:

[0100] • Actor (current network): Responsible for iteratively updating the policy network parameters μ, selecting the current action A based on the current state S, and interacting with the environment to generate the next state S′ and reward R;

[0101] • Actor target network: Responsible for selecting the next action A′ from the next state S′ sampled from the experience replay buffer. The network parameters μ′ of this network are periodically updated by copying μ from the current actor network;

[0102] • The current network critic is responsible for iteratively updating the value network parameters θ and calculating the current Q-value Q(S, A|θ), i.e.: y i =R+γQ'(S', A', θ');

[0103] • The critic target network: responsible for calculating the Q′(S′, A′, θ′) part of the target Q-value. The network parameters θ′ of this network are periodically updated by copying θ from the current critic network.

[0104] DDPG adopts a "soft" update approach, updating only small amounts at a time.

[0105] μ′ k+1 =τμ′ k +(1-τ)μ′ k (12)

[0106] θ′ k+1 =τθ′ k +(1-τ)θ′ k (13)

[0107] Where τ is the update coefficient, this update method can greatly improve the stability of learning.

[0108] The current actor network uses a deterministic strategy to generate deterministic actions, and the loss gradient is:

[0109]

[0110] c riti c The current network's loss function uses mean squared error:

[0111] J(θ)=E[(y k -Q(S,A|θ) 2 (15).

[0112] Specifically, the following DDPG-based optimization algorithm can be used to determine the optimal unloading strategy.

[0113]

[0114] The DDPG-based optimization algorithm described above can search for the optimal action (i.e., the unloading strategy for each WD) in a continuous space, while taking into account the fairness among various mobile devices and the differences in the size of their unloading tasks.

[0115] The meanings of the parameters involved in the above task unloading method can be referenced as follows.

[0116]

[0117] The task offloading method provided in this application, while considering the remaining offloading data size of each WD and the remaining battery energy, determines the optimal offloading strategy in the time frames including backscatter and active transmission by using the DDPG model. It optimizes the time allocation for each WD in the backscatter and active transmission stages, maximizes the immediate reward and future reward in each time frame, ensures that all WDs can complete the offloading task at the same time, reduces the overall task offloading latency of the system, and improves the system energy efficiency to avoid accumulating too much redundant energy in the battery.

[0118] Figure 6 This is a schematic diagram of the structure of a task unloading device according to an embodiment of this application, as shown below. Figure 6 As shown, the task unloading device 200 may include: an acquisition module 201, a DDPG module 202, and a control module 203, wherein:

[0119] The acquisition module 201 can be used to acquire each of the N WDs. j The environmental parameters at the start of the current time frame, where j = 1, 2, ..., N, the environmental parameters include each WD j The first amount of data to be unloaded, L jand each of the aforementioned WD j Remaining energy E j The first amount of data to be unloaded, L j For each of the WD j The remaining amount of data to be uninstalled;

[0120] DDPG module 202 can be used to transfer each of the WD j The environment parameters at the start of the current time frame are input to a deterministic policy deep gradient DDPG model, and the output of the DDPG model is obtained for each WD. j The offloading strategy for the current time frame, wherein the offloading strategy includes the backscattering time t b,j Active transmission time t a,j and active transmission power P a,j ;as well as

[0121] Control module 203 can be used to control each of the WDs. j According to the backscattering time t in the current time frame b,j Perform reverse transmission and according to the active transmission time t a,j and the active transmission power P a,j Active transmission is performed to transfer the first amount of data to be unloaded, L. j Transmitted to network-side nodes.

[0122] The task unloading device is used to implement the aforementioned method embodiments, and its implementation principle and technical effect are similar, so it will not be described again here.

[0123] Figure 7 This is a schematic diagram of the structure of a task unloading device according to another embodiment of this application, as shown below. Figure 7 As shown, the task unloading device 300 may include: a memory 301, a DDPG model 302, and a processor 303, which can be connected via a bus (such as... Figure 7 Interconnection (shown by thick solid lines).

[0124] The memory 301 can be used to store program instructions, WD environmental parameters, and unloading strategies. The environmental parameters include the amount of remaining unloaded data and remaining energy of the WD at the beginning of a WPCN time frame. The unloading strategy includes the backscattering time t transmitted by the WD in the WPCN time frame. b,j Active transmission time t a,j and active transmission power P a,j It should be noted that the memory 301 may also store other data or information used for executing the task unloading method, and this application does not impose any restrictions on this.

[0125] The DDPG model 302 can be used to determine the unloading strategy based on the environmental parameters. The implementation of this DDPG model can be found in the preceding descriptions and will not be repeated here.

[0126] The processor 303 can be used to call the program code and the DDPG model to execute the task unloading method as described in the embodiments of this application.

[0127] The task unloading device in this application embodiment is used to implement the aforementioned method embodiment, and its implementation principle and technical effect are similar, so it will not be described again here.

[0128] Figure 7 Only a schematic diagram of the structure that can be used to implement the task offloading device 300 disclosed herein is shown. The task offloading device 300 may utilize all or only a subset of the components shown, and the structure may vary depending on the specific circumstances. Furthermore, the task offloading device may include multiple instances of components, such as multiple processors, memory, transmitters, receivers, network interfaces, etc. The task offloading device may also include a processing unit equipped with one or more input / output devices such as a mouse, touchscreen, keypad, keyboard, printer, and monitor. The processing unit may include a central processing unit (CPU), graphics processing unit (GPU), memory, mass storage device, video adapter, I / O interface, and network interface connected to a bus.

[0129] The bus can be one or more of any type of bus architecture, including memory buses or memory controllers, peripheral buses, video buses, etc. The CPU can include any type of electronic data processor. The memory can include any type of system memory, such as static random access memory (SRAM), dynamic random access memory (DRAM), synchronous DRAM (SDRAM), read-only memory (ROM), and combinations thereof. In some embodiments, the memory may include ROM for use at startup, and DRAM for program and data storage during program execution.

[0130] Mass storage devices can include any type of storage device used to store data, programs, and other information, and to make that data, programs, and other information accessible via a bus. Mass storage devices can include, for example, one or more of solid-state drives, hard disk drives, disk drives, optical disk drives, or the like.

[0131] Video adapters and I / O interfaces provide interfaces for coupling external input and output devices to the processing unit. As shown in the figure, examples of input and output devices include a monitor coupled to the video adapter and a mouse / keyboard / printer coupled to the I / O interface. Other devices can be coupled to the processing unit, and additional or fewer interface cards may be used. For example, a serial interface such as Universal Serial Bus (USB) (not shown) can be used to interface to a printer.

[0132] The processing unit also includes one or more network interfaces, which may include wired links, such as Ethernet cables, and / or wireless links to access nodes or different networks. The network interface allows the processing unit to communicate with remote units over a network. For example, the network interface may provide wireless communication via one or more transmitters / transmit antennas and one or more receivers / receive antennas, wherein the transmitters and receivers may be integrated together. In one embodiment, the processing unit is coupled to a local area network (LAN) or a wide area network (WAN) for data processing and communication with remote devices, such as other processing units, the Internet, remote storage devices, etc.

[0133] Those skilled in the art will understand that all or part of the steps of the above-described method embodiments can be implemented by hardware related to program instructions. The aforementioned program can be stored in a computer-readable storage medium. When executed, the program performs the steps of the above-described method embodiments; and the aforementioned storage medium includes various media capable of storing program code, such as ROM, RAM, magnetic disks, or optical disks.

[0134] This application also supports computer program products comprising computer-executable code or computer-executable instructions, which, when executed, cause at least one computer to perform the operational and computational steps described herein, particularly the steps of the methods described above. Such computer program products may include a readable, non-transitory storage medium on which program code is stored for computer use. The program code can perform the processing and computational steps described herein, particularly the methods described above.

[0135] While a particular feature or aspect of this disclosure may be disclosed with reference to only one of several embodiments, such features or aspects may be combined with one or more other features or aspects of other embodiments as needed or advantageous for any given application or particular application. Furthermore, the terms “comprising,” “possessing,” “having,” or other variations are used in the specific embodiments or claims, and these terms are intended to be inclusive, similar to “including.” Additionally, the terms “exemplary,” “for example,” and “such” merely mean examples and not best or optimal. The terms “coupled” and “connected,” as well as derivatives, may be used. It should be understood that these terms can be used to indicate that two elements cooperate or interact with each other, whether they are in direct physical or electrical contact or not in direct contact with each other.

[0136] While specific aspects have been illustrated and described herein, those skilled in the art will understand that various alternative and / or equivalent implementations may be used in place of the specific aspects shown and described without departing from the scope of this disclosure. This application is intended to cover any modifications or variations to the specific aspects discussed herein.

[0137] Although the elements in the following claims are listed in a specific order of their respective designations, these elements are not necessarily intended to be limited to implementation in that specific order unless the claims indicate a specific sequence of some or all of the elements for implementing these elements.

[0138] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A task offloading method for a Power over Wireless Communication Network (WPCN), wherein the WPCN comprises N wireless devices (WD), where N is an integer not less than 2, characterized in that, The method includes: Step S101: Obtain each of the N WDs j The environmental parameters at the start of the current time frame, where j = 1, 2, ..., N, the environmental parameters include each WD j The first amount of data to be unloaded, L j and each of the aforementioned WD j Remaining energy E j The first amount of data to be unloaded, L j For each of the WD j The remaining amount of data to be uninstalled; Step S102: Place each WD j The environment parameters at the start of the current time frame are input to a deterministic policy deep gradient DDPG model, wherein the DDPG model introduces the Jain fairness index, and the Jain fairness index is determined based on the cumulative amount of unloaded data transmitted by the N WDs; Step S103: Obtain each WD output by the DDPG model. j The offloading strategy for the current time frame, wherein the offloading strategy includes the backscattering time t b,j Active transmission time t a,j and active transmission power P a,j ; Step S104: Control each WD j According to the backscattering time t in the current time frame b,j Perform reverse transmission and according to the active transmission time t a,j and the active transmission power P a,j Active transmission is performed to transfer the first amount of data to be unloaded, L. j Transmitted to network-side nodes; and Step S105: According to each WD j The offloading strategy in the current time frame determines each WD. j Environmental parameters at the start of the next time frame; and According to each WD j Based on the environmental parameters at the start of the next time frame, steps S101 to S104 are repeated until the N WDs are reached. j Both tasks were successfully unloaded and transferred simultaneously.

2. The method according to claim 1, characterized in that, Step S105 includes: Step S1051: According to each WD j active transmission rate r a,j Backscattering rate r b,j The backscattering time t b,j and the active transmission time t a,j Determine each WD j The second amount of data to be uninstalled The second amount of data to be unloaded For the WD j The amount of data to be unloaded transmitted in the current time frame; Step S1052: Based on the first amount of data to be unloaded L j and the second amount of data to be uninstalled Determine each WD j The first amount of data to be unloaded at the start of the next time frame; Step S1053: According to each WD j The backscattering time t b,j Determine each WD j During the energy harvesting time of the current time frame, and according to each WD j The active transmit power, downlink channel gain, energy conversion efficiency, and energy harvesting time determine each WD j Energy collected in the current time frame; Step S1054: According to each WD j The active transmission time t a,j and the active transmission power p a,j Determine each WD j Energy consumed in the current time frame; and Step S1055: According to each WD j The energy collected in the current time frame and each WD j The energy consumed in the current time frame determines each WD. j The remaining energy at the start of the next time frame.

3. The method according to claim 2, characterized in that, The remaining energy at the start of the next time frame shall not exceed each WD. j The maximum capacity of the battery.

4. The method according to claim 2, characterized in that, Also includes: The DDPG model is trained using tuples (S, A, R) consisting of states, actions, and rewards, where each WD is a tuple of states, actions, and rewards. j In the environmental parameters of the current time frame, action A includes each WD. j In the unloading policy of the current time frame, the reward R is based on each WD. j The first amount of data to be unloaded L j and the second amount of data to be uninstalled It's confirmed.

5. The method according to claim 4, characterized in that, Also includes: According to each WD j The first amount of data to be unloaded L j The second amount of data to be unloaded The reward R is determined by the Jain Fairness Index.

6. The method according to claim 4, characterized in that, The tuple (S, A, R) also includes each of the WDs. j The environmental parameter S' in the next time frame.

7. A task unloading device, characterized in that, include: The memory stores program code, environmental parameters of the wireless device (WD), and offloading strategies. The environmental parameters include the amount of data to be offloaded and the remaining energy of the WD at the start of a Power over Wireless Communication Network (WPCN) time frame. The offloading strategy includes the backscattering time t transmitted by the WD during that WPCN time frame. b,j Active transmission time t a,j and active transmission power P a,j ; A deterministic strategy deep gradient DDPG model is used to determine the unloading strategy based on the environmental parameters. as well as A processor is configured to invoke the program code and the DDPG model to execute the task unloading method as described in any one of claims 1 to 6.

8. A task offloading system, comprising a hybrid access node (HAP) and multiple wireless devices (WD), characterized in that, The HAP includes the task unloading device as described in claim 7.

Images