Multi-cell task unloading method based on non-orthogonal multiple access

Abstract

The invention relates to the technical field of electric power internet of things, and discloses a multi-cell task unloading method based on non-orthogonal multiple access, which is characterized by comprising the following steps of: constructing a multi-cell task unloading system model based on NOMA (Non-Orthogonal Multiple Access); based on the NOMA-based multi-cell task unloading system model, the total cost in the aspects of minimum energy and delay is optimized; and performing joint optimization on task unloading and resource allocation. Aiming at multi-user multi-cell edge computing task unloading in an electric power internet of things system, all factors influencing non-orthogonal multiple access efficiency, namely intra-cell and inter-cell interference, power control and sub-channel allocation, are jointly considered, and wireless resource allocation and task unloading are jointly optimized, so that the weighted sum of time delay and energy consumption in the system is reduced.

Description

Technical Field

[0001] The present invention relates to the field of electric power Internet of Things, and more specifically, to a multi-cell task unloading method based on non-orthogonal multiple access. Background Art

[0002] With the popularization of smart devices, various services and applications such as artificial intelligence-based object and fault identification in power systems continue to emerge, greatly facilitating the automated monitoring of power systems. However, these applications are usually computationally intensive and latency-sensitive, while the limited resources of smart devices cannot meet the real-time computing requirements of applications. Edge computing (MEC) provides computing, communication, and storage resources to the edge of the network, which can provide high-quality-of-service (QoS) solutions for these applications.

[0003] In the prior art, multi-user computing offloading is based on orthogonal multiple access (OMA). With the development of 5G and the Internet of Things, the number of devices has increased dramatically. Because the number of supported devices is strictly limited by the available orthogonal communication resources, traditional OMA-based multi-user offloading cannot support real-time QoS for massive devices.

[0004] Therefore, for the offloading of multi-user multi-cell edge computing tasks in the power Internet of Things system, the present invention provides a multi-cell task offloading method based on non-orthogonal multiple access, which improves the above technical problems. Summary of the invention

[0005] The disclosed embodiments aim to address the deficiencies of the prior art and provide a multi-cell task offloading method based on non-orthogonal multiple access. The present invention jointly considers all factors that affect the efficiency of non-orthogonal multiple access, namely, intra-cell and inter-cell interference, power control and sub-channel allocation, and jointly optimizes wireless resource allocation and task offloading to reduce the weighted sum of delay and energy consumption in the system; and solves the technical problem that multi-user offloading based on OMA cannot support real-time QoS for massive devices.

[0006] The above technical objective of the present invention is achieved through the following technical solution: A multi-cell task offloading method based on non-orthogonal multiple access comprises the following steps:

[0007] S1. Construct a multi-cell task offloading system model based on NOMA;

[0008] S2, a multi-cell task offloading system model based on NOMA, which optimizes and minimizes the total cost in terms of energy and delay;

[0009] S3,Joint optimization of task offloading and resource allocation.

[0010] As a preferred technical solution of the present invention, the process of constructing the communication model is:

[0011] When the device decides to offload tasks to the MEC server, there will be both intra-cell and inter-cell interference in the NOMA-assisted multi-cell MEC network; and They are respectively represented as the transmit power and channel coefficient from device n to MEC server m on subchannel k;

[0012] set up It determines the order of the following SIC; when device n offloads its task to MEC server m through subchannel k, the received signal to interference plus noise ratio can be expressed as:

[0013]

[0014] Where N0 is the noise power, is the interference power within the cell, Indicates the power of inter-cell interference; and is a binary variable representing the subchannel allocation result; indicates that subchannel k is allocated to device i for transmission to MEC server m; otherwise, represents the number of devices transmitting data through channel k in cell m;

[0015] Given the channel allocation result of the nth device The uplink data rate transmitted from device n to MEC server m through channel k can be described as:

[0016]

[0017] Where B is the bandwidth of each subchannel, is a binary variable representing the result of subchannel allocation; assuming that each device can occupy at most one subchannel for data transmission and there is a limit on the maximum number of devices occupying the same subchannel, the offloading decision needs to meet the following constraints:

[0018]

[0019]

[0020] in, Indicates the maximum number of devices that can be assigned to subchannel k.

[0021] As a preferred technical solution of the present invention, the process of constructing the calculation model is:

[0022] Assume that each device has a task T n , you can use the parameter tuple Tn = {b n , d n} to represent, where b n is the task size in bits, d n It is the CPU cycles required to compute one bit of the task;

[0023] For local computation, when device n performs its task locally, the task latency can be expressed as:

[0024]

[0025] Among them, f n is the computing power of device n; the energy consumption of local processing tasks is:

[0026]

[0027] Among them, z n It is a user-related constant that represents the energy consumption per CPU cycle;

[0028] The local computing cost function is expressed as the weighted sum of energy consumption and delay cost, and the function formula is:

[0029]

[0030] Among them, W e，n and w t，n are the weights of energy consumption and delay, respectively.

[0031] As a preferred technical solution of the present invention, for edge computing, when When , device n will offload its tasks to the MEC server through a sub-channel; the transmission delay of device n offloading to MEC server m through sub-channel k can be calculated as:

[0032]

[0033] Since the computing power of each MEC server is usually limited, the MEC server evenly distributes its idle computing resources to the devices connected to it; the task execution delay of the MEC server is described as:

[0034]

[0035] Among them, f m，n is the computing capacity allocated to device n by MEC server m; therefore, the latency of processing task n offloaded to MEC server m through subchannel k is:

[0036]

[0037] When offloading tasks to MEC servers, the total energy cost includes execution energy cost and transmission energy cost. Since MEC servers are usually deployed together with BSs, they have stable power supplies, so the execution energy cost of MEC servers can be ignored. The total energy cost is expressed as:

[0038]

[0039] Since the amount of data of the computational results in many computing applications is much smaller than the size of its input data, this paper ignores the download delay; the relative weighted total cost when offloading the task to the MEC server is:

[0040]

[0041] The total cost of processing task n is:

[0042]

[0043] in, It is a binary indicator of local calculation.

[0044] As a preferred technical solution of the present invention, based on the multi-cell task offloading system model of NOMA, an optimization function is constructed to optimize the total cost of minimizing energy and delay;

[0045] The optimization function is:

[0046]

[0047] The constraints are:

[0048]

[0049]

[0050]

[0051]

[0052] As a preferred technical solution of the present invention, the process of channel allocation is:

[0053] Let N m and are the number of devices and the device set in cell m respectively; the alliance game with transferable utility consists of a pair Definition, where is the set of game participants, U is a real-valued utility function; for each alliance Is a representation of the alliance C i The real number of the utility;

[0054] For subchannel allocation, the devices in the unit will perform tasks locally or offload tasks to the MEC server through K subchannels; the devices occupying the same subchannel k are represented as alliance C k , devices with local computing capabilities form alliance C0; there are K+1 alliances in total; C i The utility function can be expressed as:

[0055]

[0056] To maximize utility, the preference relations of players in the alliance game should be made clear to compare any two alliances and decide whether a device should join or leave an alliance;

[0057] The decision of device n depends on whether the utility of both coalitions will increase;

[0058] For both Alliance C i and C j , Indicates that device n strictly prefers to be in alliance C i Members of Alliance C j members of

[0059] Given a channel allocation result S, device n leaves its current alliance C j Join Alliance C i ,when This is a switching operation.

[0060] As a preferred technical solution of the present invention, the switching operation occurs during two stages of the subchannel allocation problem; the first is that the device chooses to offload tasks instead of local computing tasks; assuming that all devices are initialized to process their tasks locally, when device n finds that Under the condition that offloading via subchannel k is beneficial, the device will leave its current alliance C0 and join alliance C k , because its preference relation satisfies

[0061] The other is when the device finds a better subchannel i for offloading than its current subchannel j, then

[0062] When a device wants to switch to a subchannel k that is already occupied by its maximum number of devices, that is: Devices cannot simply join Alliance C k .

[0063] As a preferred technical solution of the present invention, the power control process is: given the result of channel allocation, it is used to show that the power control problem can be simplified to a convex problem and can be solved by a convex programming algorithm;

[0064] because in,

[0065] and

[0066] is a randomly selected variable;

[0067] The weighted cost upper bound of the nth device can be expressed as:

[0068]

[0069] in,

[0070]

[0071]

[0072] definition and The power control problem can be equivalently converted to the following problem:

[0073]

[0074] The constraints are:

[0075] In summary, the present invention has the following beneficial effects: for multi-user multi-cell edge computing task offloading in the power Internet of Things system, by jointly considering all factors affecting the efficiency of non-orthogonal multiple access, namely intra-cell and inter-cell interference, power control and sub-channel allocation, the wireless resource allocation and task offloading are jointly optimized to reduce the weighted sum of delay and energy consumption in the system. BRIEF DESCRIPTION OF THE DRAWINGS

[0076] Figure 1 A flowchart of a multi-cell task offloading method based on non-orthogonal multiple access provided by an embodiment of the present invention;

[0077] Figure 2 A model framework diagram of a multi-cell task offloading system based on NOMA provided in an embodiment of the present invention;

[0078] Figure 3 A schematic diagram of a channel allocation algorithm based on alliance game provided in an embodiment of the present invention;

[0079] Figure 4 A schematic diagram showing the comparison between the average system overhead and the number of task nodes provided in an embodiment of the present invention;

[0080] Figure 5 A schematic diagram for comparing the average system overhead and task size provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0081] The present application is described in detail below in conjunction with specific embodiments. The following embodiments will help those skilled in the art to further understand the present application, but do not limit the present application in any form. It should be noted that, for those of ordinary skill in the art, several variations and improvements can also be made without departing from the concept of the present application. These all belong to the protection scope of the present application.

[0082] In order to make the purpose, technical solution and advantages of the present application more clearly understood, the present application is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0083] It should be noted that, if there is no conflict, the various features in the embodiments of the present application can be combined with each other, all within the scope of protection of the present application. In addition, although the functional module division is performed in the device schematic diagram and the logical order is shown in the flow chart, in some cases, the steps shown or described can be performed in a sequence different from the module division in the device or the flow chart. In addition, the words "first", "second", "third", etc. used herein do not limit the data and execution order, but only distinguish the same items or similar items with basically the same functions and effects.

[0084] Unless otherwise defined, all technical and scientific terms used in this specification have the same meaning as those commonly understood by those skilled in the art to which this application belongs. The terms used in this specification and in the specification of this application are only for the purpose of describing specific embodiments and are not intended to limit this application. The term "and / or" used in this specification includes any and all combinations of one or more of the related listed items.

[0085] In addition, the technical features involved in each embodiment of the present application described below can be combined with each other as long as they do not conflict with each other.

[0086] The disclosed embodiments are intended to solve the problem that the number of supported devices is strictly limited by available orthogonal communication resources, and the traditional OMA-based multi-user offloading cannot support the real-time QoS of massive devices. In view of this, the disclosed embodiments propose a multi-cell task offloading method based on non-orthogonal multiple access, which jointly optimizes wireless resource allocation and task offloading by jointly considering all factors that affect the efficiency of non-orthogonal multiple access, namely, intra-cell and inter-cell interference, power control and sub-channel allocation, so as to reduce the weighted sum of delay and energy consumption in the system.

[0087] Please refer to Figure 1 , Figure 1The flowchart of the multi-cell task offloading method based on non-orthogonal multiple access according to the embodiment of the present disclosure is shown. The overall process mainly includes the following three steps:

[0088] Step 1: Multi-cell task offloading system model based on NOMA.

[0089] The NOMA-based multi-cell task offloading system model consists of a communication model and a computing model.

[0090] Specifically, Figure 2 As shown, the present invention provides multi-user offloading in a NOMA-assisted multi-cell MEC network, where there are M cells, and a MEC server is deployed with a base station (BS) in each cell. The multi-cell MEC network has N devices, each of which has a task that can be offloaded to the MEC server. Each device can only be connected to one base station, which is selected based on long-term channel quality measurements.

[0091] Assuming that the total system bandwidth is evenly divided into K orthogonal subchannels, shared by each cell, a device that needs to offload tasks to its associated MEC server can transmit through one subchannel.

[0092] Furthermore, subchannels can be shared by multiple devices based on NOMA, where successive interference cancellation (SIC) can be applied at the BS for signal decoding.

[0093] S1.1,Specifically, the steps for building the communication model are as follows:

[0094] When a device decides to offload tasks to the MEC server, both intra-cell and inter-cell interference will exist in the NOMA-assisted multi-cell MEC network. and They are respectively expressed as the transmission power and channel coefficient from device n to MEC server m on subchannel k, without loss of generality.

[0095] Assumptions It determines the order of the following SIC. When device n offloads its tasks to MEC server m through subchannel k, the received signal to interference plus noise ratio (SINR) can be expressed as:

[0096]

[0097] Where N0 is the noise power, is the interference power within the cell, Indicates the power of inter-cell interference. and is a binary variable representing the subchannel allocation result.

[0098] Specifically, Indicates that subchannel k is allocated to device i for transmission to MEC server m. Otherwise, Indicates the number of devices transmitting data through channel k in cell m.

[0099] Given the channel allocation result of the nth device The uplink data rate transmitted from device n to MEC server m through channel k can be described as:

[0100]

[0101] Where B is the bandwidth of each subchannel, is a binary variable representing the result of subchannel allocation. Assuming that each device can occupy at most one subchannel for data transmission and there is a limit on the maximum number of devices occupying the same subchannel, the offloading decision needs to meet the following constraints:

[0102]

[0103]

[0104] in Indicates the maximum number of devices that can be assigned to subchannel k.

[0105] S1.2, specifically, the steps for constructing the computational model are as follows:

[0106] Assume that each device has a task T n , you can use the parameter tuple T n = {b n , d n} to represent, where b n is the task size in bits, d n It is the number of CPU cycles required to compute one bit of a task.

[0107] For local computation, when device n performs its task locally, the task latency can be expressed as:

[0108]

[0109] Among them, f n is the computing power of device n (CPU cycles per second). The energy consumption of local processing tasks is:

[0110]

[0111] Among them, z n Is a user-related constant representing the energy consumption per CPU cycle.

[0112] The local computing cost function is expressed as the weighted sum of energy consumption and delay cost, and the function formula is:

[0113]

[0114] Among them, W e，n and w t，n are the weights of energy consumption and delay, respectively.

[0115] For edge computing, When , device n will offload its tasks to the MEC server through a sub-channel. The transmission delay of device n offloading to MEC server m through sub-channel k can be calculated as:

[0116]

[0117] Since the computing power of each MEC server is usually limited, the MEC server evenly distributes its idle computing resources to the devices connected to it. The task execution delay of the MEC server is described as:

[0118]

[0119] Among them, f m，n is the computing capacity allocated to device n by MEC server m. Therefore, the latency of processing task n by offloading to MEC server m through subchannel k is:

[0120]

[0121] When offloading tasks to MEC servers, the total energy cost includes execution energy cost and transmission energy cost. Since MEC servers are usually deployed together with BSs, they have stable power supplies, so the execution energy cost of MEC servers can be ignored. In this case, the total energy cost is expressed as:

[0122]

[0123] Since the amount of data of the computation results in many computing applications is much smaller than the size of its input data, this paper ignores the download delay. The relative weighted total cost when offloading the task to the MEC server is:

[0124]

[0125] In summary, the total cost of processing task n is:

[0126]

[0127] in It is a binary indicator of local calculation.

[0128] Step 2: A multi-cell task offloading system model based on NOMA is optimized to minimize the total cost in terms of energy and delay.

[0129] The optimization function that minimizes the total cost in terms of energy and delay is as follows:

[0130]

[0131] The constraints are as follows:

[0132]

[0133]

[0134]

[0135]

[0136] Step 3: Joint optimization of task offloading and resource allocation.

[0137] The task offloading and resource allocation are decomposed into two sub-problems, namely, channel allocation and power control, and the original problem is solved by iteratively optimizing the two sub-problems.

[0138] S3.1: Specifically, channel allocation. This application formulates the subchannel allocation problem as a coalition game.

[0139] Let N m and are the number of devices and the set of devices in cell m respectively. The coalition game with transferable utility consists of a pair Definition, where is the set of game participants, and U is a real-valued utility function. Is a representation of the alliance C i The real number of the utility.

[0140] For the subchannel allocation problem, the devices in the unit will perform tasks locally or offload tasks to the MEC server through K subchannels. The devices occupying the same subchannel k are represented as alliance C k , devices with local computing capabilities form alliance C0. There are K+1 alliances in total. i The utility function can be expressed as:

[0141]

[0142] To maximize utility, the preference relations of players in the alliance game should be made clear to compare any two alliances and decide whether a device should join or leave an alliance.

[0143] The decision of device n depends on whether the utility of both coalitions will increase.

[0144] Specifically, for the two alliances C i and C j , Indicates that device n strictly prefers to be in alliance C i Members of Alliance C j Members of.

[0145] Given a channel allocation result S, device n leaves its current alliance C j Join Alliance C i ,when This is a switching operation.

[0146] Switching operations may occur during two phases of the subchannel allocation problem. The first is when a device chooses to offload tasks instead of computing them locally. Assuming that all devices initially process their tasks locally, when device n finds Under the condition that offloading via subchannel k is beneficial, the device will leave its current alliance C0 and join alliance C k , because its preference relation satisfies Another case is when the device finds a better subchannel i for offloading compared to its current subchannel j, which means

[0147] However, when a device wants to switch to a subchannel k that is already occupied by its maximum number of devices, that is: Devices cannot simply join Alliance C k . At this point, another exchange operation is introduced.

[0148] Specifically, for device n∈C j and device n′∈C i , devices n and n′ are said to be in exchange if their utility functions satisfy the following condition:

[0149] U(C i \n′∪n)+U(C i \n∪n′)>U(C i )+(U j )

[0150] Then, based on the exchange operation and the exchange operation, it can be designed as: the coalition game algorithm for optimizing channel allocation, such as Figure 3 As shown, the device performs a swap operation and a swap operation until a final Nash equilibrium channel allocation result is achieved.

[0151] S3.2: Specifically, power control; given the results of channel allocation, it is used to show that the power control problem can be simplified to a convex problem and can be solved by a convex programming algorithm.

[0152] because in,

[0153] and

[0154] is a randomly selected variable.

[0155] Based on the above conditions, the weighted cost upper limit of the nth device can be expressed as:

[0156]

[0157] in,

[0158]

[0159]

[0160] definition and The power control problem can be equivalently converted to the following problem:

[0161]

[0162] The constraints are:

[0163] By taking the second-order derivative, we can see that the above problem is a convex optimization problem, which can be solved by commonly used convex optimization algorithms such as the interior point method.

[0164] Example: Assume there is a 100m×100m square area with 4 MEC servers at coordinates (25,25), (25,75), (75,25) and (75,75). All devices are randomly scattered and each device can only connect to one MEC server, which is usually selected based on long-term channel quality measurements.

[0165] like Figure 4-5 As shown, the simulation parameters that are not specified are summarized in the table and the following baselines are proposed for the following methods:

[0166] 1. Only swap operations are considered when allocating channels.

[0167] 2. The device first selects the channel that maximizes its utility function and then optimizes it using a swap operation (sortswap).

[0168] 3. Only switching operations are considered when allocating channels.

[0169] 4. The method that does not consider the interference from neighboring cells (SingleCell).

[0170] 5. Methods that do not consider power allocation (NoPowerControl).

[0171] like Figure 4 As shown, Figure 4 The average weighted cost versus the number of devices is shown.

[0172] Specifically, it can be seen that the average overhead increases with the number of devices. NoPowerContrl achieves a relatively higher overhead than the method proposed in this application, which may be due to two main reasons. On the one hand, NoPowerControl uses the maximum transmit power and does not perform power control, resulting in higher energy consumption. On the other hand, since the devices unload tasks at the same power, this will have an adverse effect on the efficiency of NOMA, resulting in higher overhead.

[0173] Since the swap operation only exists when all subchannels are occupied by the maximum number of devices, the switch operation is the main operation in the coalition game based on channel allocation when the number of devices is small. Furthermore, when the number of devices is small, the overhead of Swap is relatively high because it does not consider the situation where the device can switch to a better channel. In contrast, the switch operation occurs when there is a subchannel that can be occupied by other devices. When the number of devices is large, the overhead of Switch is high because it does not consider the situation where two devices occupying two different subchannels can swap with each other to achieve a smaller overhead. Compared with Swap, SortSwap has a performance improvement when the number of devices is small because it adds an initialization cycle where each device is matched to its most preferred subchannel before the swap operation.

[0174] Since the method proposed in this application takes into account all factors affecting NOMA efficiency, namely intra-cell and inter-cell interference, power control and sub-channel allocation, to fully unleash the potential of NOMA, the method proposed in this application achieves minimal overhead.

[0175] SingleCell achieves higher overhead because it does not consider inter-cell interference, resulting in an overestimation of the transmission rate. In this case, devices prefer to offload tasks to MEC servers, aiming to reduce the overhead of processing tasks. However, inappropriate offloading decisions will cause more devices to offload to MEC servers, resulting in more intense competition for communication and computing resources, leading to higher overhead.

[0176] like Figure 5 As shown, Figure 5The relationship between the average weighted total cost and the task bit size is shown. The number of devices is set to 40 and 240 respectively to fully utilize the performance comparison under different experimental environments.

[0177] Specifically, when the number of devices is 40, the number of subchannels is sufficient, and only switching operations are used when allocating subchannels. When the number of devices is 240, the communication channel is overloaded, and exchange and switching operations may occur. In general, the larger the task scale, the greater the weighted total cost. SingleCell still has the highest overhead, while the method proposed in this application achieves the lowest overhead because switching operations and exchange operations in subchannel allocation based on coalition game are considered. In addition, interference between and within cells and power control are optimized.

[0178] The above is only a preferred embodiment of the present invention, and the protection scope of the present invention is not limited to the above embodiments. All technical solutions under the concept of the present invention belong to the protection scope of the present invention. It should be pointed out that for ordinary technicians in this technical field, some improvements and modifications without departing from the principle of the present invention should also be regarded as the protection scope of the present invention.

Claims

1. A multi-cell task offloading method based on non-orthogonal multiple access, characterized in that: The method comprises the following steps: S1. Construct a multi-cell task offloading system model based on NOMA; S2, a multi-cell task offloading system model based on NOMA, which optimizes and minimizes the total cost in terms of energy and delay; S3,Joint optimization of task offloading and resource allocation.

2. The multi-cell task offloading method based on non-orthogonal multiple access according to claim 1, characterized in that: The NOMA-based multi-cell task offloading system model consists of a communication model and a computing model.

3. The multi-cell task offloading method based on non-orthogonal multiple access according to claim 2, characterized in that: The process of constructing the communication model is: When the device decides to offload tasks to the MEC server, there will be both intra-cell and inter-cell interference in the NOMA-assisted multi-cell MEC network; and They are respectively represented as the transmit power and channel coefficient from device n to MEC server m on subchannel k; set up It determines the order of the following SIC; when device n offloads its task to MEC server m through subchannel k, the received signal to interference plus noise ratio can be expressed as: Where N0 is the noise power, is the interference power within the cell, Indicates the power of inter-cell interference; and is a binary variable representing the subchannel allocation result; indicates that subchannel k is allocated to device i for transmission to MEC server m; otherwise, represents the number of devices transmitting data through channel k in cell m; Given the channel allocation result of the nth device The uplink data rate transmitted from device n to MEC server m through channel k can be described as: Where B is the bandwidth of each subchannel, is a binary variable representing the result of subchannel allocation; assuming that each device can occupy at most one subchannel for data transmission and there is a limit on the maximum number of devices occupying the same subchannel, the offloading decision needs to meet the following constraints: in, Indicates the maximum number of devices that can be assigned to subchannel k.

4. The multi-cell task offloading method based on non-orthogonal multiple access according to claim 2, characterized in that: The process of constructing the calculation model is: Assume that each device has a task T n , you can use the parameter tuple T n = {b n , d n }, where b n is the task size in bits, d n It is the CPU cycles required to compute one bit of the task; For local computation, when device n performs its task locally, the task latency can be expressed as: Among them, f n is the computing power of device n; the energy consumption of local processing tasks is: Among them, z n It is a user-related constant that represents the energy consumption per CPU cycle; The local computing cost function is expressed as the weighted sum of energy consumption and delay cost, and the function formula is: Among them, w e，n and w t，n are the weights of energy consumption and delay, respectively.

5. The multi-cell task offloading method based on non-orthogonal multiple access according to claim 4, characterized in that: For edge computing, When , device n will offload its tasks to the MEC server through a sub-channel; the transmission delay of device n offloading to MEC server m through sub-channel k can be calculated as: Since the computing power of each MEC server is usually limited, the MEC server evenly distributes its idle computing resources to the devices connected to it; the task execution delay of the MEC server is described as: Among them, f m，n is the computing capacity allocated to device n by MEC server m; therefore, the latency of processing task n offloaded to MEC server m through subchannel k is: When offloading tasks to MEC servers, the total energy cost includes execution energy cost and transmission energy cost. Since MEC servers are usually deployed together with BSs, they have stable power supplies, so the execution energy cost of MEC servers can be ignored. The total energy cost is expressed as: Since the amount of data of the computational results in many computing applications is much smaller than the size of its input data, this paper ignores the download delay; the relative weighted total cost when offloading the task to the MEC server is: The total cost of processing task n is: in, It is a binary indicator of local calculation.

6. The multi-cell task offloading method based on non-orthogonal multiple access according to claim 1, characterized in that: Based on the multi-cell task offloading system model of NOMA, an optimization function is constructed to optimize the total cost of minimizing energy and delay; The optimization function is: The constraints are:

7. The multi-cell task offloading method based on non-orthogonal multiple access according to claim 1, characterized in that: The task offloading and resource allocation are decomposed into two sub-problems, namely, channel allocation and power control, and the original problem is solved by iteratively optimizing the two sub-problems.

8. The multi-cell task offloading method based on non-orthogonal multiple access according to claim 7, characterized in that: The process of channel allocation is as follows: Let N m and are the number of devices and the device set in cell m respectively; the alliance game with transferable utility consists of a pair Definition, where is the set of game participants, U is a real-valued utility function; for each alliance Is a representation of the alliance C i The real number of the utility; For subchannel allocation, the devices in the unit will perform tasks locally or offload tasks to the MEC server through K subchannels; the devices occupying the same subchannel k are represented as alliance C k , devices with local computing capabilities form alliance C0; there are K+1 alliances in total; C i The utility function can be expressed as: To maximize utility, the preference relations of players in the alliance game should be made clear to compare any two alliances and decide whether a device should join or leave an alliance; The decision of device n depends on whether the utility of both coalitions will increase; For both Alliance C i and C j , Indicates that device n strictly prefers to be in alliance C i Members of Alliance C j members of Given a channel allocation result S, device n leaves its current alliance C j Join Alliance C i ,when This is a switching operation.

9. The multi-cell task offloading method based on non-orthogonal multiple access according to claim 8, characterized in that: The switching operation occurs during two phases of the subchannel allocation problem; the first is when a device chooses to offload tasks instead of computing them locally; assuming that all devices initially process their tasks locally, when device n finds Under the condition that offloading via subchannel k is beneficial, the device will leave its current alliance C0 and join alliance C k , because its preference relation satisfies The other is when the device finds a better subchannel i for offloading than its current subchannel j, then When a device wants to switch to a subchannel k that is already occupied by its maximum number of devices, that is: Devices cannot simply join Alliance C k .

10. The multi-cell task offloading method based on non-orthogonal multiple access according to claim 7, characterized in that: The power control process is as follows: given the result of channel allocation, it is used to show that the power control problem can be simplified to a convex problem and can be solved by a convex programming algorithm; because in, and ; is a randomly selected variable; The weighted cost upper bound of the nth device can be expressed as: in, definition and The power control problem can be equivalently converted to the following problem: The constraints are:

Images