Coding calculation-oriented power control and node selection method and equipment in edge environment

Abstract

The invention discloses a coding calculation-oriented power control and node selection method and equipment in an edge environment. The method comprises the following steps: acquiring task deadline, a base station power upper limit and channel and calculation related parameters, and presetting that at least a specified number of edge nodes finish subtasks on time as a success standard; establishing an optimization model by taking minimization of task failure probability as a target and taking the condition that the total transmission power does not exceed the upper limit of a base station and the power of a single node is non-negative as constraints; a non-convex problem corresponding to the model is approximately decomposed into multiple sub-problems, a transmission threshold power threshold value is introduced to be converted into a convex optimization problem, and a transmission power distribution scheme of each node is obtained based on a Lagrange function, a KKT condition and Newton iteration solution; and channel weights of candidate nodes are calculated and sorted, subsets of different scales are constructed, and an optimal scheme is selected to execute subtask distribution and power distribution. According to the method, the task success rate can be remarkably improved under the conditions of time delay constraint and lagging-behind nodes, and the system robustness and the resource utilization efficiency are improved.

Description

Technical Field

[0001] This invention relates to the field of distributed edge computing, and more specifically to a power control and node selection method and device for coding computation in an edge environment. Background Technology

[0002] With the rapid development of mobile internet and the Internet of Things (IoT), edge computing has become a crucial infrastructure supporting massive data processing and real-time applications. Distributed computing systems can significantly improve the parallel processing capabilities of tasks through multi-node collaboration; however, in actual operation, they still face the problem of lagging nodes. Due to factors such as wireless network fluctuations, system overload, and hardware failures, edge nodes are prone to excessive latency or even failure. The overall task completion time is often dominated by these slow nodes, thereby reducing system efficiency and reliability.

[0003] To alleviate the performance bottleneck caused by lagging nodes, distributed coding technology has been introduced into edge environments. Coding can improve fault tolerance through redundant subtask mechanisms. Taking Maximum Distance Separable (MDS) coding as an example, the system only needs to receive the results from any k fastest nodes to recover the complete task, thereby effectively reducing the computational risk caused by node latency or failure. However, in wireless edge scenarios, the channel conditions between the base station and multiple nodes differ significantly. If a uniform task allocation method is still used, the response time differences between different nodes will be further amplified, making it difficult to guarantee task deadlines.

[0004] On the other hand, edge computing tasks typically have strict latency constraints. Applications such as augmented reality, autonomous driving, and artificial intelligence inference services all require tasks to be completed within a specified timeframe. This necessitates that the system maximize the success rate of tasks with limited resources. In wireless transmission, the power allocation of the base station directly impacts the transmission rate and task completion probability of each node, making power control strategies crucial in such systems.

[0005] Existing research largely focuses on task coding strategies and redundancy selection, or only considers differences in node computational performance, but rarely systematically incorporates base station transmission power allocation as an optimization variable into the model. While some works address energy consumption and power issues, they often aim to maximize throughput or fairness, rather than maximizing the success rate of latency-sensitive tasks. Furthermore, existing methods typically do not adequately integrate node selection and power allocation, leading to insufficient resource utilization efficiency in heterogeneous channel environments.

[0006] Therefore, how to construct an optimization model oriented towards deadline constraints in a heterogeneous wireless channel environment by combining node selection and power control, so as to maximize the mission success rate, alleviate the lag effect and improve resource utilization efficiency, has become an important problem that urgently needs to be solved. Summary of the Invention

[0007] Purpose of the invention: To address the shortcomings of existing technologies, this invention provides a power control and node selection method and device for coding computation in edge environments. In environments with heterogeneous wireless channels and frequent node failures, it can effectively improve the success probability of tasks within the deadline by rationally allocating base station transmission power and selecting appropriate node sets. At the same time, in scenarios with uncertain parameters, it can reduce the task failure rate and improve resource utilization efficiency.

[0008] Technical solution: To achieve the above-mentioned objectives, the present invention adopts the following technical solution:

[0009] A power control and node selection method for coding computation in an edge environment includes the following steps:

[0010] Obtain task-related parameters, including task deadline, base station power limit, power attenuation parameter, ambient noise power, subtask data volume, wireless communication subchannel bandwidth, ideal minimum computing time for edge nodes, and edge node dropout probability parameter. The preset task success criterion is that at least a specified number of edge nodes complete the subtask within the task deadline.

[0011] An optimization model is constructed with the goal of minimizing the probability of task failure, and the constraints are that the sum of the transmission power of all edge nodes participating in the task does not exceed the upper limit of the base station power and the transmission power of each edge node participating in the task is non-negative.

[0012] The non-convex optimization problem corresponding to the optimization model is approximated and decomposed into multiple sub-problems, each sub-problem corresponding to a fixed-size edge node selection set; a transmission threshold power threshold is introduced for each sub-problem, and each sub-problem is transformed into a convex optimization problem. The solution conditions for each convex optimization problem are constructed based on the Lagrangian function, and the single-variable equations are solved using the KKT conditions and Newton's iteration method until the convergence condition is met, thus obtaining the transmission power allocation scheme for each edge node in the corresponding edge node selection set;

[0013] Calculate the channel weight of each candidate edge node, sort the candidate edge nodes according to the channel weight, and construct candidate node subsets of different sizes; for each candidate node subset, calculate the task success rate corresponding to each candidate node subset in combination with the corresponding transmission power allocation scheme; select the scheme with the highest task success rate from all candidate node subsets and corresponding transmission power allocation schemes as the final power control and node selection scheme; distribute subtasks and allocate transmission power according to the final scheme.

[0014] Furthermore, the optimization model is expressed as:

[0015]

[0016] The number of candidate edge nodes. Number of subtasks The set of nodes that participate in the task execution. The set of nodes that were successfully executed and returned. For power attenuation parameters, For nodes The probability of successfully completing the task. For nodes The probability of successfully completing the task. For nodes Transmission power, This represents the upper limit of base station power.

[0017] Furthermore, the transformed convex optimization problem is expressed as:

[0018]

[0019] This is a parameter representing the probability of edge nodes falling behind. , The deadline for the task. The minimum computation time under ideal conditions. The amount of data for each subtask For nodes Channel weights, The set of available edge nodes within the base station's communication range. The bandwidth of the sub-channel. This is the transmission power threshold.

[0020] Furthermore, the channel weight is determined by the edge node channel gain, the distance between the base station and the edge node, the power attenuation parameter, and the ambient noise power, and the calculation formula is as follows:

[0021]

[0022] Let be the channel gain of node i. Let i be the distance between node i and the base station. For environmental noise power, For power attenuation parameters;

[0023] The top n nodes with the largest weights are selected for calculation.

[0024] Furthermore, the transmission threshold power threshold is calculated based on the subtask data volume, subchannel bandwidth, task deadline, ideal minimum computation time, edge node channel gain, distance between the base station and the edge node, power attenuation parameters, and ambient noise power. The calculation formula is as follows:

[0025]

[0026] Let be the channel gain of node i. Let i be the distance between node i and the base station. For environmental noise power, For power attenuation parameters; The amount of data for each subtask The bandwidth of the sub-channel. The deadline for the task. This represents the minimum computation time under ideal conditions.

[0027] Furthermore, the Lagrange function is composed of specific function values ​​corresponding to each edge node, Lagrange multipliers, and transmission power constraint terms:

[0028]

[0029] The KKT conditions include:

[0030]

[0031] For nodes Transmission power, This is the upper limit of base station power. For transmission power threshold, This refers to the set of available edge nodes within the communication range of the base station. , Let n = t, t = k, k+1, k+2, ..., m be the set of nodes that participate in the task execution. For limiting conditions The Lagrange multiplier corresponding to )=0, For limiting conditions The corresponding Lagrange multipliers, For limiting conditions The corresponding Lagrange multipliers, For function The first derivative function, , The amount of data for each subtask The bandwidth of the sub-channel. , The deadline for the task. This represents the minimum computation time under ideal conditions. For nodes Channel weights.

[0032] Furthermore, the success rate of the task is calculated as follows:

[0033] Let the set of edge nodes participating in the task be... The set of edge nodes that successfully executed the task and returned the result is The probability of a single edge node i successfully completing the task is The number of edge nodes that successfully execute tasks When the target number falls within the range of a preset specified number to the total number of edge nodes participating in the task, the task success rate is the percentage of all nodes meeting this requirement. The sum of the product of the success probabilities of all successful edge nodes and the product of the failure probabilities of all unsuccessful edge nodes within the range is expressed as:

[0034]

[0035] When at least Among the selected nodes A task is considered successful if a subtask is completed within the deadline.

[0036] The present invention also provides an electronic device, comprising: one or more processors; a memory; and one or more programs, wherein the one or more programs are stored in the memory and configured to be executed by the one or more processors, wherein when the programs are executed by the processors, they implement the steps of the power control and node selection method for coding computation in an edge environment as described above.

[0037] The present invention also provides a computer-readable storage medium having a computer program stored thereon, wherein the computer program, when executed by a processor, implements the steps of the power control and node selection method for coding computation in an edge environment as described above.

[0038] The present invention also provides an edge computing system, characterized in that the system includes an edge server and a plurality of edge nodes, the edge server being integrated in a base station and wirelessly connected to the edge nodes; the edge server is configured to, according to the power control and node selection method for coding computation in an edge environment as described in the present invention, select a corresponding edge node to participate in the computation for each newly arrived task request, and allocate optimal transmission power to the selected edge node.

[0039] Beneficial Effects: The power control and node selection method for coding computation in edge environments provided by this invention improves the success rate of tasks under strict time constraints by jointly optimizing power control and node selection, effectively alleviating the performance bottleneck caused by heterogeneous channels and lagging nodes in edge environments. The use of approximation and convexity methods reduces the complexity of the original problem, enabling efficient solution of the optimization problem in real-world scenarios, improving decision-making efficiency, and avoiding power resource waste. This invention effectively mitigates the lagging effect and improves system robustness and resource utilization efficiency in situations with heterogeneous wireless channels and frequent lagging nodes, adapting to the practical application needs of latency-sensitive scenarios such as mobile edge computing and distributed coding computation. Attached Figure Description

[0040] Figure 1 This is a schematic diagram of task execution in a coded distributed computing system;

[0041] Figure 2 This is a flowchart of an optimized power control method for latency-sensitive coding computation in edge environments. Detailed Implementation

[0042] The technical solutions in the embodiments of the present invention will now be clearly and completely described in conjunction with the accompanying drawings.

[0043] Figure 1 This is a schematic diagram of task execution in a coded distributed computing system. The system consists of a base station and... It consists of available edge nodes within its communication range, and these nodes form a set. When the base station performs task offloading, it will select... Each node participates in the execution of the task; these nodes constitute a set. Subsequently, the task was divided into multiple parts, which the base station generated using coding techniques. There are 1 subtask, and the data volume of each subtask is 1. These encoded subtasks are assigned to a set. The edge nodes perform calculations, and after completing their respective calculations, the edge nodes return the results in sequence. Finally, the base station decodes the returned results to obtain the final result.

[0044] In this system, the base station communicates with edge nodes via a wireless network, and the base station's transmit power is limited by... There is a connection between the base station and each node. There are several channels, each with different channel gains. The base station allocates different power to these channels. (Base station and node) The power allocation and channel gain between them are denoted as follows: and Considering the heterogeneity of distances between the base station and each edge node, the distances between the base station and the nodes... The transmission rate between them is expressed as:

[0045]

[0046] in, Indicates the bandwidth of the sub-channel. For base stations and edge nodes The distance between them This represents the noise power. The corresponding transmission time is:

[0047]

[0048] The wireless environment is dynamic, meaning these parameters may vary with each task. The computation time for each node is also considered, taking into account the impact of lagging nodes. It is modeled as a random variable, and its cumulative distribution function is as follows:

[0049]

[0050] in, This represents the minimum computation time under ideal conditions. It indicates the possibility of falling behind.

[0051] Under the above conditions, node Deadline The probability of completing the task before completion is:

[0052]

[0053] like Then the node The inability to complete the task on time indicates that task execution is only meaningful when the transmission power exceeds a certain threshold. This threshold power is:

[0054]

[0055] make The effective channel quality weight represents the probability of completion. Represented as:

[0056] (1)

[0057] A task is considered successful when at least k out of n selected nodes complete their subtasks within the deadline. Let S represent the number of successful nodes, then the probability of task success is:

[0058]

[0059] The overall optimization problem (denoted as P1) aims to minimize the probability of task failure under the total power constraint:

[0060]

[0061] Problem P1 is difficult to solve directly. Therefore, it is transformed into multiple subproblems, which are solved sequentially. Specifically, The minimum value is (Without adding redundancy), the maximum value is (All available nodes participate in task execution). The solution is obtained by enumerating all possible values ​​of n. The original problem can be solved by solving several subproblems and comparing the results of each subproblem. Therefore, solving the optimal power allocation problem with a fixed number of n solves the original problem. Definition Let n be the set of nodes that participate in task execution when n = t (t = k, k+1, k+2, ..., m). The following lemma is proposed:

[0062] Lemma: When n is fixed, choose the one with the maximum The first n nodes are the optimal node selection strategy.

[0063] Proof: From equation (1) Follow Monotonically increasing. The set Replace any node k with another node that has Node j, in the same Down, By strictly increasing the value, the overall probability of success for the task is increased. Therefore, the choice is... The top n largest nodes are optimal.

[0064] Therefore, in order to maximize the probability of task success, we should start from the set Select The largest front A set of nodes However, the problem remains complex, and directly allocating power to the selected nodes is not feasible in real-time scenarios. To reduce complexity and computation time, the original problem is approximated, resulting in the reconstructed problem:

[0065]

[0066] However It remains a piecewise function of power P, and the problem is still a non-convex optimization problem, which is difficult to solve. Therefore, we define:

[0067]

[0068] The following lemma is proposed:

[0069] Lemma: For any , about It is a convex function.

[0070] Proof: Let ,in,

[0071] right about By finding the first and second derivatives, we can obtain:

[0072] ;

[0073] ;

[0074] because and , ,therefore and .because It is a convex and monotonically increasing function, while It is a convex function, therefore about It is a convex function.

[0075] Using the lemma, problem P2 can be rewritten as:

[0076]

[0077] Since P3 is a convex optimization problem, its optimal solution can be obtained by analyzing the KKT conditions. The Lagrangian function of P3 is:

[0078]

[0079] The KKT conditions for P3 are:

[0080] (2)

[0081] (3)

[0082] (4)

[0083] (5)

[0084] (6)

[0085] (7)

[0086] (8)

[0087] (9)

[0088] Let n be the set of nodes that participate in task execution when n = t (t = k, k+1, k+2, ..., m). For limiting conditions The Lagrange multiplier corresponding to )=0, For limiting conditions The corresponding Lagrange multipliers, For limiting conditions The corresponding Lagrange multipliers, For function The first derivative function, .

[0089] Based on this, the following conclusions can be drawn:

[0090] Lemma: The necessary and sufficient conditions for the optimal solution of P3 are as follows:

[0091] (10)

[0092] (11)

[0093] (12)

[0094] (13)

[0095] Proof: When =0 When the solution is not optimal, it is definitely not the optimal solution. Therefore, according to equation (9) in the KKT conditions, it can be known that... Furthermore, for (8), consider two cases: (i) At this point, according to equation (1), the probability of success is zero, therefore it is not feasible; (ii) From (6) we get , And obtained from (7) .because Therefore, (3) is automatically true, and in When (7) is satisfied, the above are the necessary and sufficient optimal conditions.

[0096] Since f'(x) is a monotonic function, it can be derived from (10) Represented as And (31) can be transformed into a statement about A univariate equation. If the solution to this equation satisfies (11)-(13), then the corresponding... This is the optimal solution to the subproblem; otherwise, no feasible optimal solution exists. Through... The global problem can be solved by enumerating each case and applying Newton's method to each case. The flowchart of the algorithm is as follows: Figure 2 .

[0097] The goal of this algorithm is to select the optimal set of nodes and allocate power for distributed computing under a total power constraint, thereby maximizing the probability of task success. Specifically, it first enumerates the minimum required number of nodes... Up to the maximum number of available nodes For all possible nodes, select the channel quality weights for each node. The first n nodes are selected as the candidate set. Then, using the Lagrange multiplier method combined with Newton's iteration, the power allocation for each node is solved under the power constraint, ensuring that the derivative of the failure probability function for each node satisfies the equality condition. Next, the success probability of the task under this power allocation is calculated and compared with the current best result. If it is better, the optimal power allocation scheme and the corresponding success probability are updated. Finally, the algorithm outputs the set of nodes that maximizes the success probability of the task and its power allocation scheme. (The diagram shows...) This indicates the probability of success in the current round. f' represents the best result among all rounds, and f' represents the first derivative of f.

[0098] According to an embodiment of the present invention, a power control and node selection method for coding computation in an edge environment is provided, specifically including the following steps:

[0099] Obtain task-related parameters, including task deadline, base station power limit, power attenuation parameter, ambient noise power, subtask data volume, wireless communication subchannel bandwidth, ideal minimum computing time for edge nodes, and edge node dropout probability parameter. The preset task success criterion is that at least a specified number of edge nodes complete the subtask within the task deadline.

[0100] An optimization model is constructed with the goal of minimizing the probability of task failure, and the constraints are that the sum of the transmission power of all edge nodes participating in the task does not exceed the upper limit of the base station power and the transmission power of each edge node participating in the task is non-negative.

[0101] The non-convex optimization problem corresponding to the optimization model is approximated and decomposed into multiple sub-problems, each sub-problem corresponding to a fixed-size edge node selection set; a transmission threshold power threshold is introduced for each sub-problem, and each sub-problem is transformed into a convex optimization problem. The solution conditions for each convex optimization problem are constructed based on the Lagrangian function, and the single-variable equations are solved using the KKT conditions and Newton's iteration method until the convergence condition is met, thus obtaining the transmission power allocation scheme for each edge node in the corresponding edge node selection set;

[0102] Calculate the channel weight of each candidate edge node, sort the candidate edge nodes according to the channel weight, and construct candidate node subsets of different sizes; for each candidate node subset, calculate the task success rate corresponding to each candidate node subset in combination with the corresponding transmission power allocation scheme; select the scheme with the highest task success rate from all candidate node subsets and corresponding transmission power allocation schemes as the final power control and node selection scheme; distribute subtasks and allocate transmission power according to the final scheme.

[0103] The optimization model is expressed as follows:

[0104]

[0105] The number of candidate edge nodes. Number of subtasks The set of nodes that participate in the task execution. The set of nodes that were successfully executed and returned. For power attenuation parameters, For nodes The probability of successfully completing the task. For nodes The probability of successfully completing the task. For nodes Transmission power, This represents the upper limit of base station power.

[0106] The transformed convex optimization problem is represented as:

[0107]

[0108] This is a parameter representing the probability of edge nodes falling behind. , The deadline for the task. The minimum computation time under ideal conditions. The amount of data for each subtask For nodes Channel weights, The set of available edge nodes within the base station's communication range. The bandwidth of the sub-channel. This is the transmission power threshold.

[0109] Furthermore, the channel weight is determined by the edge node channel gain, the distance between the base station and the edge node, the power attenuation parameter, and the ambient noise power, and the calculation formula is as follows:

[0110]

[0111] Let be the channel gain of node i. Let i be the distance between node i and the base station. For environmental noise power, For power attenuation parameters;

[0112] The top n nodes with the largest weights are selected for calculation.

[0113] Furthermore, the transmission threshold power threshold is calculated based on the subtask data volume, subchannel bandwidth, task deadline, ideal minimum computation time, edge node channel gain, distance between the base station and the edge node, power attenuation parameters, and ambient noise power. The calculation formula is as follows:

[0114]

[0115] Let be the channel gain of node i. Let i be the distance between node i and the base station. For environmental noise power, For power attenuation parameters; The amount of data for each subtask The bandwidth of the sub-channel. The deadline for the task. This represents the minimum computation time under ideal conditions.

[0116] Furthermore, the Lagrange function is composed of specific function values ​​corresponding to each edge node, Lagrange multipliers, and transmission power constraint terms:

[0117]

[0118] The KKT conditions include:

[0119]

[0120] For nodes Transmission power, This is the upper limit of base station power. For transmission power threshold, This refers to the set of available edge nodes within the communication range of the base station. , Let n = t, t = k, k+1, k+2, ..., m be the set of nodes that participate in the task execution. For limiting conditions The Lagrange multiplier corresponding to )=0, For limiting conditions The corresponding Lagrange multipliers, For limiting conditions The corresponding Lagrange multipliers, For function The first derivative function, , The amount of data for each subtask The bandwidth of the sub-channel. , The deadline for the task. This represents the minimum computation time under ideal conditions. For nodes Channel weights.

[0121] Furthermore, the success rate of the task is calculated as follows:

[0122] Let the set of edge nodes participating in the task be... The set of edge nodes that successfully executed the task and returned the result is The probability of a single edge node i successfully completing the task is The number of edge nodes that successfully execute tasks When the target number falls within the range of a preset specified number to the total number of edge nodes participating in the task, the task success rate is the percentage of all nodes meeting this requirement. The sum of the product of the success probabilities of all successful edge nodes and the product of the failure probabilities of all unsuccessful edge nodes within the range is expressed as:

[0123]

[0124] When at least Among the selected nodes A task is considered successful if a subtask is completed within the deadline.

[0125] The present invention also provides an electronic device, comprising: one or more processors; a memory; and one or more programs, wherein the one or more programs are stored in the memory and configured to be executed by the one or more processors, wherein when the programs are executed by the processors, they implement the steps of the power control and node selection method for coding computation in an edge environment as described above.

[0126] The present invention also provides a computer-readable storage medium having a computer program stored thereon, wherein the computer program, when executed by a processor, implements the steps of the power control and node selection method for coding computation in an edge environment as described above.

[0127] The present invention also provides an edge computing system, characterized in that the system includes an edge server and a plurality of edge nodes, the edge server being integrated in a base station and wirelessly connected to the edge nodes; the edge server is configured to, according to the power control and node selection method for coding computation in an edge environment as described in the present invention, select a corresponding edge node to participate in the computation for each newly arrived task request, and allocate optimal transmission power to the selected edge node.

Claims

1. A power control and node selection method for coding computation in an edge environment, characterized in that, The method includes the following steps: Obtain task-related parameters, including task deadline, base station power limit, power attenuation parameter, ambient noise power, subtask data volume, wireless communication subchannel bandwidth, ideal minimum computing time for edge nodes, and edge node dropout probability parameter. The preset task success criterion is that at least a specified number of edge nodes complete the subtask within the task deadline. An optimization model is constructed with the goal of minimizing the probability of task failure, and the constraints are that the sum of the transmission power of all edge nodes participating in the task does not exceed the upper limit of the base station power and the transmission power of each edge node participating in the task is non-negative. The non-convex optimization problem corresponding to the optimization model is approximated and decomposed into multiple sub-problems, each sub-problem corresponding to a fixed-size edge node selection set; a transmission threshold power threshold is introduced for each sub-problem, and each sub-problem is transformed into a convex optimization problem. The solution conditions for each convex optimization problem are constructed based on the Lagrangian function, and the single-variable equations are solved using the KKT conditions and Newton's iteration method until the convergence condition is met, thus obtaining the transmission power allocation scheme for each edge node in the corresponding edge node selection set; Calculate the channel weight of each candidate edge node, sort the candidate edge nodes according to the channel weight, and construct candidate node subsets of different sizes; for each candidate node subset, calculate the task success rate corresponding to each candidate node subset in combination with the corresponding transmission power allocation scheme; select the scheme with the highest task success rate from all candidate node subsets and corresponding transmission power allocation schemes as the final power control and node selection scheme; distribute subtasks and allocate transmission power according to the final scheme.

2. The method according to claim 1, characterized in that, The optimization model is represented as: ； The number of candidate edge nodes. Number of subtasks The set of nodes that participate in the task execution. The set of nodes that were successfully executed and returned. For power attenuation parameters, For nodes The probability of successfully completing the task. For nodes The probability of successfully completing the task. For nodes Transmission power, This represents the maximum power limit for the base station.

3. The method according to claim 2, characterized in that, The transformed convex optimization problem is represented as: ； This is a parameter representing the probability of edge nodes falling behind. , The deadline for the task. The minimum computation time under ideal conditions. The amount of data for each subtask For nodes Channel weights, The set of available edge nodes within the base station's communication range. The bandwidth of the sub-channel. This is the transmission power threshold.

4. The method according to claim 1, characterized in that, The channel weight is determined by the edge node channel gain, the distance between the base station and the edge node, the power attenuation parameter, and the ambient noise power. The calculation formula is as follows: ； Let be the channel gain of node i. Let i be the distance between node i and the base station. For environmental noise power, For power attenuation parameters; The top n nodes with the largest weights are selected for calculation.

5. The method according to claim 1, characterized in that, The transmission threshold power is calculated based on the subtask data volume, subchannel bandwidth, task deadline, ideal minimum computation time, edge node channel gain, distance between the base station and the edge node, power attenuation parameters, and ambient noise power. The calculation formula is as follows: ； Let be the channel gain of node i. Let i be the distance between node i and the base station. For environmental noise power, For power attenuation parameters; The amount of data for each subtask The bandwidth of the sub-channel. The deadline for the task. This represents the minimum computation time under ideal conditions.

6. The method according to claim 1, characterized in that, The Lagrange function consists of specific function values ​​corresponding to each edge node, Lagrange multipliers, and transmission power constraint terms: ； The KKT conditions include: ； For nodes Transmission power, This is the upper limit of base station power. For transmission power threshold, This refers to the set of available edge nodes within the communication range of the base station. , Let n = t, t = k, k+1, k+2, ..., m be the set of nodes that participate in the task execution. For limiting conditions The Lagrange multiplier corresponding to )=0, For limiting conditions The corresponding Lagrange multipliers, For limiting conditions The corresponding Lagrange multipliers, For function The first derivative function, , The amount of data for each subtask The bandwidth of the sub-channel. , The deadline for the task. This represents the minimum computation time under ideal conditions. For nodes Channel weights.

7. The method according to claim 1, characterized in that, The success rate of the task is calculated as follows: Let the set of edge nodes participating in the task be... The set of edge nodes that successfully executed the task and returned the result is The probability of a single edge node i successfully completing the task is The number of edge nodes that successfully execute tasks When the target number falls within the range of a preset specified number to the total number of edge nodes participating in the task, the task success rate is the percentage of all nodes meeting this requirement. The sum of the product of the success probabilities of all successful edge nodes and the product of the failure probabilities of all unsuccessful edge nodes within the range is expressed as: ； When at least Among the selected nodes A task is considered successful if a subtask is completed within the deadline.

8. An electronic device, comprising: One or more processors; Memory; And one or more programs, wherein the one or more programs are stored in the memory and configured to be executed by the one or more processors, wherein when the programs are executed by the processors, they implement the steps of the power control and node selection method for coding computation in an edge environment as described in any one of claims 1 to 7.

9. A computer-readable storage medium having a computer program stored thereon, the computer program being executed by a processor to implement the steps of the power control and node selection method for coding computation in an edge environment as described in any one of claims 1 to 7.

10. An edge computing system, characterized in that, The system includes an edge server and several edge nodes. The edge server is integrated into a base station and wirelessly connects with the edge nodes. The edge server is configured to perform power control and node selection for coding computation in an edge environment according to any one of claims 1 to 7. For each newly arrived task request, the corresponding edge node is selected to participate in the computation, and the optimal transmission power is allocated to the selected edge node.

Images