A collaborative service multi-dimensional resource rapid optimization method based on a generalized service model

Abstract

The application discloses a kind of collaborative service multidimensional resource fast optimization method based on generalization service model, when the service to be handled arrives, each service is divided into a limited number of different types of subtasks according to the different structural characteristics of each service, according to the basic task flow logical topology structure, the internal logical relationship between the subtasks contained in a service is described, each subtask is assigned to an internet of things equipment to complete, an internet of things equipment can only handle a subtask in a service at the same time;In the service processing stage, first, allocate bandwidth, power to internet of things equipment and determine the subtask that each internet of things equipment needs to process to algorithm convergence, and the optimal strategy that the internet of things equipment can minimize the total energy consumption of the system is obtained and bandwidth, power allocation and internet of things equipment scheduling are carried out according to the strategy. In the multi-internet of things equipment system containing a variety of different types of collaborative services, the optimal bandwidth, power allocation and equipment scheduling decision can be made.

Description

Technical Field

[0001] This invention belongs to the field of wireless communication technology, specifically relating to a method for rapid optimization of multi-dimensional resources in cooperative services based on a generalized service model. Background Technology

[0002] The Internet of Things (IoT) is considered a technology capable of solving various social problems. IoT technology will play a crucial role in connecting everything to achieve ubiquitous intelligence, including smart homes, smart cities, and Industry 4.0. Furthermore, various IoT devices, such as smart home appliances, sensors, and drones, can connect to networks and interact with each other. These devices are used in various fields such as home automation, smart grids, traffic management, and healthcare. In multi-IoT smart home scenarios, multiple heterogeneous devices are deployed within a home. Through internet access, these devices can interconnect to provide collaborative services, and a nearby base station equipped with a server enhances computing power, meeting the needs of computationally intensive and latency-sensitive tasks to achieve ubiquitous computing. However, there are complex resource allocation optimization problems among massive heterogeneous IoT devices. This requires optimizing network bandwidth, device power allocation, and device scheduling to reduce overall system energy consumption, and also overcoming the increased algorithm time costs caused by the increasing number of IoT devices.

[0003] While existing research has explored various device-to-device resource sharing models and task flow logical topologies, it has not delved into the joint optimization of bandwidth allocation, power allocation, and device scheduling, nor has it addressed the increased algorithm time costs caused by the growing number of IoT devices. With the development of communication equipment and internet technology, network services will become more complex and diverse, rendering existing service models inapplicable to future needs. Therefore, for the diverse and numerous services of the future, a generalized service model is needed to optimize channel bandwidth and power allocation in the network, determine the optimal device scheduling decision, accelerate algorithm execution time, and obtain the optimal strategy that minimizes system energy consumption. Summary of the Invention

[0004] The technical problem to be solved by the present invention is to provide a method for rapid optimization of collaborative business multi-dimensional resources based on a generalized business model, addressing the shortcomings of the prior art.

[0005] To achieve the above-mentioned technical objectives, the technical solution adopted by the present invention is as follows:

[0006] A collaborative multi-dimensional resource optimization method based on a generalized business model includes:

[0007] Step 1: When a collaborative business occurs, obtain the types of subtasks contained in the collaborative business, the inherent logical relationships between each subtask, and the data size of each subtask based on the generalized business model;

[0008] Step 2: Allocate bandwidth and power to IoT devices based on the data obtained in Step 1, and obtain the energy required for each IoT device to complete the processing of each sub-task in this iteration;

[0009] Step 3: Based on the energy consumption data obtained in Step 2, schedule IoT devices to determine the sub-tasks that each IoT device needs to handle in this iteration;

[0010] Step 4: Repeat steps 1-3 above to complete the next iteration until the algorithm converges. At this point, the bandwidth and power allocation and IoT device scheduling decisions are the optimal decisions for the rapid allocation of multi-dimensional resources for collaborative services.

[0011] To optimize the above technical solution, the specific measures also include:

[0012] In step 1 above, the generalized business model divides each business into a finite number of different types of subtasks according to the different structural characteristics of each business. According to the basic task flow logic topology, it describes the inherent logical relationship between the subtasks contained in a business. Each subtask is assigned to an IoT device to complete. An IoT device can only process one subtask of a business at any given time.

[0013] The basic task flow logical topology structures mentioned above are divided into linear, tree-like, and... Figure 3 kind.

[0014] Step 2 above, which involves allocating bandwidth to IoT devices, includes the following steps:

[0015] Step 2.1: Given the power allocation coefficient p = {p1,...,p} N Initial equipment scheduling x = {x1, ..., x} N}, to handle the base station transmission power consumption of a service As the objective function, the time T for processing each business z is... z The constraints are that the threshold T is not exceeded and the sum of the bandwidth allocation coefficients of each IoT device is not more than 1.

[0016] Step 2.2: For the objective function Add a logarithmic barrier function and denote the constraint as f. z (a) ≤ 0, z = 1, ..., Z, the logarithmic barrier function is the logarithm of the negative values ​​of the constraints summed over all constraints, i.e., the logarithmic barrier function satisfies the formula: The objective function is transformed into

[0017] Step 2.3: Initialize bandwidth allocation coefficients. Given initial values ​​for the bandwidth allocation coefficients, calculate the objective function. By calculating the gradient and setting it to zero, we obtain the update formula for the bandwidth allocation coefficients.

[0018] Step 2.4: Update parameter k a =ν a k a , where ν a >1;

[0019] Step 2.5: Repeat steps 2.2-2.4 until the objective function is achieved. The value converges, and the bandwidth allocation coefficient 'a' for this iteration is obtained. * This refers to the bandwidth allocation strategy.

[0020] The method for initializing the bandwidth allocation coefficients described above is as follows: bandwidth is evenly allocated according to the number of devices N, meaning the bandwidth allocation coefficient a is initialized to satisfy...

[0021] Step 2 above, which involves power allocation for IoT devices, includes the following steps:

[0022] Step 2.6: Given the bandwidth allocation coefficient 'a' obtained in Step 2.5 * ={a1,...,a N The initial device scheduling is x = {x1, ..., x}. N As given in step 2.1, calculate the transmission energy E for processing a service z. z,trans The objective function is the total time T for processing each business z. z The transmit power of each IoT device does not exceed the threshold T, and the sum of the transmit power of each IoT device does not exceed the threshold P. sum For constraints;

[0023] Step 2.7: For the objective function E z,trans Add a logarithmic barrier function and denote the constraint as f. z (p)≤0, z=1,...,Z, the logarithmic barrier function is the logarithm of the negative values ​​of the constraints summed over all constraints, i.e., the logarithmic barrier function satisfies the formula: The objective function is transformed into

[0024] Step 2.8: Power Allocation Coefficient Initialization. Given the initial values ​​of the power allocation coefficients for the devices, calculate the objective function. By calculating the gradient and setting it to zero, we obtain the update formula for the power allocation coefficient.

[0025] Step 2.9: Update parameter kp =ν p k p , where ν p >1.

[0026] Step 2.10: Repeat steps 2.7-2.9 until the objective function is achieved. When the value converges, the device power allocation strategy p for this iteration is obtained. * .

[0027] The above-mentioned method for initializing the power allocation coefficient includes: setting the total power P... sum Power is allocated evenly to each device, meaning the device's initial power meets the requirements.

[0028] In step 3 above, l is used p -Box ADMM performs IoT device scheduling, including:

[0029] Step 3.1: Apply the binary constraint x∈{0,1} n The equivalent replacement is a set of continuous constraints, i.e., a cuboid. and an l2 norm sphere The intersection of these two sets of coordinates, and the formula for equivalent substitution, is expressed as follows: Where y1 remains inside the cuboid, and y2 is on the moving sphere l2;

[0030] Step 3.2: Give the augmented Lagrangian function of the equipment scheduling subproblem. f(x) represents the sum of energy consumption of Z services, i.e. The variable a is fixed. * ,p * Time is a function of x, η1 and η2 are Lagrange multipliers of the two equality constraints in the equipment scheduling problem, ρ1 and ρ2 are positive penalty parameters, and the variables in the k-th iteration are expressed as... The Lagrange multipliers at the k-th iteration are represented as follows:

[0031] Step 3.3: Project onto the cuboid First, update variable y1 so that The square of the L2 norm;

[0032] Step 3.4: Project onto the sphere First, update variable y2 so that

[0033] Step 3.5: Update the binary decision variable x, and fix the variable. Let the derivative of L(x,y1,y2,η1,η2) with respect to x be 0, then we get

[0034] Step 3.6: Update parameters η1, η2, ρ1, ρ2, and fix variables. The update of η1 satisfies The update of η2 satisfies The update method for ρ1 is as follows: The update method for ρ2 is as follows Where μ > 1 is a given parameter;

[0035] Step 3.7: Repeat steps 3.3-3.6 until the objective function is achieved. Once the values ​​converge, the device scheduling strategy for this iteration is obtained.

[0036] The present invention has the following beneficial effects:

[0037] When a service request arrives, this invention, based on a generalized service model, divides each service into a finite number of different types of subtasks according to their structural characteristics. Following three basic task flow logical topologies, it characterizes the inherent logical relationships between the subtasks within a service. Each subtask is assigned to an IoT device for completion, and an IoT device can only process one subtask from a single service at a time. During the service processing phase, the system first allocates bandwidth and power to the IoT devices and determines the subtasks each device needs to process. Once the collaborative service multi-dimensional resource allocation algorithm converges, the IoT devices obtain the optimal strategy that minimizes the system's total energy consumption and perform bandwidth and power allocation and IoT device scheduling accordingly. By utilizing joint multi-dimensional resource management and IoT device scheduling, it can make optimal bandwidth, power allocation, and device scheduling decisions in multi-IoT device systems containing various types of collaborative services. Attached Figure Description

[0038] Figure 1 This is a system model diagram of the present invention;

[0039] Figure 2 This is a flowchart of the allocation method of the present invention;

[0040] Figure 3 This is a flowchart of the bandwidth allocation process in this invention;

[0041] Figure 4 This is a flowchart of the power allocation process in this invention;

[0042] Figure 5 This is a flowchart of the equipment scheduling process in this invention;

[0043] Figure 6 There are three basic task flow logical topologies. Detailed Implementation

[0044] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0045] Although the steps in this invention are arranged by reference numerals, this is not intended to limit the order of the steps. Unless the order of the steps is explicitly stated or the execution of a step requires other steps as a basis, the relative order of the steps can be adjusted. It is understood that the term "and / or" as used herein refers to and covers any and all possible combinations of one or more of the associated listed items.

[0046] This invention discloses a collaborative multi-dimensional resource rapid allocation method based on a generalized business model. The model is business-centric and oriented towards multiple business types. It determines the devices executing sub-tasks through IoT device scheduling decisions and uses three basic task flow logical topologies to characterize the inherent logical relationships between business sub-tasks, representing the system's total latency and energy consumption, as well as the requirements for device bandwidth and transmit power allocation. The method is implemented based on a collaborative multi-dimensional resource rapid allocation system. An embodiment considers a smart home collaborative business processing system with multiple IoT devices, such as... Figure 1 As shown, when a collaborative service needs to be processed, the system, based on a generalized service model, divides each service into a finite number of different types of subtasks according to the different structural characteristics of each service. Following three basic task flow logical topologies, the system characterizes the inherent logical relationships between the subtasks within a service. Each subtask is assigned to an IoT device for completion, and an IoT device can only process one subtask from a single service at any given time. During the service processing phase, the system first allocates bandwidth and power to the IoT devices and determines the subtasks that each IoT device needs to process. Once the collaborative service multi-dimensional resource fast allocation algorithm converges, the IoT devices obtain the optimal strategy that minimizes the total system energy consumption and allocate bandwidth and power and schedule IoT devices accordingly.

[0047] Specifically, this invention provides a collaborative multi-dimensional resource rapid allocation method based on a generalized business model, employing a method of joint multi-dimensional resource management and IoT device scheduling, such as... Figure 2 As shown, the collaborative business multi-dimensional resource rapid allocation method includes:

[0048] Step 1: When a collaborative task arrives, the system obtains the type of subtasks contained in the collaborative task, the inherent logical relationship between each subtask, and the data size of the subtasks based on the generalized business model.

[0049] Step 2: The system allocates bandwidth and power to IoT devices based on the data obtained in Step 1, and obtains the energy required for each IoT device to complete the processing of each sub-task in this iteration;

[0050] Step 3: Based on the energy consumption data obtained in Step 2, schedule IoT devices to determine the sub-tasks that each IoT device needs to handle in this iteration;

[0051] Step 4: Repeat steps 1-3 above to complete the next iteration until the algorithm converges. At this point, the bandwidth and power allocation and IoT device scheduling decisions are the optimal decisions for the rapid allocation of multi-dimensional resources for collaborative services.

[0052] In this embodiment, in step 1, the inherent logical relationships between the subtasks in each business can be classified as linear, tree-like, and... Figure 3 The basic task flow logical topology structure is divided as follows: Figure 6 As shown.

[0053] In the embodiments, such as Figure 3 As shown, in step 2, bandwidth allocation for IoT devices includes the following steps:

[0054] Step 2.1: Given the power allocation coefficient p = {p1,...,p} N Initial equipment scheduling x = {x1, ..., x} N}, to handle the base station transmission power consumption of a service As the objective function, the time T for processing each business z is... z The constraints are that the threshold T is not exceeded and the sum of the bandwidth allocation coefficients of each IoT device is not more than 1.

[0055] Step 2.2: For the objective function Add a logarithmic barrier function and denote the constraint as f. z (a) ≤ 0, z = 1, ..., Z, the logarithmic barrier function is the logarithm of the negative values ​​of the constraints summed over all constraints, i.e., the logarithmic barrier function satisfies the formula: The objective function is transformed into

[0056] Step 2.3: Initialize the bandwidth allocation coefficient. The method for initializing the bandwidth allocation coefficient is as follows: distribute the bandwidth evenly according to the number of devices N, that is, initialize the bandwidth allocation coefficient a to satisfy...

[0057] Given the initial value of the bandwidth allocation coefficient Calculate the objective function By calculating the gradient and setting it to zero, we obtain the update formula for the bandwidth allocation coefficients.

[0058] Step 2.4: Update parameter k a =ν a k a , where ν a >1;

[0059] Step 2.5: Repeat steps 2.2-2.4 until the objective function is achieved. The value converges, and the bandwidth allocation coefficient 'a' for this iteration is obtained. * This refers to the bandwidth allocation strategy.

[0060] like Figure 4 As shown, step 2, power allocation for IoT devices includes the following steps:

[0061] Step 2.6: Given the bandwidth allocation coefficient 'a' obtained in Step 2.5 * ={a1,...,a N The initial device scheduling is x = {x1,...,x}. N As given in step 2.1, calculate the transmission energy E for processing a service z. z,trans The objective function is the total time T for processing each business z. z The transmit power of each IoT device does not exceed the threshold T, and the sum of the transmit power of each IoT device does not exceed the threshold P. sum For constraints;

[0062] Step 2.7: For the objective function E z,trans Add a logarithmic barrier function and denote the constraint as f. z (p)≤0, z=1,...,Z, the logarithmic barrier function is the logarithm of the negative values ​​of the constraints summed over all constraints, i.e., the logarithmic barrier function satisfies the formula: The objective function is transformed into

[0063] Step 2.8: Power allocation coefficient initialization, the method for initializing the power allocation coefficient includes: setting the total power P... sum Power is allocated evenly to each device, meaning the device's initial power meets the requirements.

[0064] Given the initial value of the power allocation factor of the device Calculate the objective function By calculating the gradient and setting it to zero, we obtain the update formula for the power allocation coefficient.

[0065] Step 2.9: Update parameter k p =ν p k p , where ν p >1.

[0066] Step 2.10: Repeat steps 2.7-2.9 until the objective function is achieved. When the value converges, the device power allocation strategy p for this iteration is obtained. * .

[0067] In the embodiments, such as Figure 5 As shown, in step 3, l is used p -Box ADMM performs IoT device scheduling, including:

[0068] Step 3.1: Apply the binary constraint x∈{0,1} n The equivalent replacement is a set of continuous constraints, i.e., a cuboid. and an l2 norm sphere The intersection of the two sets of y = x, y, and y = y can be represented by the formula for equivalent substitution as follows: Where y1 remains inside the cuboid, and y2 is on the moving sphere l2;

[0069] Step 3.2: Give the augmented Lagrangian function of the equipment scheduling subproblem. f(x) represents the sum of energy consumption of Z services, i.e. The variable a is fixed. * ,p * Time is a function of x, η1 and η2 are Lagrange multipliers of the two equality constraints in the equipment scheduling problem, ρ1 and ρ2 are positive penalty parameters, and the variables in the k-th iteration are expressed as... The Lagrange multipliers at the k-th iteration are represented as follows:

[0070] Step 3.3: Project onto the cuboid The variable y1 is updated to minimize the square of the L2 norm of the differences between y1, the value of x at the (k+1)th iteration, and the ratio of the Lagrange multiplier η1 to the positive penalty parameter ρ1 at the kth iteration.

[0071] Step 3.4: Project onto the sphere The variable y2 is updated to minimize the square of the L2 norm of the differences between y2, the value of x at the (k+1)th iteration, and the ratio of the Lagrange multiplier η2 to the positive penalty parameter ρ2 at the kth iteration.

[0072] Step 3.5: Update the binary decision variable x, and fix the variable. Let the derivative of L(x,y1,y2,η1,η2) with respect to x be 0. The closed-form solution for x is obtained by summing the product of ρ1 and y1 and the product of ρ2 and y2 at the k-th iteration, then subtracting this product from the product of η1 and η2 at the k-th iteration, and finally comparing this difference with the sum of ρ1 and ρ2 at the k-th iteration.

[0073] Step 3.6: Update parameters η1, η2, ρ1, ρ2, and fix variables. The update method for η1 is to multiply the value of ρ1 at the k-th iteration by the difference between x and y1 at the (k+1)-th iteration, and then sum this product with the value of η1 at the k-th iteration. In other words, the update of η1 satisfies... The update method for η2 is to multiply the value of ρ2 at the k-th iteration by the difference between x and y2 at the (k+1)-th iteration, and then sum this product with the value of η2 at the k-th iteration. That is, the update of η2 satisfies... The update method for ρ1 is to multiply the value of a given parameter μ > 1 with the value of ρ1 in the k-th iteration, that is... The update method for ρ2 is to multiply the value of a given parameter μ > 1 by the value of ρ2 in the k-th iteration, that is...

[0074] Step 3.7: Repeat steps 3.3-3.6 until the objective function is achieved. Once the values ​​converge, the device scheduling strategy for this iteration is obtained.

[0075] Example

[0076] like Figure 1 As shown, consider a smart home collaborative service processing multi-IoT device system consisting of N IoT devices and a base station with a server deployed on it. There are Z different types of services in the entire network, denoted as [example of service set]. A business process can have at most K subtasks that need to be processed, and the set of subtasks is denoted as K. These subtasks are completed according to a specific topology. Each subtask is assigned to an IoT device, and an IoT device can only handle one subtask from a given service. When IoT device n processes a service locally... subtasks When the time delay requirement cannot be met, the IoT device can offload the subtask to the base station for processing. Binary decision variable X offloadn,z,k, ∈{0,1} represents the task offloading decision. When task k in business z is processed on the local IoT device n, X offload,n,z,k =0; When task k in service z is offloaded to the base station for processing, X offload,n,z,k =1. Binary decision variable X offload,n,z,kLet x be the set of data. When processing a service z, orthogonal frequency division multiple access (OFDMA) technology is used to divide the total bandwidth W into N channels. Each IoT device occupies a different bandwidth depending on the subtask it is processing. Each subtask of service z can be represented as τ. z,k ={d z,k ,c z,k}, where d z,k c represents the input data size of subtask k of business z. z,k This indicates the CPU cycles required to complete the subtask computation.

[0077] With h z and Γ z Let η and d represent small-scale channel fading and large-scale channel fading, respectively. n Let represent the path loss exponent and the distance between the base station and device n, respectively. The channel power gain between the base station and IoT device n is the product of the squared magnitudes of the large-scale channel fading and the small-scale channel fading, and the negative power of the path loss exponent of the distance between the base station and device n. In other words, the channel power gain between the base station and device n is calculated according to the following formula:

[0078]

[0079] With matrix H z =(H z,i,g ) K×K Describe the inherent relationships between the subtasks processing business z. When subtask i needs the output data of subtask g, then H... z,i,g =1; otherwise, H z,i,g =0. The inherent correlation between the subtasks of business z can be expressed as:

[0080] H z,i,g ={0,1},1≤g<i≤K

[0081] f n and f BS Let represent the computing power of IoT device n and base station, respectively. When subtask k is processed on the local IoT device n, the local device execution time is the ratio of the CPU cycles required to complete subtask k to the computing power of IoT device n, i.e., the local device execution time T. n,z,k,exe Satisfy the following formula:

[0082]

[0083] When subtask k is offloaded to the base station for processing by device n, the base station execution time is the ratio of the CPU cycles required to complete the calculation of subtask k to the base station's computing power, i.e., the base station execution time T. BS,z,k,exe Satisfy the following formula:

[0084]

[0085] Subtask k is processed on one of two platforms. The total execution time of a service z is the local device execution time or base station execution time for processing subtask k, which is then summed over K subtasks and N IoT devices to obtain the total execution time T. z,exe Satisfy the following formula:

[0086]

[0087] When IoT device n offloads subtask k of service z to the base station, the achievable transmission rate R is... n,z,k,BS The signal power is the product of the bandwidth occupied by IoT device n and a logarithmic function, where the base of the logarithmic function is 2, the argument is 1, and the signal-to-noise ratio is the sum. The signal power is the channel power gain between the base station and IoT device n, multiplied by the transmit power p of IoT device n. n The noise power is the product of the bandwidth occupied by IoT device n and the noise power spectral density N0. That is, the feasible transmission rate R when IoT device n is offloading tasks. n,z,k,BS The calculation satisfies the following formula:

[0088]

[0089] Where a n,z,k This represents the bandwidth allocation coefficient a when the subtask k of service z is offloaded from IoT device n to the base station. n,z,k Let the set be denoted as a, and the transmit power be p. n The set of is denoted as p.

[0090] During the processing of a business transaction z, the sum of the bandwidth allocation coefficients of all IoT devices should not exceed 1, i.e., the bandwidth allocation coefficient a. n,z,k Satisfy the following formula:

[0091]

[0092] The total power of the devices in the entire system is denoted as P. sum The sum of the transmit power of all IoT devices should not exceed the threshold P. sum That is, the transmit power p n Satisfy the following formula:

[0093]

[0094] The base station transmission time T for processing a service z z,BS,trans The input data size of subtask k for business z and the feasible transmission rate R when the IoT device n unloads the task. n,z,k,BS The ratio of the two variables, X and X, is then used to determine the binary decision variable X. offload,n,z,kMultiply them, then sum them over the K subtasks and N IoT devices respectively, which is the base station transmission time T. z,BS,trans Satisfy the following formula:

[0095]

[0096] When device n transmits a subtask k of a service z to device m, what is the feasible transmission rate R for device-to-device communication? n,m,z,k,d2d The feasible transmission rate for device-to-device communication is the product of the device-to-device communication bandwidth and a logarithmic function, where the base of the logarithmic function is 2, the argument is 1, and the signal-to-noise ratio is the sum. Signal power is the product of the channel power gain between device n and device m and the transmit power of device n. Noise power is the product of the device-to-device communication bandwidth occupied by device n and the noise power spectral density N0. Therefore, the feasible transmission rate for device-to-device communication can be calculated according to the following formula:

[0097]

[0098] The device-to-device transmission time T for processing a service z z,d2d,trans The input data size of subtask k for business z and the feasible transmission rate R for device-to-device communication of IoT device n. n,m,z,k,d2d The ratio, and then the matrix H representing the subtask topology of business z. z,i,g Multiply them, then sum them over the K subtasks and N IoT devices respectively, which is the device-to-device transmission time T. z,d2d,trans Satisfy the following formula:

[0099]

[0100] The transmission time for processing a service z is the base station transmission time T. z,BS,trans With device-to-device transmission time T z,d2d,trans The sum, i.e., the transmission time T for processing a service z. z,trans Satisfy the following formula:

[0101] T z,trans =T z,BS,trans +T z,d2d,trans

[0102] In summary, the total time T for processing a single business transaction z is... z The total time T is the sum of the execution time and the transmission time. z Satisfy the following formula:

[0103] T z =T z,exe +T z,trans

[0104] Let T represent the time threshold for processing each service z, then the time T for processing each service z is... z It should not exceed the threshold T, i.e., Tz Satisfy the following formula:

[0105] T z ≤T

[0106] With e n,exe and e BS Let n and z represent the unit energy consumption of IoT device n and base station respectively when processing service z. Then the energy consumption E for executing a service z is... z,exe The execution energy consumption E generated by the device involved in processing business z z,device,exe The execution energy consumption E generated by the base station z,BS,exe Composition. The energy consumption E generated by the devices involved in processing business z. z,device,exe This can be expressed as the local device execution time T. n,z,k,exe The product of the energy consumption per unit of work when device n processes task z, and then summed over K subtasks and N IoT devices respectively, gives the execution energy consumption E generated by the device. z,device,exe Satisfy the following formula:

[0107]

[0108] The execution power consumption E generated by the base station z,BS,exe This can be expressed as the base station execution time T. BS,z,k,exe The energy consumption per unit of time that the base station consumes when processing service z is multiplied by the energy consumption of the base station, and then summed over K subtasks, which gives the execution energy consumption E generated by the base station. z,BS,exe Satisfy the following formula:

[0109]

[0110] The execution energy consumption E for processing a business z z,exe This can be expressed as the energy consumption E generated by the device. z,device,exe The execution energy consumption E generated by the base station z,BS,exe The sum of these is the energy consumption E. z,exe Satisfy the following formula:

[0111] E z,exe =(1-X offload,n,z,k E z,device,exe +X offload,n,z,k E z,BS,exe

[0112] With e n,trans This represents the unit energy consumption of IoT device n transmitting services. The transmission energy consumption for processing one service z can be divided into base station transmission energy consumption E. z,BS,trans Energy consumption E for device-to-device transmission z,d2d,trans E z,BS,trans The base station transmission time T for processing a service z can be represented. z,BS,trans The unit energy consumption e of subtask k of the transmission service z of device n n,trans Multiplication, i.e., E z,BS,transSatisfy the following formula:

[0113] E z,BS,trans =T z,BS,trans e n,trans

[0114] With E z,d2d,trans The device-to-device transmission time T represents the time to process a service z. z,d2d,trans The unit energy consumption e of subtask k of the transmission service z of device n n,trans Multiplication, i.e., E z,d2d,trans Satisfy the following formula:

[0115] E z,d2d,trans =T z,d2d,trans e n,trans

[0116] The transmission energy consumption E for processing a service z z,trans This can be expressed as the base station transmission energy consumption E. z,BS,trans Energy consumption E for device-to-device transmission z,d2d,trans The sum of, i.e., E z,trans Satisfy the following formula:

[0117] E z,trans =E z,BS,trans +E z,d2d,trans

[0118] In summary, the total energy consumption E for processing a single service z is... z It can be expressed as the sum of execution energy consumption and transmission energy consumption, i.e., total energy consumption E. z Satisfy the following formula:

[0119] E z =E z,exe +E z,trans

[0120] The objective function is to minimize the sum of energy consumption for Z services, while satisfying the constraint that the processing time T for each service z must be... z The total bandwidth allocation coefficients of all IoT devices should not exceed threshold T; the sum of the bandwidth allocation coefficients of all IoT devices should not exceed 1; and the sum of the transmit power of all IoT devices should not exceed threshold P. sum The value of the binary decision variable x is either 0 or 1. The joint optimization problem of network resource allocation and equipment scheduling involved in cooperative service processing can be represented as follows:

[0121]

[0122]

[0123]

[0124]

[0125] x∈{0,1}

[0126] Given power allocation coefficients p = {p1,...,p} N Initial equipment scheduling x = {x1, ..., x} N},by As the objective function, the time T for processing each business z is... z The sum of the bandwidth allocation coefficients for each IoT device should not exceed the threshold T, i.e., the sum of the threshold T and the bandwidth allocation coefficients for each IoT device should not exceed 1. and Let f be the constraint condition. z (a) ≤0, z=1,...,Z, for the objective function Add barrier function The objective function is transformed into Given the initial value of the bandwidth allocation coefficient Calculate the objective function By calculating the gradient and setting it to zero, we obtain the update formula for the bandwidth allocation coefficients. Update parameter k a =ν a k a , where ν a >1. Repeat the above steps until the objective function is achieved. When the value converges, the bandwidth allocation strategy 'a' for this iteration is obtained. * .

[0127] Given the bandwidth allocation strategy a obtained in this iteration * ={a1,...,a N} and the initial device scheduling x = {x1,...,x N},by The objective function is the total time T for processing each business z. z The transmit power of each IoT device should not exceed threshold T, and the sum of the transmit power of each IoT device should not exceed threshold P. sum , that is to and Let f be the constraint condition. z (p)≤0, z=1,...,Z, for the objective function Add barrier function The objective function is transformed into Given initial values ​​of power allocation coefficient Calculate the objective function By calculating the gradient and setting it to zero, we obtain the update formula for the power allocation coefficient. Update parameter k p =ν p k p , where ν p>1. Repeat the above steps until the objective function is achieved. When the value converges, the power allocation strategy p for this iteration is obtained. * .

[0128] Given the bandwidth allocation strategy a obtained in this iteration * ={a1,...,a N Power allocation strategy p * ={p1,...,p N},by Let x be the objective function, and let x be the binary decision variable with values ​​of 0 or 1 as the constraint. Let the binary constraint x∈{0,1} n The equivalent replacement is a set of continuous constraints, i.e., a cuboid. and an l2 norm sphere The intersection of the two sets of y = x, y, and y = y can be represented by the formula for equivalent substitution as follows: Where y1 remains within the cuboid, and y2 lies on the moving l2 sphere. Give the objective function. The augmented Lagrange function, f(x) represents the sum of energy consumption of Z services, i.e. The variable a is fixed. * ,p * Time is a function of x, η1 and η2 are Lagrange multipliers of the two equality constraints in the equipment scheduling problem, ρ1 and ρ2 are positive penalty parameters, and the variables in the k-th iteration are expressed as... The Lagrange multipliers at the k-th iteration are represented as follows:

[0129] By projecting onto the cuboid The variable y1 is updated to minimize the square of the L2 norm of the differences between y1, the value of x at the (k+1)th iteration, and the ratio of the Lagrange multiplier η1 to the positive penalty parameter ρ1 at the kth iteration. The closed-form solution can be obtained as follows:

[0130]

[0131] It is in [0,1] n The projection onto the rectangular prism space can be expressed as an element-wise function:

[0132]

[0133] By projecting onto the sphere The variable y2 is updated to minimize the square of the L2 norm of the differences between y2, the value of x at the (k+1)th iteration, and the ratio of the Lagrange multiplier η2 to the positive penalty parameter ρ2 at the kth iteration. The optimal solution can be obtained from lp sphere The result is obtained by the Euclidean projection on the surface, i.e.

[0134]

[0135] For any vector The two candidate values ​​for the upward projection can be calculated using the following formula.

[0136]

[0137]

[0138] Therefore, b in The projection on can be calculated using the following formula.

[0139]

[0140] Fixed variables Setting the derivative of the augmented Lagrangian function L(x,y1,y2,η1,η2) with respect to x to 0, we obtain a closed-form solution for x. This solution is obtained by summing the product of ρ1 and y1 and the product of ρ2 and y2 at the k-th iteration, then subtracting this product from the product of η1 and η2 at the k-th iteration, and finally comparing this difference with the sum of ρ1 and ρ2 at the k-th iteration.

[0141] Fixed variables The update method for η1 is to multiply the value of ρ1 at the k-th iteration by the difference between x and y1 at the (k+1)-th iteration, and then sum this product with the value of η1 at the k-th iteration. That is, the update of η1 satisfies... The update method for η2 is to multiply the value of ρ2 at the k-th iteration by the difference between x and y2 at the (k+1)-th iteration, and then sum this product with the value of η2 at the k-th iteration. That is, the update of η2 satisfies... The update method for ρ1 is to multiply the value of a given parameter μ > 1 with the value of ρ1 in the k-th iteration, that is... The update method for ρ2 is to multiply the value of a given parameter μ > 1 by the value of ρ2 in the k-th iteration, that is... Repeat the above steps until the objective function is achieved. The value converges.

[0142] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above, and that the invention can be implemented in other specific forms without departing from its spirit or essential characteristics. Therefore, the embodiments should be considered in all respects as exemplary and non-limiting, and the scope of the invention is defined by the appended claims rather than the foregoing description. Thus, all variations falling within the meaning and scope of equivalents of the claims are intended to be included within the present invention. No reference numerals in the claims should be construed as limiting the scope of the claims.

[0143] Furthermore, it should be understood that although this specification describes embodiments, not every embodiment contains only one independent technical solution. This narrative style is merely for clarity. Those skilled in the art should consider the specification as a whole, and the technical solutions in each embodiment can also be appropriately combined to form other embodiments that can be understood by those skilled in the art.

Claims

1. A collaborative multi-dimensional resource optimization method based on a generalized business model, characterized in that, include: Step 1: When a collaborative service arrives, the generalized service model is used to obtain the types of subtasks contained in the collaborative service, the inherent logical relationships between the subtasks, and the data size of the subtasks. The generalized service model divides each service into a finite number of subtasks of different types according to the different structural characteristics of each service. According to the basic task flow logical topology, the inherent logical relationships between the subtasks contained in a service are characterized. Each subtask is assigned to an IoT device for completion. An IoT device can only process one subtask of a service at a time. The basic task flow logical topology is divided into three types: linear, tree, and graph. Step 2: Allocate bandwidth and power to IoT devices based on the data obtained in Step 1, and obtain the energy required for each IoT device to complete the processing of each sub-task in this iteration; Step 3: Based on the energy consumption data obtained in Step 2, schedule IoT devices to determine the sub-tasks that each IoT device needs to handle in this iteration; use -Box ADMM performs IoT device scheduling, including: Step 3.1: Add binary constraints The equivalent replacement is a set of continuous constraints, i.e., a cuboid. and Norm sphere The intersection of these two sets of coordinates, and the formula for equivalent substitution, is expressed as follows: in Keep it within a cuboid. Moving on the sphere; Step 3.2: Give the augmented Lagrangian function of the equipment scheduling subproblem. , express The sum of the energy consumption of each business, that is It is a fixed variable , Time about The function, , These are the Lagrange multipliers of the two equality constraints in the equipment scheduling problem. , It is a positive penalty parameter, the first The variables at the next iteration are represented as follows: , No. The Lagrange multiplier representation at the next iteration is: ; Step 3.3: Project onto the cuboid Update variables , making , The square of the L2 norm; Step 3.4: Project onto the sphere Update variables , making ; Step 3.5: Update the binary decision variables Fixed variables ,make about The derivative is 0, therefore we get ; Step 3.6: Update parameters Fixed variables ,in The update satisfies ; The update satisfies ; The update method is ; The update method is ,in For the given parameters; Step 3.7: Repeat steps 3.3-3.6 until the objective function is achieved. Once the values ​​converge, the device scheduling strategy for this iteration is obtained. Step 4: Repeat steps 1-3 above to complete the next iteration until the algorithm converges. At this point, the bandwidth and power allocation and IoT device scheduling decisions are the optimal decisions for the rapid allocation of multi-dimensional resources for collaborative services.

2. The method for rapid optimization of collaborative business multi-dimensional resources based on a generalized business model according to claim 1, characterized in that, In step 2, bandwidth allocation for IoT devices includes the following steps: Step 2.1: Given the power allocation coefficient Initial equipment scheduling To handle the base station transmission power consumption of a service As the objective function, to process each business time Not exceeding the threshold The constraint is that the sum of the bandwidth allocation coefficients of each IoT device does not exceed 1. Step 2.2: For the objective function Add a logarithmic barrier function and denote the constraint as follows: The logarithmic barrier function is the negative logarithm of the constraint and the summation over all constraints; that is, the logarithmic barrier function satisfies the formula: The objective function is transformed into ; Step 2.3: Initialize bandwidth allocation coefficients. Given initial values ​​for the bandwidth allocation coefficients, calculate the objective function. By calculating the gradient and setting it to zero, we obtain the update formula for the bandwidth allocation coefficients. ; Step 2.4: Update parameters ,in ; Step 2.5: Repeat steps 2.2-2.4 until the objective function is achieved. The values ​​converge, yielding the bandwidth allocation coefficients for this iteration. This refers to the bandwidth allocation strategy.

3. The method for rapid optimization of collaborative business multi-dimensional resources based on a generalized business model according to claim 2, characterized in that, The method for initializing the bandwidth allocation coefficient is as follows: based on the number of devices. Uniform distribution, i.e., bandwidth allocation coefficient Initialization satisfies .

4. The method for rapid optimization of collaborative business multi-dimensional resources based on a generalized business model according to claim 2, characterized in that, Step 2, power allocation for IoT devices includes the following steps: Step 2.6: Given the bandwidth allocation coefficients obtained in Step 2.5 Initial equipment scheduling As given in step 2.1, process a business transaction. Transmission energy consumption As the objective function, to process each business Total time Not exceeding the threshold And the sum of the transmission power of all IoT devices does not exceed the threshold. For constraints; Step 2.7: For the objective function Add a logarithmic barrier function and denote the constraint as follows: The logarithmic barrier function is the negative logarithm of the constraint and the summation over all constraints; that is, the logarithmic barrier function satisfies the formula: The objective function is transformed into ; Step 2.8: Power Allocation Coefficient Initialization. Given the initial values ​​of the power allocation coefficients for the devices, calculate the objective function. By calculating the gradient and setting it to zero, we obtain the update formula for the power allocation coefficient. ; Step 2.9: Update parameters ,in . Step 2.10: Repeat steps 2.7-2.9 until the objective function is achieved. The values ​​converge, yielding the device power allocation strategy for this iteration. .

5. The method for rapid optimization of collaborative business multi-dimensional resources based on a generalized business model according to claim 4, characterized in that, The method for initializing the power allocation coefficient includes: setting the total power... Power is allocated evenly to each device, meaning the device's initial power meets the requirements. .

Images