Satellite computing resource allocation method based on weighted particle swarm algorithm

Abstract

The application discloses a satellite computing resource allocation method based on a weighted particle swarm algorithm, and relates to the field of low-orbit satellite network computing resource allocation.The application firstly defines the computing capacity of a cooperative computing network, then sets the weight of the cooperative computing network according to the computing capacity, then initializes the position and speed of particles in a particle swarm, and calculates the fitness, and finally obtains the optimal solution by updating the position and speed of the particles in the particle swarm, so that the satellite computing resource allocation is completed.The application is aimed at the scene that multiple satellites in a low-orbit satellite network cooperatively process a computing task, and proposes a method for calculating the weight based on the computing capacity of an on-board computer and the transmission bandwidth of an inter-satellite link, and the method is used for optimizing the search process of a particle swarm algorithm, so that the purpose of shortening the satellite computing task processing time is achieved.

Description

Technical Field

[0001] This invention relates to the field of computing resource allocation in low-Earth orbit (LEO) satellite networks, and is applicable to scenarios where multiple satellites in a LEO satellite network collaboratively process computing tasks. Specifically, it relates to a satellite computing resource allocation method based on a weighted particle swarm optimization algorithm. Background Technology

[0002] Low Earth Orbit (LEO) satellite networks typically consist of multiple satellites and a ground control center. In traditional LEO satellite networks, satellites generally transmit computing tasks to the ground control center for processing. In recent years, with the increasing computing power of onboard computers, it has become possible to perform computing tasks on satellites. Since the processing capacity of a single satellite is limited, computing tasks can be broken down and distributed via inter-satellite links with higher bandwidth, allowing multiple satellites to form a collaborative computing network for processing. In this process, to obtain computing results in the shortest possible time, it is necessary to study how to rationally distribute computing tasks to each satellite, i.e., computing resource allocation methods.

[0003] The basic idea of ​​Particle Swarm Optimization (PSO) is to find the optimal solution through cooperation and information sharing among individuals in a swarm, and it is widely used in fields such as function optimization. PSO initializes a swarm of particles in the feasible solution space. Each particle represents a potential optimal solution to an extreme value optimization problem. The characteristics of a particle are represented by three metrics: position, velocity, and fitness. Position represents the optimal solution, velocity is the value after each change, and fitness is the value corresponding to the optimal solution.

[0004] Currently, existing satellite computing resource allocation methods are inefficient, and a faster convergence method is needed. Summary of the Invention

[0005] In view of this, this invention proposes a method for obtaining the optimal allocation of computing resources using a weighted particle swarm optimization algorithm, specifically for scenarios involving multiple satellites collaboratively processing computing tasks in low-Earth orbit satellite networks. This method has a fast convergence speed and can shorten the processing time of satellite computing tasks.

[0006] To achieve the above objectives, the present invention adopts the following technical solution:

[0007] A satellite computing resource allocation method based on weighted particle swarm optimization algorithm includes the following steps:

[0008] Step 1: Combining the computing power of the onboard computer and the transmission bandwidth of the inter-satellite link, define the computing power S = S0, S1, ..., S... of the collaborative computing network. n The computing power of satellite i, S i for:

[0009]

[0010] Among them, C i To enhance the computing power of the satellite's onboard computer, The inter-satellite link transmission bandwidth from the satellite where the computing mission is located to the cooperating computing satellite;

[0011] Step 2, set the weights of the collaborative computing network to W = w0, w1, ..., w n , where w i The weight of satellite i is represented by W; W is derived through the computing power of the collaborative computing network.

[0012]

[0013] Where |||| represents modulo;

[0014] Step 3: Initialize the particles in the particle swarm, including random positions and velocities, and calculate their fitness:

[0015]

[0016] Where, p i It calculates the percentage of task D assigned to satellite i, C i For the computing power of the satellite's onboard computer, To calculate the inter-satellite link transmission bandwidth from the satellite hosting the computational mission to the cooperating computing satellite, l SiSj Let be the transmission delay from satellite i to satellite j;

[0017] Step 4: Update the particle velocity V of the particle swarm in each satellite. i =v i1 ,v i2 ,...,v iN :

[0018]

[0019] Where N is the particle swarm size, rand is a random number in the range (0,1), and X... i =x i1 ,x i2 ,...,x iN P represents the current position of the particle swarm. i =p i1 ,p i2 ,...,p iN G represents the local optimum of the particle swarm. i =g i1 ,g i2 ,...,g iN This is the globally optimal position for the particle swarm. and Here, K is the weighting factor, and K is the convergence factor. The calculation method is as follows:

[0020]

[0021] V i The maximum value is 1, and the minimum value is -1, when V i When >1, V i =1;

[0022] Then, based on the computing power of satellite i, V is recalculated. i Update:

[0023] V i =w i *V i (6)

[0024] Step 5, update the particle positions of the particle swarm in each satellite:

[0025] X i =X i +V i (7)

[0026] Limitation X i The value range of X is 0 to 1. i If the updated value exceeds this range, then the corresponding endpoint value will be used.

[0027] Step 6: Calculate the fitness based on the updated particle positions and record the optimal position P in this iteration. i If P i The result is better than the current global optimal position G. i Then update G i For P i Otherwise G i The fitness remains unchanged; then, it is compared with the user's acceptable computation time. If the user's requirements are met, the computation stops, and X... i If the maximum number of iterations has been reached, the calculation is complete; otherwise, steps 4 and 5 are repeated to continue iterating.

[0028] Compared with the prior art, the present invention has the following advantages:

[0029] 1. This invention calculates weights based on the computing power of the onboard computer and the transmission bandwidth of the inter-satellite link, and uses these weights to calculate particle velocity, thereby optimizing the search process and finding the optimal solution of the model in a shorter time, thus shortening the processing time of satellite computing tasks.

[0030] 2. The method of the present invention is simple, easy to implement, and can significantly improve the efficiency of satellite computing resource allocation. Attached Figure Description

[0031] Figure 1 This is a flowchart of an embodiment of the present invention. Detailed Implementation

[0032] The present invention will now be described in further detail with reference to the accompanying drawings and specific embodiments.

[0033] A satellite computing resource allocation method based on a weighted particle swarm optimization (PSO) algorithm is proposed. This method calculates weights based on the onboard computer's computing power and inter-satellite link transmission bandwidth, optimizing the PSO algorithm's search process to find the optimal solution for the model in a shorter time. The method specifically includes the following steps:

[0034] Step 1: Combining the computing power of the onboard computer and the transmission bandwidth of the inter-satellite link, define the computing power S = S0, S1, ..., S... of the collaborative computing network. n The computing power of satellite i, S i for:

[0035]

[0036] Among them, C i For the computing power of the satellite's onboard computer, The inter-satellite link transmission bandwidth from the satellite where the computing mission is located to the cooperating computing satellite;

[0037] Step 2, set the weights of the collaborative computing network to W = w0, w1, ..., w n , where w i The weight of satellite i is represented by W; W is derived through the computing power of the collaborative computing network.

[0038]

[0039] Where |||| represents modulo;

[0040] Step 3: Initialize the particles in the particle swarm, including random positions and velocities, calculate the task processing time, and use it as the fitness:

[0041]

[0042] Where, p i It calculates the percentage of task D assigned to satellite i, C i For the computing power of the satellite's onboard computer, To calculate the inter-satellite link transmission bandwidth from the satellite hosting the computational mission to the cooperating computing satellite, l SiSj Let be the transmission delay from satellite i to satellite j;

[0043] Step 4: Update the particle velocity V of the particle swarm in each satellite. i =v i1 ,v i2,...,v iN :

[0044]

[0045] Where N is the particle swarm size, rand is a random number in the range (0,1), and X... i =x i1 ,x i2 ,...,x iN P represents the current position of the particle swarm. i =p i1 ,p i2 ,...,p iN G represents the local optimum of the particle swarm. i =g i1 ,g i2 ,...,g iN This is the globally optimal position for the particle swarm. and Here, K is the weighting factor, and K is the convergence factor. The calculation method is as follows:

[0046]

[0047] Typically, K is set to 0.729. and The values ​​are 2.1 and 2 respectively.

[0048] After obtaining the particle velocity, if the velocity is too high, the particle may overshoot the local minimum; if the velocity is too low, the particle cannot explore sufficiently beyond the local minimum. Therefore, a limit is placed on the maximum and minimum velocity values: V i The maximum value is 1, and the minimum value is -1, when V i When >1, V i =1;

[0049] Then, based on the computing power of satellite i, V is recalculated. i Update:

[0050] V i =w i *V i (6)

[0051] Step 5, update the particle positions of the particle swarm in each satellite:

[0052] X i =X i +V i (7)

[0053] After the particle position is updated, the result needs to be evaluated. If the updated position is outside the search range, it needs to be restricted. Since the particle position X is...i That is, calculate the percentage p of tasks assigned to satellite i. i Therefore X i The value range of X is 0 to 1. If X is updated... i If the value exceeds this range, the corresponding endpoint value will be used.

[0054] Step 6: Calculate the fitness based on the updated particle positions and record the optimal position P in this iteration. i If P i The result is better than the current global optimal position G. i Then update G i For P i Otherwise G i The fitness remains unchanged; then, it is compared with the user's acceptable computation time. If the user's requirements are met, the computation stops, and X... i If the maximum number of iterations has been reached, the calculation is complete; otherwise, steps 4 and 5 are repeated to continue iterating.

[0055] Here is a more specific example:

[0056] like Figure 1 As shown, a satellite computing resource allocation method based on the weighted particle swarm optimization algorithm includes the following steps:

[0057] (1) Combining the computing power of onboard computers and the transmission bandwidth of inter-satellite links, the computing power of the collaborative computing network is defined as S = S0, S1, ..., S... n The computing power of each satellite is S i This can be represented by the following formula:

[0058]

[0059] Among them, C i For the computing power of the satellite's onboard computer, This refers to the inter-satellite link transmission bandwidth between the satellite where the computing mission is located and the cooperating computing satellite.

[0060] In a collaborative computing network, a computing task is generated by a satellite, which then breaks down the task and distributes it to other collaborative computing satellites for processing. The factors determining the task processing time are the computing power of each satellite and the transmission time of the inter-satellite links. The transmission time is mainly affected by the transmission bandwidth of the inter-satellite links. Therefore, the computing power of each satellite can be evaluated by considering the computing power of its onboard computer and the transmission bandwidth of the inter-satellite links.

[0061] (2) The weights of the collaborative computing network are W = w0, w1, ..., w n , where w iW represents the weight of satellite i. W can be obtained through the computational power of the collaborative computing network, as shown in the following formula:

[0062]

[0063] Where |||| represents modulo;

[0064] In a collaborative computing network, the weight of each satellite is determined by its computing power. This weight represents a trend in the iterative process of the particle swarm optimization algorithm: the stronger the computing power, the more tasks are allocated to that satellite.

[0065] (3) Initialize the particles in the particle swarm, including random positions and velocities, and calculate their fitness:

[0066]

[0067] (4) Update the particle velocities of the particle swarm in each satellite, in V i =v i1 ,v i2 ,...,v iN This indicates that an initial update will be performed using the following formula:

[0068]

[0069] Where N is the particle swarm size, X i =x i1 ,x i2 ,...,x iN P represents the current position of the particle swarm. i =p i1 ,p i2 ,...,p iN G represents the local optimum of the particle swarm. i =g i1 ,g i2 ,...,g iN This is the globally optimal position for the particle swarm. and Here, K is the weighting factor, and K is the convergence factor. The calculation formula is as follows:

[0070]

[0071] After obtaining the particle velocity, if the velocity is too high, the particle may overshoot the local minimum; if the velocity is too low, the particle cannot explore sufficiently beyond the local minimum. Therefore, the maximum and minimum velocities V must be limited. i The maximum value is 1, and the minimum value is -1, when V i When >1, V i =1.

[0072] When calculating particle velocities, each particle knows its best known position P so far. i We also know the best location G for finding all particles in the entire group. i The particles use the above information as their own experience and that of their companions to determine their next movement. In formula (4), the convergence factor is determined by the weighting factor, where... As an inertial factor, it allows the particles to maintain their inertia. The learning factor enables particles to explore new regions. The convergence factor determines the convergence speed; if the speed is too fast, the particle may overshoot the optimal solution, while if the convergence speed is too slow, it may wander outside the optimal solution region. Generally, K is set to 0.729. and The values ​​are 2.1 and 2 respectively.

[0073] After the initial velocity update is completed, Vi is weighted according to the computing power of satellite i, and the velocity is updated a second time according to the following formula:

[0074] V i =w i *V i (6)

[0075] (5) Update the particle positions of the particle swarm in each satellite, as shown in the following formula:

[0076] X i =X i +V i (7)

[0077] After the particle position is updated, the result needs to be evaluated. If the updated position is not within the search range, it needs to be restricted.

[0078] (6) Calculate the fitness based on the updated particle position and record the optimal position in this iteration as P. i If P i The result is better than the current global optimal position G. i Then update G i For P i Otherwise G i The fitness remains unchanged. Then, the fitness is compared with the expected result; if the threshold is met, the computation stops, and X... i If the maximum number of iterations has been reached, the calculation is complete; otherwise, step 4 is executed.

[0079] In summary, this invention first defines the computing power of the collaborative computing network, then sets the weights of the collaborative computing network based on the computing power, followed by initializing the positions and velocities of particles in the particle swarm and calculating their fitness. By updating the positions and velocities of the particles in the particle swarm, the optimal solution is obtained, thereby completing the allocation of satellite computing resources. This invention addresses the scenario of multiple satellites collaboratively processing computing tasks in low-Earth orbit satellite networks, proposing a method to calculate weights based on the computing power of onboard computers and the transmission bandwidth of inter-satellite links, and using this method to optimize the particle swarm optimization algorithm search process, thus achieving the goal of shortening the processing time of satellite computing tasks.

Claims

1. A satellite computing resource allocation method based on a weighted particle swarm algorithm, characterized in that, Comprising the steps of: Step 1, the computing capability of the cooperative computing network is defined as S=S0, S1,..., S n , where S0 is the computing capability of the onboard computer of satellite i, and S1,..., S are the computing capabilities of the inter-satellite links of satellite i. i The computing capability S of satellite i is: wherein C i is the computing power of the satellite i on-board computer, is the inter-satellite link transmission bandwidth from the satellite where the computing task is located to the cooperative computing satellite; Step 2, set the weights of the collaborative computing network as W = w0, w1,..., wn n where w i represent the weight of satellite i; W is derived from the computing power of the collaborative computing network: where ||| denotes modulo; Step 3, initialize the particles in the particle swarm, including random position and velocity, and calculate its fitness: where p i is the percentage of the computing task D assigned to satellite i, C i is the computing capacity of the satellite i on-board computer, is the inter-satellite link transmission bandwidth from the satellite where the computing task is located to the cooperative computing satellite, l SiSj is the transmission delay from satellite i to satellite j; Step 4, update the particle velocity V of the particle group in each satellite i = v i1 , v i2 ,..., v iN : where N is the size of the particle swarm, rand is a random number in the range of (0, 1), X i = x i1 , x i2 ,..., x iN is the current position of the particle swarm, P i = p i1 , p i2 ,..., p iN is the local optimal position of the particle swarm, G i = g i1 , g i2 ,..., g iN is the global optimal position of the particle swarm, and are weight factors, and K is a convergence factor, and the calculation is as follows: V i maximum of 1 and a minimum of -1, V i = 1 when V i > 1; Then, V is updated again according to the computing power of satellite i i V = V + (V - V) V i = w i *V i (6) Step 5, update the particle position of the particle swarm in each satellite: X i = X i + V i (7) Restriction X i X is in the range of 0 to 1, if X i If the updated value exceeds this range, the corresponding end value is taken. Step 6: Calculate the fitness based on the updated particle positions and record the optimal position P in this iteration. i If P i The result is better than the current global optimal position G. i Then update G i For P i Otherwise G i The fitness remains unchanged; then, it is compared with the user's acceptable computation time. If the user's requirements are met, the computation stops, and X... i If the maximum number of iterations has been reached, the calculation is complete; otherwise, steps 4-5 are repeated to continue iterating.

Images