Comprehensive quality assessment method of CORS station network based on particle swarm optimization

Abstract

The present invention proposes a comprehensive quality assessment method for CORS station networks based on the particle swarm optimization algorithm, which includes the following steps: A. Obtain the urban coordinates of the CORS network and form an undirected graph G; B. Initialize; C. Take any point in the undirected graph G with positive weights as the starting point S, initialize a batch of initial solutions x and initialize the particle velocity v as the current population; D. Calculate the fitness value fit[i] of each particle; E. If fit[i] < Pbest[i], then replace Pbest[i] with fit[i]; F. If fit[i] < gbest[i], then replace gbest[i] with fit[i]; G. Iteratively update the velocity v i and the position x i ; H. Perform boundary condition processing; I. Control the final convergence of the system behavior based on the method of compression factor; J. Judge whether the algorithm termination condition is satisfied. If so, end the algorithm and output the optimization result; otherwise, return to step A. The comprehensive quality assessment method for CORS station networks based on the particle swarm optimization algorithm of the present invention realizes a dynamic global quality assessment from a whole network to even a region.

Description

Technical Field

[0001] The present invention relates to the technical field of satellite positioning data processing, and in particular to a satellite navigation positioning reference station network (Continuous Operational Reference Station, CORS) comprehensive quality assessment method based on particle swarm algorithm. Background Art

[0002] The satellite navigation and positioning base station network is the product of the integration of multi-source technologies such as the Global Navigation Satellite System (GNSS) positioning technology, network communication technology, and modern geodetic measurement. It is the most important component of the modern surveying and mapping spatial data infrastructure. It can monitor the transient and long-term changes in regional and even global crustal movement and meteorological state changes, and provide various high-precision spatial positioning services and diversified information services. Whether it is the International Earth Rotation and Framework Service (IGS) or many countries or regions, they are establishing global and regional coordinate reference frames. However, with the establishment of more and more CORS station networks in provinces, cities and enterprises, the development of various industries has been closely linked to it, especially the promotion of the modernization of "digital cities" in various cities, such as natural resources, geological disasters, smart agriculture, real-time disaster prediction, collaborative navigation, precision agriculture and modern management of traffic information, which are increasingly inseparable from the information services of the CORS network. Therefore, as a CORS system that enables applications in various industries, its stable operation is the basis of information services. Therefore, evaluating its data quality and stability is a key technology.

[0003] The particle swarm optimization (PSO) algorithm is inspired by the predation behavior of birds and imitates this behavior. It compares the search space of the optimization problem to the flight space of birds, and abstracts each bird as a particle. The particle has no mass and no volume, which is used to represent a feasible solution to the problem. The optimal solution to the optimization problem is equivalent to the food source that birds are looking for, and then a simplified model is established using swarm intelligence. The PSO algorithm formulates simple behavioral rules similar to bird movement for each particle, so that the movement of the entire particle group shows similar characteristics to bird predation, so that complex optimization problems can be solved. Like other evolutionary algorithms, it is also based on the concepts of "population" and "evolution", and realizes the search for the optimal solution in complex space through cooperation and competition between individuals. At the same time, it does not perform evolutionary operator operations such as crossover, mutation, and selection on individuals like other evolutionary algorithms. Due to the small number of parameters and easy implementation, it has a strong global search capability for nonlinear and multi-peak problems, and has received widespread attention in scientific research and engineering practice. Recently, the algorithm has been widely used in function optimization, neural network training, pattern classification, fuzzy control and other fields.

[0004] At present, the quality assessment of the CORS station network generally uses the collected GNSS observation data for multipath, data rejection rate and other quality analysis, and combines the static results after static network solution or network adjustment results for quality analysis. Since these methods are limited to the local level of data quality assessment or the relative accuracy of the weakest point after CORS station network adjustment, there is no dynamic global quality assessment result for the entire network or even the region, and there is a lack of multi-system, multi-level and multi-dimensional assessment methods. Summary of the invention

[0005] The present invention aims to at least solve the technical problems existing in the prior art, and in particular innovatively proposes a CORS station network comprehensive quality assessment method based on a particle swarm algorithm.

[0006] In order to achieve the above-mentioned object of the present invention, the present invention provides the following technical steps to achieve:

[0007] A. Obtain the city coordinates of the CORS network and form an undirected graph G.

[0008] B. Initialize iteration parameters iter, group size N, inertia weight ω, individual learning weight c1, and social learning weight c2.

[0009] C. Take any point in the undirected graph G with positive weights as the starting point S and initialize a batch of initial solutions x (x < x max ), initialize particle velocity v(v<v max ), as the current population (current solution set).

[0010] wherein, x and x max respectively represent the position of the particle and the set boundary position;

[0011] v and v max respectively represent the movement speed of the particle and the set boundary speed, that is, the maximum movement speed of the particle;

[0012] D. Calculate the fitness value fit[i] of each particle.

[0013] E. For each particle, compare its fitness value fit[i] with the individual extreme value Pbest[i]. If fit[i] < Pbest[i], then replace Pbest[i] with fit[i] and execute the next step. If fit[i] ≥ Pbest[i], then execute the next step.

[0014] F. For each particle, compare its fitness value fit[i] with the global extreme value gbest[i]. If fit[i] < gbest[i], then replace gbest[i] with fit[i] and execute the next step. If fit[i] ≥ gbest[i], then execute the next step.

[0015] G. Iteratively update the velocity v i and the position x i .

[0016] H. Perform boundary condition processing. When the position or speed exceeds the set value, adopt the boundary condition processing strategy to limit the position of the particle within the feasible search space. This can avoid the expansion and divergence of the population, and also avoid the large-scale blind search of the particle, thereby improving the search efficiency. There are many specific methods, as shown in the following formula. By setting the maximum position limit x max and the maximum speed limit v max , when exceeding the maximum position limit or the maximum speed limit, randomly generate a value within the range to replace it, or set it to the maximum value, that is, boundary absorption.

[0017] I. Control the final convergence of the system behavior based on the method of the compression factor. This method can effectively compress the search space and obtain high-quality solutions. The specific formula is as follows:

[0018]

[0019] wherein,

[0020] In the formula: λ is the compression factor calculated in the current iteration;

[0021] is the sum of the individual learning and social learning weights;

[0022] c1 represents the individual learning weight;

[0023] c2 represents the social learning weight;

[0024] J. Determine whether the algorithm termination condition is met. If so, end the algorithm and output the optimization result; otherwise, return to step A.

[0025] K. Calculate the angle closure error of the entire CORS network and the coordinate increment values ​​in each direction based on the searched global optimal route, and perform difference analysis with previous periods.

[0026]

[0027] Where: (x i ,y i ) represents the position coordinates of the i-th reference station in the shortest path;

[0028] (x i-1 ,y i-1 ) represents the position coordinates of the i-1th reference station in the shortest path;

[0029] f a is the azimuth closure error of the optimal path found in the undirected graph G; f Δx 、f Δy is the closure error in the X and Y directions of the optimal path plane found in the undirected graph G;

[0030] α b , α i They are the starting position and the recursive position respectively;

[0031] n is the dimension of the undirected graph G vector;

[0032] s is the sum of the optimal paths found in the undirected graph G;

[0033] L. Perform a comprehensive evaluation based on the calculation in step K to determine the accuracy of the current entire network.

[0034] In summary, due to the adoption of the above technical solutions, the present invention realizes a dynamic global quality assessment of the CORS station network based on the particle swarm algorithm, which is a whole network or even a region. The particle swarm algorithm utilizes the intelligent bionic optimization algorithm of the biological mechanism in nature. It does not rely on the strict mathematical properties of the optimization problem itself. It adapts to the environment better through mutual cooperation, shows the ability to interact with the environment, and has self-organization, evolution and memory functions. Combined with the results of the CORS network adjustment and the observation data collected in real time by the base station network, a multi-level quality assessment of the CORS network is achieved from the bottom data layer, the station network topology level, the physical comprehensive level, and the previous long time series. Provide a reasonable evaluation method for the quality stability of the surveying and mapping benchmark framework.

[0035] Additional aspects and advantages of the present invention will be given in part in the following description and in part will be obvious from the following description, or will be learned through practice of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0036] The above and / or additional aspects and advantages of the present invention will become apparent and easily understood from the description of the embodiments in conjunction with the following drawings, in which:

[0037] Figure 1 It is a schematic block diagram of the process of the present invention.

[0038] Figure 2 It is a schematic diagram of the implementation principle and process of the present invention. DETAILED DESCRIPTION

[0039] Embodiments of the present invention are described in detail below, examples of which are shown in the accompanying drawings, wherein the same or similar reference numerals throughout represent the same or similar elements or elements having the same or similar functions. The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain the present invention, and cannot be understood as limiting the present invention.

[0040] The present invention discloses a method for comprehensively evaluating the quality of CORS station network data. Figures 1-2 As shown in the figure: Initialize the particle swarm, including initializing the particle swarm parameters, initializing the coordinate sequence of the CORS station network, and establishing a combined optimization problem; then, based on the number of iterations, realize the optimal iterative search, calculate the distance vector between each station network in real time, calculate the individual optimal value and the social optimal value, and continuously update the speed and position vectors. For the speed and position vectors that exceed the boundary, perform boundary processing until an optimal path between all cities is found, that is, the Hamilton circuit with the smallest weight; finally, based on the optimal path found, recurse the angle closure error and the coordinate increment, and calculate the cumulative changes of the angle closure error and the coordinate closure error and the current change amount in combination with the previous data, and perform a comprehensive quality evaluation.

[0041] The specific steps include:

[0042] Step 1: Initialize iteration parameters iter, group size N, inertia weight ω, individual learning weight c1, social learning weight c2, CORS network undirected graph G, and initialize particle position x i , speed v i .

[0043] x i =(x ix ,x iy ),i=1,2…N,

[0044] v i =(v ix ,v iy ),i=1,2…N,

[0045] Where: N is the number of initialized particle population;

[0046] x ix 、x iy Respectively represent the position information of the i-th particle;

[0047] v ix 、v iy Respectively represent the velocity direction information of the i-th particle;

[0048] x i is the position search matrix consisting of N-dimensional particles;

[0049] v i is the velocity search matrix composed of N-dimensional particles;

[0050] Step 2: Take any point in the undirected graph G with positive weights as the starting point S and initialize a batch of initial solutions x (x < x max ), initialize particle velocity v(v<v max ), as the current population (current solution set).

[0051] Step 3: Calculate the fitness value of the individual particle. Based on the iterative sequence initialized in step 1, perform an iterative search in the feasible space of the undirected graph G of the entire CORS station network. The fitness value fit[i] of each particle can be calculated by formula (1). In order to enable local particles to find the optimal sequence, the exchanger s and exchange sequence ss are introduced. The definition of the exchanger s is as follows, where ss means that multiple exchanges are performed on the particles to obtain a new solution sequence x i , and calculate the individual fitness fit[i]. Determine whether fit[i] is less than the individual historical optimal value (individual extreme value) Pbest[i]. Pbest[i] can also be written as P best (t), if a replacement is to be made.

[0052] s=swapx(i,j),

[0053] ss=[swapx1,swapx2,…swapx i ],

[0054] Where: i, j represents the solution sequence x i middle index position;

[0055] swapx(i,j) means in the solution sequence x i Exchange the particles at index i and j;

[0056] s represents the solution sequence after one exchange sequence;

[0057] ss means multiple exchanges of x i The swap sequence of index positions in ;

[0058] Step 4: Determine whether the individual fitness of the particle fit[i] is less than the global historical optimal value (global extreme value) gbest[i]. gbest[i] can also be written as g best (t), if a replacement is to be made.

[0059] Step 5: Speed ​​update, based on the above steps, iteratively update the next generation particle speed v i (t+1), the formula is as follows:

[0060]

[0061] Where: v i (t+1) represents the next generation speed update value;

[0062] ω represents the inertia weight;

[0063] v i (t) represents the current generation speed update value;

[0064] Indicates that two exchange sequences are merged into one exchange sequence;

[0065] c1 and c2 are individual and social learning weights;

[0066] r1(t) and r2(t) represent a random number in this iteration, with a value between 0.1 and 0.6;

[0067] P best (t), g best (t) represent the individual and social (global) historical optimal values ​​respectively;

[0068] x i (t) represents the updated value of the particle position of the current generation;

[0069] Step 6: Update particle position x i , using the calculated individual fitness value and speed update value v i (t+1), calculate the position of the next generation of particles and provide the initial value for the next iterative search.

[0070] x i (t+1)=x i (t)+v i (t+1) (1)

[0071] Step 7: Perform boundary condition processing. When the position or speed exceeds the set value, the boundary condition processing strategy can limit the position of the particle within the feasible search space. This can avoid the expansion and divergence of the population, and also avoid the particles from blindly searching in a large range, thereby improving the search efficiency. There are many specific methods, as shown in the following formula: By setting the maximum position limit x max and the maximum speed limit v max ,When the maximum position limit or maximum speed limit is exceeded, a value is randomly generated within the ,range instead, or it is set to the maximum value, i.e. boundary absorption.

[0072] Step 8: Since the inertia weight, individual and social learning weight are all constants, it is not easy to iterate and converge quickly. In the iterative process, the above weight coefficients are adaptively iterated, and a method based on compression factor is used to control the final convergence of the system behavior. This method can effectively compress the search space and obtain high-quality solutions. The specific formula is as follows:

[0073]

[0074] in,

[0075] Where: λ is the compression factor calculated in the current iteration;

[0076] It is the sum of the weights of individual learning and social learning;

[0077] Step 9: Determine whether the algorithm termination condition is met. If so, end the algorithm and output the optimization result. Otherwise, return to step 1.

[0078] Step 10: Based on the searched global optimal route, calculate the angle closure error of the entire CORS network and the coordinate increment values ​​in each direction, and perform difference analysis with the previous period.

[0079]

[0080] Where: f a is the azimuth closure error of the optimal path found in the undirected graph G;

[0081] f Δx 、f Δy is the closure error in the X and Y directions of the optimal path plane found in the undirected graph G;

[0082] α b , α i They are the starting position and the recursive position respectively;

[0083] n is the number of nodes on the optimal path;

[0084] s is the sum of the optimal paths found in the undirected graph G;

[0085] Step 11: Based on the results calculated in step 10 and the historical closure error data of multiple periods, calculate the change in angle and coordinate closure error of this period and the historical cumulative change. The calculation formula is as follows, and then a comprehensive evaluation is performed to determine the accuracy of the current entire network.

[0086] Δα=α i -α0,

[0087] Δα i =α i -α i-1 ,

[0088] Where: Δα is the historical cumulative angle closure error;

[0089] α i , α0 are the angle and coordinate incremental closure errors and initial closure errors of this period respectively;

[0090] Δα i They are the change of the angle and coordinate increment closure error of this period and the closure error of the previous period respectively;

[0091] α i Indicates the change in the angle and coordinate increment closure error of this period;

[0092] α i-1 Indicates the change in angle and coordinate incremental closure error of the previous period;

[0093] When Δα, Δα i If the corresponding conditions are less than or equal to the set cumulative closure error accuracy threshold and the predetermined closure error accuracy threshold, respectively, then the condition is met.

[0094] Although the embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the claims and their equivalents.

Claims

1. A CORS station network comprehensive quality assessment method based on particle swarm algorithm, characterized in that: It includes the following steps: A. Obtain the urban coordinates of the CORS station network and form an undirected graph G; B. Initialize the iteration parameter iter, population size N, inertia weight ω, individual learning weight c1, and social learning weight c2; C. Take any point with positive weight in the undirected graph G as the starting point S, and initialize a batch of initial solutions x and initialize the particle velocity v as the current population; D. Calculate the fitness value fit[i] of each particle; E. For each particle, compare its fitness value fit[i] with the individual extreme value Pbest[i]: If fit[i] < Pbest[i], then replace Pbest[i] with fit[i]; F. For each particle, compare its fitness value fit[i] with the global extreme value gbest[i]: If fit[i] < gbest[i], then replace gbest[i] with fit[i]; G. Iteratively update the particle's velocity v i and position x i ; H. Perform boundary condition processing; I. Control the final convergence of the system behavior based on the method of compression factor; In step I, it includes: in, In the formula: λ is the compression factor calculated by the current iteration; It is the sum of the weights of individual learning and social learning; c1 represents the individual learning weight; c2 represents the social learning weight; J. Judge whether the algorithm termination condition is satisfied. If so, end the algorithm and output the optimization result; otherwise, return to step A; K. Calculate the angle closure error of the entire CORS station network and the coordinate increment values in each direction based on the searched global optimal route, and perform difference analysis with the previous period. In step K, it includes: Where: (x i ,y i ) represents the position coordinates of the i-th reference station in the shortest path; (x i-1 ,y i-1 ) represents the position coordinates of the i-1th reference station in the shortest path; f a is the azimuth closure error of the optimal path found in the undirected graph G; f Δx 、f Δy is the optimal path plane closure error found in the undirected graph G; α b , α i They are the starting position and the recursive position respectively; n is the dimension of the vector of the undirected graph G; s is the total path of the optimal path searched by the undirected graph G; L. Based on the calculation in step K, conduct a comprehensive evaluation to determine the accuracy of the current entire network. In step L, it includes: Dα=α i -a0, Da i =a i -a i-1 , In the formula: Δα is the historical cumulative angle closure error; α i , α0 are the angle and coordinate incremental closure errors and initial closure errors of this period respectively; Δα i , α i-1 They are respectively the change in the angle and coordinate incremental closure error of this period and the closure error of the previous period.

Images