CN114928882B - WLAN fingerprint positioning method based on self-adaptive Bayes comprehensive learning particle swarm optimization - Google Patents
WLAN fingerprint positioning method based on self-adaptive Bayes comprehensive learning particle swarm optimization Download PDFInfo
- Publication number
- CN114928882B CN114928882B CN202210434881.2A CN202210434881A CN114928882B CN 114928882 B CN114928882 B CN 114928882B CN 202210434881 A CN202210434881 A CN 202210434881A CN 114928882 B CN114928882 B CN 114928882B
- Authority
- CN
- China
- Prior art keywords
- fingerprint
- positioning
- particle
- self
- comprehensive learning
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
- 239000002245 particle Substances 0.000 title claims abstract description 66
- 238000005457 optimization Methods 0.000 title claims abstract description 27
- 238000000034 method Methods 0.000 title claims abstract description 26
- 238000004422 calculation algorithm Methods 0.000 claims abstract description 18
- 238000012549 training Methods 0.000 claims abstract description 11
- 238000005259 measurement Methods 0.000 claims abstract description 6
- 238000004364 calculation method Methods 0.000 claims abstract description 5
- 230000006870 function Effects 0.000 claims abstract description 4
- 230000006978 adaptation Effects 0.000 claims abstract 3
- 239000013598 vector Substances 0.000 claims description 18
- 241000764238 Isis Species 0.000 claims description 3
- 230000001186 cumulative effect Effects 0.000 claims description 3
- 241000287196 Asthenes Species 0.000 claims 1
- 238000005516 engineering process Methods 0.000 description 5
- 230000003044 adaptive effect Effects 0.000 description 4
- 238000013528 artificial neural network Methods 0.000 description 2
- 238000010586 diagram Methods 0.000 description 2
- 238000011160 research Methods 0.000 description 2
- 238000012706 support-vector machine Methods 0.000 description 2
- 238000012360 testing method Methods 0.000 description 2
- 238000007635 classification algorithm Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000002474 experimental method Methods 0.000 description 1
- 238000010801 machine learning Methods 0.000 description 1
- 230000001737 promoting effect Effects 0.000 description 1
- 239000007787 solid Substances 0.000 description 1
Abstract
The invention discloses a WLAN fingerprint positioning method based on self-adaptive Bayes comprehensive learning particle swarm optimization, which comprises similarity measurement and optimal position prediction calculation: 1) Calculating the correlation and cosine distance between the query fingerprint and the training fingerprint by using a double-panel uniformity model; 2) Establishing a positioning model fitness function; 3) And calculating the parameter of the optimal adaptation value by adopting a self-adaptive Bayes comprehensive learning particle swarm optimization algorithm and obtaining the optimal prediction position. The method can efficiently and accurately perform indoor positioning, adopts a double-sided board fingerprint uniformity model to measure the similarity of positioning fingerprints, measures the matching degree between fingerprints by utilizing the correlation, introduces cosine distances to reflect the difference of the fingerprints in the direction, and improves the diversity and the robustness of the positioning module; parameter optimization is carried out by adopting a self-adaptive Bayes comprehensive learning particle swarm optimization algorithm, and the self-adaptive mechanism adopts interval division to select the comprehensive learning probability level of particles so as to improve the performance of the particle swarm optimization algorithm. The system is more robust, and the positioning accuracy can be improved.
Description
Technical Field
The invention relates to the field of indoor positioning technology and optimization algorithm, in particular to a WLAN fingerprint positioning method based on self-adaptive Bayes comprehensive learning particle swarm optimization.
Background
As the indoor positioning system is increasingly in demand in modern society, many technologies (such as ultra wideband UWB, inertial sensor IMU, bluetooth, geomagnetic sensor, laser, etc.) are researched and developed, and meanwhile, the indoor positioning technology based on WiFi is focused on aspects of academic research and engineering application, etc. with the advantages of wide signal range, low deployment cost, no need of additional hardware deployment, no influence of non-line of sight, etc. As one of WiFi indoor positioning methods, a positioning method based on RSS fingerprint is the mainstream in WiFi indoor positioning because RSS is easy to directly obtain on a common commercial WiFi device, and the positioning method based on RSS fingerprint usually uses WiFi signals to characterize position information, and signal strength RSS, signal to noise ratio and channel state information are mainly studied and used as position identification fingerprints.
The accuracy and stability of positioning are the most focused or only focused indexes of a positioning service terminal user, however, as a wireless signal is easily influenced by external factors in a complex and changeable indoor environment, the signal receiving strength RSS has sensitive and changeable characteristics, and the fingerprint based on the RSS is not remarkable in the aspects of space recognition rate and time stability.
With more and more research on machine learning algorithms, methods based on Particle Filters (PFs), support Vector Machines (SVMs), neural Networks (NNs), classification algorithms, etc. are introduced into indoor positioning, and many positioning models are widely proposed and improve indoor positioning based on WiFi fingerprints to some extent. The limited precision and unstable performance of the existing algorithm become constraint factors for promoting the WiFi indoor positioning technology, the positioning model of many systems is not enough in robustness, and the average error is usually larger in practice.
The particle swarm optimization PSO algorithm is used as a biological evolution algorithm, and has better performance in the aspect of optimization. Many students have studied the influence of different parameter configurations on PSO algorithm, particle learning strategy, etc. on the basis. In the comprehensive learning particle swarm optimization method (BCLPSO) based on the Bayesian iteration method, particles take the particle position with the maximum posterior probability based on the Bayesian formula as a learning sample, so that the particles can be effectively prevented from being trapped into local optimization and omitting potential global optimal solutions, and the method has better applicability in the indoor positioning technology based on WiFi fingerprints.
Disclosure of Invention
The invention aims at overcoming the defects of the prior art, and provides a WLAN fingerprint positioning method based on self-adaptive Bayesian comprehensive learning particle swarm optimization. The method utilizes the correlation to measure the matching degree between fingerprints, introduces cosine distance to reflect the difference of the fingerprints in the direction, and improves the diversity and the robustness of the positioning module; and the parameter optimization is performed by adopting a self-adaptive Bayesian comprehensive learning particle swarm optimization algorithm, so that the system has higher robustness and the positioning accuracy can be improved.
In order to achieve the above purpose, the technical scheme provided by the invention comprises the following steps:
The WLAN fingerprint positioning method based on the self-adaptive Bayes comprehensive learning particle swarm optimization comprises the steps of carrying out fingerprint similarity measurement by adopting a double-sided board uniformity model, and carrying out positioning parameter optimization and position prediction by adopting a self-adaptive Bayes comprehensive learning particle swarm optimization algorithm, wherein the method comprises the following specific steps:
it is assumed that there are M signal access points AP in the indoor environment and N physical locations are uniformly selected as reference points RP. Is provided with AndRepresenting sets of APs and RPs, respectively. Is provided withRepresenting the position coordinates of the i (i=1, 2, …, N) th RP, and s ij represents the RSS of the j-th AP at that position. Defining the fingerprint (RSS vector) corresponding to the ith RP position asIn summary, the training fingerprint set of reference points RP in the positioning environment can be expressed as:
Is provided with Representing RSS vectors collected by the mobile device during the online phase for mobile device positioning. If a plurality of users perform positioning inquiry, the fingerprint set is expressed as:
Where Γ is the number of query fingerprints. If the jth AP cannot be detected in the positioning, a very small value is assigned to s ij.
1. Fingerprint similarity metric
The similarity of different fingerprints is measured by adopting a double-panel fingerprint uniformity graph representation method.
1) For the first panel, the correlation is used to evaluate the similarity of different fingerprints. For fingerprints Calculating the correlation between the two:
cor(u,v)＝1-r(u,v) (10)
Wherein, AndRespectively isIs defined as the mean value and standard deviation of (c),AndRespectively isMean and standard deviation of (c).
Selecting h AND gates according to equation (1)The fingerprint with the highest correlation is expressed as:
Wherein, Represents the q-th similar fingerprint and,And (3) withThe training fingerprints in (a) constitute the edges of the first panel graph.
2) For the second panel, for the fingerprintThe cosine distance between the two is calculated (reflecting the divergence of the fingerprint vector in terms of directionality):
similarly, h of the second panels are The most similar fingerprint of (2) is expressed as:
Wherein, Representing the q' th similar fingerprint in the second panel,AndThe training fingerprints in (a) constitute edges of the second panel graph.
3) For on the panel fingerprintThe corresponding position coefficients are determined according to the similarity and expressed as Then, calculate using the softmax functionIs added to the fingerprint coefficients of the fingerprint database so that the cumulative amount is equal to 1. Integrating the results of the two panels to obtainIs represented by the optimal predicted position of (a):
2. calculating the optimal prediction position by adopting an adaptive Bayes comprehensive learning particle swarm optimization algorithm
1) First, a particle swarm (ps particles) is initialized, and equation (6) is used as a fitness function, and each particle position vector can be represented as a particle position vector X n＝(xn,yn,hn with d=3 dimensions, n=1, 2, 3. An initial fitness value for each particle is calculated and a level of integrated learning (CL) probability is randomly assigned to each particle.
2) Dividing posterior probability intervals and adaptively determining comprehensive learning probability: in the iterative calculation, the particle posterior probability p t of the t-th iteration is calculated by using the Bayesian theorem, p t is divided into L intervals, and the subintervals are defined as:
Wherein the method comprises the steps of The L i th subinterval for the t-th iteration; And The maximum and minimum of the posterior probability vector, respectively. Δp t is the deviation interval of the posterior probability; And The lower and upper bounds of the subinterval, respectively.Is the CL probability for the eta subinterval.
The posterior probabilities of all particles are divided into subintervals S I according to equation (7), and the CL probability of each interval is the median value of the interval. When the CL probability of a particle needs to be updated, the self-adaptive mechanism adopts interval division to select the CL probability level of the particle, thereby realizing self-adaptation of comprehensive learning CL probability and determining a particle learning sample.
3) And finally, updating the speed, the position, the adaptive value and the like of the particles to finish iteration, and stopping iteration if the iteration times T reach the set maximum iteration times T. Taking the final optimal position of the particle swarm as an optimal solution of the formula (6), and taking the corresponding (x, y) as an optimal prediction position for positioning.
Drawings
FIG. 1 is a schematic diagram of a positioning model and algorithm of the present invention;
FIG. 2 is a schematic diagram of a testing environment according to an embodiment of the present invention;
Detailed Description
The present invention will be further described with reference to the following examples and figures, wherein the examples are provided for illustration only and not for the purpose of limiting the invention in any way. All other embodiments, which can be made by those skilled in the art based on the embodiments of the invention without making any inventive effort, are intended to be within the scope of the invention.
Examples:
The WLAN fingerprint positioning method based on the self-adaptive Bayes comprehensive learning particle swarm optimization comprises the steps of carrying out fingerprint similarity measurement by adopting a double-sided board uniformity model, and carrying out positioning parameter optimization and position prediction by adopting a self-adaptive Bayes comprehensive learning particle swarm optimization algorithm, wherein the method comprises the following specific steps:
it is assumed that there are M signal access points AP in the indoor environment and N physical locations are uniformly selected as reference points RP. Is provided with AndRepresenting sets of APs and RPs, respectively. Is provided withRepresenting the position coordinates of the i (i=1, 2, …, N) th RP, and s ij represents the RSS of the j-th AP at that position. Defining the fingerprint (RSS vector) corresponding to the ith RP position asIn summary, the training fingerprint set of reference points RP in the positioning environment can be expressed as:
Is provided with Representing RSS vectors collected by the mobile device during the online phase for mobile device positioning. If a plurality of users perform positioning inquiry, the fingerprint set is expressed as:
Where Γ is the number of query fingerprints. If the jth AP cannot be detected in the positioning, a very small value is assigned to s ij.
To verify the effectiveness and robustness of the algorithm, we performed experiments in a real indoor scenario. In this environment, m=11 APs are deployed, n=53 reference points are selected in the whole area, RSS data are collected and processed into 53 groups of fingerprint vectors to form a training fingerprint set in an offline stage, and in addition, RSS data are collected at Γ=136 different positions and processed into 136 groups of fingerprint vectors to form an online stage query (test) fingerprint set, as shown by the open square dots and solid dots in fig. 2.
1. Fingerprint similarity metric
The similarity of different fingerprints is measured by adopting a double-panel fingerprint uniformity graph representation method.
1) For the first panel, the correlation is used to evaluate the similarity of different fingerprints. For fingerprints Calculating the correlation between the two:
cor(u,v)＝1-r(u,v) (19)
Wherein, AndRespectively isIs defined as the mean value and standard deviation of (c),AndRespectively isMean and standard deviation of (c).
Selecting h AND gates according to equation (1)The fingerprint with the highest correlation is expressed as:
Wherein, Represents the q-th similar fingerprint and,And (3) withThe training fingerprints in (a) constitute the edges of the first panel graph.
2) For the second panel, for the fingerprintThe cosine distance between the two is calculated (reflecting the divergence of the fingerprint vector in terms of directionality):
similarly, h of the second panels are The most similar fingerprint of (2) is expressed as:
Wherein, Representing the q' th similar fingerprint in the second panel,AndThe training fingerprints in (a) constitute edges of the second panel graph.
3) For on the panel fingerprintThe corresponding position coefficients are determined according to the similarity and expressed as Then, calculate using the softmax functionIs added to the fingerprint coefficients of the fingerprint database so that the cumulative amount is equal to 1. Integrating the results of the two panels to obtainIs represented by the optimal predicted position of (a):
2. calculating the optimal prediction position by adopting an adaptive Bayes comprehensive learning particle swarm optimization algorithm
1) First, a particle swarm (ps particles) is initialized, and equation (6) is used as a fitness function, and each particle position vector can be represented as a particle position vector X n＝(xn,yn,hn with d=3 dimensions, n=1, 2, 3. An initial fitness value for each particle is calculated and a level of integrated learning (CL) probability is randomly assigned to each particle.
2) Dividing posterior probability intervals and adaptively determining comprehensive learning probability: in the iterative calculation, the particle posterior probability p t of the t-th iteration is calculated by using the Bayesian theorem, p t is divided into L intervals, and the subintervals are defined as:
Wherein the method comprises the steps of The L i th subinterval for the t-th iteration; And The maximum and minimum of the posterior probability vector, respectively. Δp t is the deviation interval of the posterior probability; And The lower and upper bounds of the subinterval, respectively.Is the CL probability for the eta subinterval.
In particular, L may be 3 in implementation to reduce computational complexity.
The posterior probabilities of all particles are divided into subintervals S I according to equation (7), and the CL probability of each interval is the median value of the interval. When the CL probability of a particle needs to be updated, the self-adaptive mechanism adopts interval division to select the CL probability level of the particle, thereby realizing self-adaptation of comprehensive learning CL probability and determining a particle learning sample.
3) And finally, updating the speed, the position, the adaptive value and the like of the particles to finish iteration, and stopping iteration if the iteration times T reach the set maximum iteration times T. Taking the final optimal position of the particle swarm as an optimal solution of the formula (6), and taking the corresponding (x, y) as an optimal prediction position for positioning.
Claims (1)
1. The WLAN fingerprint positioning method based on the self-adaptive Bayes comprehensive learning particle swarm optimization is characterized by comprising the steps of establishing a similarity between fingerprints calculated by a double-panel uniformity measurement model and calculating an optimal prediction position by a self-adaptive Bayes comprehensive learning particle swarm optimization algorithm, respectively carrying out fingerprint similarity measurement on the similarity part between fingerprints calculated by the double-panel uniformity measurement model by adopting correlation and cosine distance, and combining the correlation and cosine distance with the similarity part for calculating the optimal prediction position, wherein the method comprises the following specific steps of:
Assuming that M signal Access Points (AP) exist in an indoor environment, and uniformly selecting N physical positions as Reference Points (RP); let a= { AP 1,AP2,…APM } and r= { RP 1,RP2,…RPN } represent sets of AP and RP, respectively; is provided with Representing the position coordinates of the ith RP, i=1, 2, …, N, s ij representing the RSS of the jth AP at that position; defining the corresponding fingerprint of the ith RP position asThe training fingerprint set of reference points RP in a positioning environment can be expressed as:
Let S query＝[squery1,squery2,…,squeryM]T denote RSS vectors collected by the mobile device during the online phase for mobile device positioning; if a plurality of users perform positioning inquiry, the fingerprint set is expressed as:
Where Γ is the number of query fingerprints; if the jth AP cannot be detected in the positioning, assigning a very small value to s ij;
1) For a first panel, using the correlation to evaluate similarity of different fingerprints; for fingerprints Calculating the correlation between the two:
cor(u,v)＝1-r(u,v) (1)
Wherein, AndRespectively isIs defined as the mean value and standard deviation of (c),AndRespectively isAverage and standard deviation of (a);
Selecting h AND gates according to equation (1) The fingerprint with the highest correlation is expressed as:
Wherein, Represents the q-th similar fingerprint and,And (3) withThe training fingerprints in (a) constitute the edges of the first panel graph;
2) For the second panel, for the fingerprint Calculating the cosine distance between the two:
similarly, h of the second panels are The most similar fingerprint of (2) is expressed as:
Wherein, Representing the q' th similar fingerprint in the second panel,AndThe training fingerprints in (a) constitute the edges of a second panel graph;
3) For on the panel fingerprint The corresponding position coefficients are determined according to the similarity and expressed asThen, calculate using the softmax functionThe cumulative amount of all fingerprint coefficients in (2) is equal to 1; integrating the results of the two panels to obtainIs represented by the optimal predicted position of (a):
；
in the improved Bayes comprehensive learning particle swarm optimization algorithm, the particle posterior probability is divided into intervals, the comprehensive learning CL probability is determined in a self-adaptive mode, and the calculation of the optimal prediction position is realized in an iterative mode, wherein the method comprises the following specific steps:
1) First initializing a population of particles of ps particles, using formula (6) as a fitness function, each particle position vector being representable as a particle position vector X n＝(xn,yn,hn having d=3 dimensions, n=1, 2,3,..ps; calculating an initial adaptation value of each particle, and randomly distributing a horizontal comprehensive learning CL probability to each particle;
2) Dividing posterior probability intervals and self-adaptively determining comprehensive learning CL probabilities: in the iterative calculation, the particle posterior probability p t of the t-th iteration is calculated by using the Bayesian theorem, p t is divided into L intervals, and the subintervals are defined as:
Wherein the method comprises the steps of The L i th subinterval for the t-th iteration; And The maximum value and the minimum value of the posterior probability vector are respectively; Δp t is the deviation interval of the posterior probability; And Respectively a lower bound and an upper bound of the subinterval; the comprehensive learning CL probability of the eta subinterval;
The posterior probabilities of all particles are divided into subintervals S I according to equation (7), and the comprehensive learning CL probability of each interval is the median value of the interval; when the comprehensive learning CL probability of a particle needs to be updated, the self-adaptive mechanism adopts interval division to select the comprehensive learning CL probability level of the particle, so as to realize self-adaptation of the comprehensive learning CL probability and determine a particle learning sample;
3) Finally updating the speed, position, adaptation value and the like of the particles to finish iteration, and stopping iteration if the iteration times T reach the set maximum iteration times T; taking the final optimal position of the particle swarm as an optimal solution of the formula (6), and taking the corresponding (x, y) as an optimal prediction position for positioning.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210434881.2A CN114928882B (en) | 2022-04-24 | WLAN fingerprint positioning method based on self-adaptive Bayes comprehensive learning particle swarm optimization |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210434881.2A CN114928882B (en) | 2022-04-24 | WLAN fingerprint positioning method based on self-adaptive Bayes comprehensive learning particle swarm optimization |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN114928882A CN114928882A (en) | 2022-08-19 |
| CN114928882B true CN114928882B (en) | 2024-06-28 |
Family
ID=
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109298389A (en) * | 2018-08-29 | 2019-02-01 | 东南大学 | Indoor pedestrian based on multiparticle group optimization combines position and orientation estimation method |
| CN112887902A (en) * | 2021-01-22 | 2021-06-01 | 湖南大学 | Indoor positioning method of WiFi fingerprint based on Gaussian clustering and hybrid measurement |
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109298389A (en) * | 2018-08-29 | 2019-02-01 | 东南大学 | Indoor pedestrian based on multiparticle group optimization combines position and orientation estimation method |
| CN112887902A (en) * | 2021-01-22 | 2021-06-01 | 湖南大学 | Indoor positioning method of WiFi fingerprint based on Gaussian clustering and hybrid measurement |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Ibrahim et al. | CNN based indoor localization using RSS time-series | |
| CN109298389B (en) | Indoor pedestrian combination pose estimation method based on multi-particle swarm optimization | |
| Saucan et al. | A multisensor multi-Bernoulli filter | |
| Bilik et al. | GMM-based target classification for ground surveillance Doppler radar | |
| Hero et al. | Foundations and applications of sensor management | |
| Li et al. | SmartLoc: Smart wireless indoor localization empowered by machine learning | |
| EP2981136B1 (en) | Terminal device, position estimation method, position estimation system, and program | |
| CN112887902B (en) | Indoor positioning method of WiFi fingerprint based on Gaussian clustering and hybrid measurement | |
| US7747084B2 (en) | Methods and apparatus for target discrimination using observation vector weighting | |
| Ji et al. | Nonmyopic multiaspect sensing with partially observable Markov decision processes | |
| Kwak et al. | Magnetic field based indoor localization system: A crowdsourcing approach | |
| CN105913078A (en) | Multi-mode soft measurement method for improving adaptive affine propagation clustering | |
| Caso et al. | Frequentist inference for WiFi fingerprinting 3D indoor positioning | |
| Dou et al. | Top-down indoor localization with Wi-fi fingerprints using deep Q-network | |
| Ma et al. | A novel indoor fingerprint localization system based on distance metric learning and AP selection | |
| CN109121081B (en) | Indoor positioning method based on position candidate set and EM algorithm | |
| Tu et al. | A theoretical investigation of several model selection criteria for dimensionality reduction | |
| Yan et al. | A noise reduction fingerprint feature for indoor localization | |
| CN114928882B (en) | WLAN fingerprint positioning method based on self-adaptive Bayes comprehensive learning particle swarm optimization | |
| Rosic et al. | An improved chaos driven hybrid differential evolution and butterfly optimization algorithm for passive target localization using TDOA measurements | |
| Maheswararajah et al. | Sensor scheduling for target tracking by suboptimal algorithms | |
| Denzler et al. | Optimal selection of camera parameters for state estimation of static systems: An information theoretic approach | |
| CN114928882A (en) | WLAN fingerprint positioning method based on adaptive Bayes comprehensive learning particle swarm optimization | |
| Yong Kang et al. | A robust indoor mobile localization algorithm for wireless sensor network in mixed LOS/NLOS environments | |
| Kim et al. | An efficient scheme of target classification and information fusion in wireless sensor networks |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant |