CN109598320A - A kind of RFID indoor orientation method based on locust algorithm and extreme learning machine - Google Patents
A kind of RFID indoor orientation method based on locust algorithm and extreme learning machine Download PDFInfo
- Publication number
- CN109598320A CN109598320A CN201910046012.0A CN201910046012A CN109598320A CN 109598320 A CN109598320 A CN 109598320A CN 201910046012 A CN201910046012 A CN 201910046012A CN 109598320 A CN109598320 A CN 109598320A
- Authority
- CN
- China
- Prior art keywords
- locust
- value
- rssi
- learning machine
- extreme 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.)
- Pending
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06K—GRAPHICAL DATA READING; PRESENTATION OF DATA; RECORD CARRIERS; HANDLING RECORD CARRIERS
- G06K17/00—Methods or arrangements for effecting co-operative working between equipments covered by two or more of main groups G06K1/00 - G06K15/00, e.g. automatic card files incorporating conveying and reading operations
- G06K17/0022—Methods or arrangements for effecting co-operative working between equipments covered by two or more of main groups G06K1/00 - G06K15/00, e.g. automatic card files incorporating conveying and reading operations arrangements or provisious for transferring data to distant stations, e.g. from a sensing device
- G06K17/0029—Methods or arrangements for effecting co-operative working between equipments covered by two or more of main groups G06K1/00 - G06K15/00, e.g. automatic card files incorporating conveying and reading operations arrangements or provisious for transferring data to distant stations, e.g. from a sensing device the arrangement being specially adapted for wireless interrogation of grouped or bundled articles tagged with wireless record carriers
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/004—Artificial life, i.e. computing arrangements simulating life
- G06N3/006—Artificial life, i.e. computing arrangements simulating life based on simulated virtual individual or collective life forms, e.g. social simulations or particle swarm optimisation [PSO]
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Biophysics (AREA)
- Data Mining & Analysis (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- Biomedical Technology (AREA)
- Computer Networks & Wireless Communication (AREA)
- Computational Linguistics (AREA)
- Health & Medical Sciences (AREA)
- Evolutionary Computation (AREA)
- General Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Radar Systems Or Details Thereof (AREA)
Abstract
The present invention relates to a kind of RFID indoor orientation method based on locust algorithm and extreme learning machine, this method does parameter optimization with hidden layer connection weight and hidden layer neuron threshold value to the input layer in extreme learning machine using locust algorithm, the performance of extreme learning machine is improved, reference label building extreme learning machine location model is recycled to be positioned.The present invention includes following two stage, off-line phase: the acquisition of indoor positioning area reference label data, the foundation of original RSSI data prediction, tranining database, the building of location model parameter optimization, location model;On-line stage: target labels signal acquisition, data processing prediction, positioning coordinate output.Compared with prior art, the advantages of present invention has positioning accuracy high, and environment resistant changing capability is strong, at low cost, performance efficiency.
Description
Technical field
The present invention relates to less radio-frequency field of locating technology, specifically a kind of to be based on locust algorithm and extreme learning machine
RFID indoor orientation method.
Background technique
Internet of Things is that its user terminal is extended to and expanded to any article on the basis of internet, carries out information exchange
With a kind of network of communication.The key technology of Internet of Things, radio frequency identification (Radio Frequency Identification,
RFID) technology is a kind of automatic identification technology that non contact angle measurement is carried out by radiofrequency signal, can be carried out to identified object
Identification.And the indoor positioning in its field is due to navigating indoors, the advantage of the positioning of personnel's cargo and emergency management and rescue etc., by
The concern of each side has been arrived, and has showed vast market prospect and extensive products application.
Common location technology includes GPS, infrared, WiFi, bluetooth etc..GPS(Global Positioning System)
That is global positioning system is the satellite navigation and location system established by the U.S., and using the system, user can be in global model
It encloses interior round-the-clock, continuous, the real-time three-dimensional navigation positioning for realizing open air and tests the speed;In addition, using the system, user can also
Carry out high-precision Time Transmission and high-precision precision positioning.But indoors, due to obstacle under the complex scenes such as house wall
Obstruction of the object for electromagnetic wave weakens and absorbs, it is limited to lead to GPS location precision, and GPS power consumption is huge, sometimes unfavorable
It is used in people.And infrared ray, the technologies such as WiFi for RFID radio-frequency technique, RFID position indoors in possessed by
Positioning accuracy is high, and strong antijamming capability is low in cost, and the series of advantages such as easy to use make it have more big advantage
And development potentiality.
By literature search, we retrieved following pertinent literature, and indoor positioning algorithms used by these documents can
To realize the indoor positioning of certain precision, but all without using locust algorithm-extreme learning machine method, such as:
Chinese patent CN201710735262.6, a kind of indoor orientation method based on RFID, apparatus and system, patent right
People: Guangdong University of Technology.Which disclose a kind of indoor orientation methods based on RFID, apparatus and system, control RFID days
Line is rotated with predetermined angle, is scanned to label, is obtained to the corresponding signal strength indication of rotation angle.According to rotation angle
Degree, signal strength indication and rotable antenna location algorithm are positioned.The inventive embodiments use only a RFID antenna, have
System structure is simple, at low cost, the deployment advantage that difficulty is low and hardware utilization is high.But it copes with complex environment, frequently results in
Positioning accuracy is not high, is easy to receive the interference of barrier.And a RFID antenna is only used only in embodiment, and it is fixed if desired to increase
The accuracy of position, then corresponding cost will increase again.
Chinese patent CN201710050791.2, a kind of indoor positioning device based on BP-Landmarc neural network and
Control method, patentee: Jilin University.Which disclose a kind of indoor positioning devices based on BP and Landmarc algorithm
And method, RFID reader read the RSSI value of RFID electronic label, construct BP-Landmarc artificial neural network, pass through
Study to reference label RSSI data obtains the weight interconnected between each layer, to establish between reference label and position
Specific non-linear mapping model realizes the mesh for exporting corresponding dynamic labels motion profile or static object location map
Mark.But it copes with environment complicated and changeable, frequently results in that positioning accuracy is not high, and robustness is not strong.
Summary of the invention
A kind of indoor orientation method technical solution based on locust algorithm and extreme learning machine of the invention is as follows: Yi Zhongji
In the indoor orientation method of locust algorithm and extreme learning machine, comprising:
RFID label tag is distributed indoors, for emitting original RSSI signal strength Value Data;
Multiple reader antennas and a reader terminal, distribution indoors, for read FRID label information and
RSSI value;
WiFi radio receiving transmitting module: for receiving and transmitting the data of reader;
PC host computer: for reading label information, the training of locust algorithm-extreme learning machine location model and output RFID
Label position.
A kind of RFID indoor orientation method based on locust algorithm and extreme learning machine of the invention, comprising the following steps:
The acquisition of step 1) off-line phase reference label signal strength indication RSSI data: it disposes and reads in localization region
The RSSI signal strength that device, reader antenna and reference label, record reference label position and different location reader receive
Value obtains original training data;
The pretreatment of step 2) off-line phase initial data: RSSI signal strength indication accessed by reader is located in advance
Reason, the exceptional value caused by removing because of various factors;And the initial data after pretreatment is integrated into training data;
The building of step 3) off-line phase location model: using the training data integrated as input, pass through locust algorithm
The hidden layer neuron weight and threshold value of extreme learning machine are optimized, locust algorithm-extreme learning machine location model is constructed;
The acquisition of step 4) on-line stage RFID target labels information: when the target to be positioned for carrying RFID label tag enters
When detection zone, reader obtains label information and RSSI value, is transmitted to PC host computer by WiFi radio receiving transmitting module, removes
PC host computer handles the information received and constructs the real time information data library about this label after exceptional value;
It accurately predicts step 5) on-line stage position: the data for the target labels that on-line stage obtains being input to and have been instructed
In the locust algorithm-extreme learning machine location model perfected, the specific location coordinate of label to be measured is exported.
As a further improvement of the present invention, the pretreated specific method of off-line phase initial data in the step 2)
Are as follows: for off-line phase first according to the actual environment situation of localization region, then reasonable layout RFID reference label disposes N number of read
It reads device antenna and is distributed in detection zone surrounding, the signal strength indication RSSI and corresponding seat of each label are collected by antenna
Mark;The signal strength that i-th of reader repeats to read same reference label is J times total, and the signal strength that kth time is read is remembered
Record is RSSIik, i=1,2 ..., N;It calculates J times and measures the average intensity value R_avg, standard deviation R_std got:
The absolute value of the difference of measured value and average value is removed greater than the sample of standard deviation, that is, assumes that the signal of some sample is strong
Angle value is RSSIikIf | RSSIik- R_avg | > R_std then deletes corresponding sample;One has finally been obtained after the completion of processing
The RSSI that size is m gathers, and the RSSI mean value calculation gathered is denoted as average signal strengthFor in localization region
With N number of reader, M reference label carries out above-mentioned pretreatment, can obtain the RSSI training sample set that a size is M × N
It is denoted as RSSIMN:
The position mark collection PI for the reference label that a corresponding size is M × 2 is obtained simultaneously, wherein including with reference to mark
The coordinate information (x, y) of label:
In order to disclose the present invention abudantly, the specific steps of the building of step 3) the off-line phase location model are as follows:
A) design food source in solution space: in all solution spaces, each food source includes following information:
[wI, j, bj]TI=1,2 ..., M j=1,2 ..., L (5)
The information of each food source generates at random, wI, jFor j-th of hidden layer neuron of connection and i-th of input layer
Weight, random number of the value between [- 1,1];bjFor j-th of hidden layer neuron threshold value, value between [0,1] with
Machine number;M and L is respectively input layer number (i.e. the number of reference label sample) and hidden layer neuron number;
B) food source fitness is designed: for the accuracy of response prediction result, using root-mean-square error:
X in above formulai, yiFor reference label actual position value, xi', yi' it is location prediction value;The lesser solution of root-mean-square error
Corresponding biggish fitness, therefore take the inverse of root-mean-square error as fitness value:
C) K food source is randomly generated in solution space, initializes Locust Swarm individual Xi, i=1,2 ..., K, random distribution
All food sources in solution space;T is setmaxFor maximum number of iterations (Tmax> 1), t is current iteration number;
D) according to the food source where each locust individual, obtain the food source information i.e. input layer and
The weight w of hidden layer neuronI, jWith hidden layer neuron threshold value bj;By the training sample set RSSI after pretreatmentMNFurther
R is obtained after linear normalization processing:
According to ri, wI, jAnd bj, then locust algorithm-extreme learning machine location model with L hidden layer neuron can table
It is shown as:
Wherein g (wI, j·ri+bj) it is output of i-th of reference label sample in j-th of neuron of hidden layer;wI, j·ri
For the inner product of vector;βiConnection weight between hidden layer neuron and output layer;By the position mark collection PI of reference label
P is obtained after being normalized:
In the neural network for being activation primitive with g (x), location prediction value tiIt can be with zero error close to the true position of label
Set pi, i.e.,
WhereinIt can state are as follows: H β=T, wherein H is hidden layer output matrix, and β is output weight,
T is desired output;H can be expressed as follows:
According to H β=T, β is solved;It is solved are as follows: β=H+T;Wherein H+For the inverse of hidden layer output matrix;According to what is solved
β, anti-normalization processing tiIt can be obtained by location prediction value xi', yi′;K food source is calculated separately according to formula (6), formula (7)
Fitness, therefrom selection obtains the current optimal adaptation degree food source in K food source
E) location updating is carried out according to food source to Locust Swarm individual:
Wherein dij=| xj-xi| indicate i-th locust at a distance from jth locust;Indicate i-th locust to
The unit vector of j locust;S function defines influence function of the locust by population active force, and expression formula is as follows:
Wherein r is the distance between population at individual;In order to preferably obtain globally optimal solution, make entire convergence in population to only
One food source, is added adaptation coefficient in location update formulaIt balances entire population in solution space for food
Material resource searches calculation ability;Therefore, the individual location updating function of locust can be replaced by:
Wherein ud, ldIndicate the upper bound and the lower bound of kth locust d dimension variable;For the mobile target food of Locust Swarm
Source, that is, the last optimal adaptation degree food source iterated to calculate out;Each locust individual behind update position is mutual
Distance r is limited in [Isosorbide-5-Nitrae];And the fitness of food source where locust individual behind update position is recalculated, it updatesIteration
Number t=t+1;
F) check whether current iteration number t is greater than maximum number of iterations Tmax;If not up to, repeating step d)-e)
Until maximum number of iterations;If the number of iterations reaches setting Tmax, from optimal adaptation degree food sourceExtract extreme learning machine
Required connection weight wI, jWith threshold value bj, complete the building of location model.
The beneficial effects of the present invention are:
1, the pretreatment of label RSSI value vector eliminates the exceptional value occurred in measurement process, effectively prevents exception
Influence of the data for positioning accuracy;Establish the location model of high quality, effective solution algorithm is by abnormal environment situation
Caused by the low problem of accuracy.
2, the extreme learning machine of locust algorithm optimization, the basic principle is that being implied using locust algorithm to extreme learning machine
Layer weight ω and threshold value b is optimized, and the RSSI value vector for recycling reader to receive is any with reference to mark as training sample
Label have a RSSI value vector RSSII, j, after removing exceptional value, with RSSII, j(x, y) value corresponding with it is as feature and mark
Note, training locust algorithm-extreme learning machine location model.When positioning, target to be detected enters area to be tested, can also obtain one
A RSSI value vector RSSII, j, as mode input, the accurate coordinates of target to be positioned will be obtained.Experimental data table
Bright, for this method compared to traditional algorithm, generalization ability is more preferable, and accuracy is higher, and can reduce under the premise of improving precision
Positioning system use cost, it is effective to reduce positioning time, overcome bring due to multipath effect, environmental change occurs for signal fixed
The low problem of position precision.
Detailed description of the invention
A kind of overall framework figure of the RFID indoor orientation method based on locust algorithm and extreme learning machine of Fig. 1 present invention;
A kind of system structure of the RFID indoor orientation method based on locust algorithm and extreme learning machine of Fig. 2 present invention is shown
It is intended to;
A kind of example schematic of the RFID indoor orientation method based on locust algorithm and extreme learning machine of Fig. 3 present invention;
A kind of algorithm flow chart of the RFID indoor orientation method based on locust algorithm and extreme learning machine of Fig. 4 present invention.
Specific embodiment
Below in conjunction with Fig. 1~4 and a kind of room RFID based on locust algorithm and extreme learning machine of the embodiment description present invention
Interior localization method.
Embodiment:
Fig. 1 is a kind of overall framework of the RFID indoor orientation method based on locust algorithm and extreme learning machine of the present invention
Figure, relates generally to two stages: off-line phase and on-line stage.Off-line phase is by RFID reference label cloth according to certain rules
Set in localization region, by RFID antenna and RFID reader terminal receive each label signal strength indication RSSI and specific position
Coordinate is set, to obtain the original training data collection needed for locust algorithm-extreme learning machine location model containing exceptional value, PC
After host computer receives initial data, remove exceptional value, using locust algorithm to the hidden layer weight ω of extreme learning machine and
Threshold value b is optimized and is constructed locust algorithm-extreme learning machine location model.Target labels are carried along into detection by on-line stage
Region, reader receive the information (RSSI value) of target labels, and information is transferred on PC by WiFi radio receiving transmitting module
Position machine terminal, removes exceptional value again, is input to trained locust algorithm-extreme learning machine model, carries out online pre-
It surveys, output result is exactly the specific location coordinate of label to be measured.
Fig. 2 is that a kind of system structure of the RFID indoor orientation method based on locust algorithm and extreme learning machine of the present invention is shown
It is intended to, including eight reader antennas 1,2,3,4,5,6,7,8.Reader antenna off-line phase obtains localization region reference label
The acquisition of information data acquisition and on-line stage object to be measured label information.RFID reader passes through wired by 8 antennas and its
It is connected, then gives the data transmission received to PC host computer by WiFi radio receiving transmitting module, PC host computer is responsible for receiving transmission
Data, send control command and output target information at building locust algorithm-extreme learning machine location model.
Fig. 3 is a kind of example signal of RFID indoor orientation method based on locust algorithm and extreme learning machine of the present invention
Figure, is placed in the surrounding of detection zone including 8 reader antennas, reference label by certain rule deployment, label number according to
It is adjacent to be separated by greater than 0.5m depending on the size of region area to be measured.Object to be positioned carries RFID label tag and enters detection zone, is
The label information (RSSI signal strength indication) carried according to object of uniting positions it.
Fig. 4 is a kind of algorithm flow of the RFID indoor orientation method based on locust algorithm and extreme learning machine of the present invention
Figure, the specific steps are as follows:
Step 1. off-line phase localization region data collection rationally divides first according to the actual environment situation of localization region
Then cloth RFID reference label disposes 8 reader antennas and is distributed in detection zone surrounding, collects each label by antenna
Signal strength indication RSSI and corresponding coordinate, to obtain the original trained number of locust algorithm-extreme learning machine location model
According to collection.For some reference label of deployment, it is assumed that marked as j, (wherein j=1,2,3 ..., 100), measure K RSSI to label
It is worth vector, obtains K sample data, the data of 8 rows, K+2 column are constituted with corresponding (x, y), structure is as follows:
Wherein, 8 readers measure 5 RSSI value vectors to 1 reference label:
Step 2. off-line phase initial data is pretreated method particularly includes: calculates 5 measurements of each reader acquisition
The average intensity value R_avg arrivedi, standard deviation R_stdi, whereinRemove the exhausted of the difference of measured value and average value
It is greater than the sample of standard deviation to value, that is, assumes that the signal strength indication of some sample is RSSIikIf | RSSIik-R_avgi| > R_
stdi, then corresponding sample is deleted;The RSSI that a size is m has finally been obtained after the completion of processing to gather, and RSSI has been gathered
Mean value calculation be denoted as average signal strengthFor the first row RSSI value data of above-mentioned set R, it reflects first
The result that a reader reads 1 label;Its R_avgi=1.36, R_stdi=0.305, according to the above method, processing
Result afterwards is deleted 1.9 as exceptional value.The each column of above-mentioned R is all done into such place
It manages back and obtains 8 readersFor in localization region have 8 readers, 100
A reference label carries out above-mentioned pretreatment, can obtain the RSSI training sample set that a size is 100 × 8 and be denoted as RSSI100×8:
The coordinate (x, y) of each label is extracted to the position for independently forming the reference label of 100 × 2 sizes
Label sets PI, wherein including the location information of reference label:
Step 3. off-line phase is constructed based on locust algorithm-extreme learning machine location model, the number obtained using preceding 2 pacing
According to as training data, wherein RSSI value vector RSSI100×8As characteristic data set, what corresponding position coordinates (x, y) were formed
Position mark collection of the set PI as sample.By locust algorithm to the hidden layer neuron weight of location model extreme learning machine
It is optimized with threshold value.It completes the training data integrated as input for locust algorithm-extreme learning machine location model
Building.
Step 4. on-line stage live signal obtains, when the object for carrying RFID label tag enters localization region, reader
Antenna repeatedly obtains the RSSI value vector of label, and it is upper by reader terminal and WiFi radio receiving transmitting module to be transferred to PC
Machine, PC host computer construct the information bank of itself label according to data obtained.
Step 5. on-line stage data prediction, the RSSI value vector T est for the label to be detected that detection is 5 times:
Exceptional value is removed according to step 2 method, it is assumed that is left 3 samples, is taken the average RSSI value vector of 3 measurements:
R_test=(R1,1, R1,2..., R1,3)/3
R after the R_test of 8 readers is carried out integration normalized is as locust algorithm-extreme learning machine positioning
The input of model.
Step 6. on-line stage real time position is accurately positioned, and the R of treated in step 5 high quality is input to and is trained
Locust algorithm-extreme learning machine location model, carry out on-line prediction, the output result of model is exactly the specific position of label to be measured
Set coordinate.
It should be noted last that the above examples are only used to illustrate the technical scheme of the present invention and are not limiting.Although ginseng
It is described the invention in detail according to embodiment, those skilled in the art should understand that, to technical side of the invention
Case is modified or replaced equivalently, and without departure from the spirit and scope of technical solution of the present invention, should all be covered in the present invention
Claim in.The present invention be by multidigit RFID indoor positioning algorithms technical staff's Long-Term Scientific Study experience accumulation, and
Gone out by creative work creation, the RSSI value vector of reference label is detected and collected as characteristic, phase
Check bit of the corresponding position coordinates (x, y) as sample, by locust algorithm to the hidden layer of location model extreme learning machine
Neuron weight and threshold value optimize, and training locust algorithm-extreme learning machine model, then PC host computer inputs target labels
RSSI value carry out on-line prediction, obtain specific location coordinate.The present invention solves that indoor positioning positioning accuracy is low and algorithm pair
The problems such as ambient noise is sensitive, and algorithm model is simple, positioning time is fast, and system cost is lower, has very strong practical value.
Claims (3)
1. a kind of RFID indoor orientation method based on locust algorithm and extreme learning machine, which comprises the following steps:
The acquisition of step 1) off-line phase reference label signal strength indication RSSI data: reader is disposed in localization region, is read
Device antenna and reference label are read, the RSSI signal strength indication that record reference label position and different location reader receive obtains
Obtain original training data;
The pretreatment of step 2) off-line phase initial data: pre-processing RSSI signal strength indication accessed by reader,
Exceptional value caused by removing because of various factors;And the initial data after pretreatment is integrated into training data;
The building of step 3) off-line phase location model: using the training data integrated as input, by locust algorithm to pole
The hidden layer neuron weight and threshold value for limiting learning machine optimize, and construct locust algorithm-extreme learning machine location model;
The acquisition of step 4) on-line stage RFID target labels information: when the target to be positioned for carrying RFID label tag enters detection
When region, reader obtains label information and RSSI value, is transmitted to PC host computer by WiFi radio receiving transmitting module, removal is abnormal
PC host computer handles the information received and constructs the real time information data library about this label after value;
It accurately predicts step 5) on-line stage position: the data for the target labels that on-line stage obtains being input to and have been trained
Locust algorithm-extreme learning machine location model in, export the specific location coordinate of label to be measured.
2. a kind of indoor orientation method based on locust algorithm and extreme learning machine according to claim 1, feature exist
In the step 2 off-line phase initial data is pretreated method particularly includes: off-line phase is first according to the reality of localization region
Then ambient conditions, reasonable layout RFID reference label dispose N number of reader antenna and are distributed in detection zone surrounding, pass through day
Line collects the signal strength indication RSSI and corresponding coordinate of each label;I-th of reader repeats to read same reference label
Signal strength it is J times total, and the signal strength that kth time is read is recorded as RSSIik, i=1,2 ..., N;Calculate J measurement
The average intensity value R_avg and standard deviation R_std got:
The absolute value of the difference of measured value and average value is removed greater than the sample of standard deviation, that is, assumes the signal strength indication of some sample
For RSSIikIf | RSSIik- R_avg | > R_std then deletes corresponding sample;A size has finally been obtained after the completion of processing
Gather for the RSSI of m, and the RSSI average value gathered is denoted as average signal strengthFor having N number of read in localization region
Device is read, M reference label carries out above-mentioned pretreatment, can obtain the RSSI training sample set that a size is M × N, be denoted as
RSSIMN:
The position mark collection PI for obtaining the reference label that a corresponding size is M × 2 simultaneously, wherein including reference label
Coordinate information (x, y):
3. a kind of indoor orientation method based on locust algorithm and extreme learning machine according to claim 1, feature exist
In, the step 3 off-line phase location model building comprising the following specific steps
A) design food source in solution space: in all solution spaces, each food source includes following information:
[wI, j, bj]TI=1,2 ..., M j=1,2 ..., L;
The information of each food source generates at random, wI, jFor the power for connecting j-th of hidden layer neuron and i-th of input layer
Value, random number of the value between [- 1,1];bjFor j-th of hidden layer neuron threshold value, value is random between [0,1]
Number;M and L is respectively input layer number (i.e. the number of reference label sample) and hidden layer neuron number;
B) food source fitness is designed: for the accuracy of response prediction result, using root-mean-square error:
Wherein xi, yiFor reference label actual position value, xi', yi' it is location prediction value;The lesser solution of root-mean-square error it is corresponding compared with
Big fitness, therefore take the inverse of root-mean-square error as fitness value:
C) K food source is randomly generated in solution space, initializes Locust Swarm individual Xi, i=1,2 ..., K are randomly dispersed in solution
All food sources in space;T is setmaxFor maximum number of iterations (Tmax> 1), t is current iteration number;
D) according to the food source where each locust individual, the information i.e. input layer and implicit of the food source are obtained
The weight w of layer neuronI, jWith hidden layer neuron threshold value bj;By the training sample set RSSI after pretreatmentMNIt is further linear
R is obtained after normalized:
According to ri, wI, jAnd bj, then locust algorithm-extreme learning machine location model with L hidden layer neuron may be expressed as:
Wherein g (wI, j·ri+bj) it is output of i-th of reference label sample in j-th of neuron of hidden layer;wI, j·riFor to
The inner product of amount;βiConnection weight between hidden layer neuron and output layer;The position mark collection PI of reference label is carried out
P is obtained after normalized:
In the neural network for being activation primitive with g (x), location prediction value tiIt can be with zero error close to label actual position pi,
I.e.
WhereinIt can state are as follows: H β=T, wherein H is hidden layer output matrix, and β is output weight, and T is scheduled to last
Hope output;H can be expressed as follows:
According to H β=T, β is solved;It is solved are as follows: β=H+T;Wherein H+For the inverse of hidden layer output matrix;According to the β solved, instead
Normalized tiIt can be obtained by location prediction value xi', yi′;According to reference label actual position value xi, yiAnd location prediction
Value xi', yi' fitness of K food source is calculated separately, therefrom selection obtains the current optimal adaptation degree food in K food source
Source
E) location updating is carried out according to food source to Locust Swarm individual:
Wherein dij=| xj-xi| indicate i-th locust at a distance from jth locust;Indicate i-th locust to jth only
The unit vector of locust;S function defines influence function of the locust by population active force, and expression formula is as follows:
Wherein r is the distance between population at individual;In order to preferably obtain globally optimal solution, make entire convergence in population to unique food
Adaptation coefficient is added in material resource in location update formulaIt balances entire population in solution space for food source
Search calculation ability;Therefore, the individual location updating function of locust can be replaced by:
Wherein ud, ldIndicate the upper bound and the lower bound of kth locust d dimension variable;For the mobile target food source of Locust Swarm,
The exactly last optimal adaptation degree food source iterated to calculate out;By the mutual distance r of each locust individual behind update position
It is limited in [Isosorbide-5-Nitrae];And the fitness of food source where locust individual behind update position is recalculated, it updatesThe number of iterations t
=t+1:
F) check whether current iteration number t is greater than maximum number of iterations Tmax;If not up to, repeating step d)-e) until
Maximum number of iterations;If the number of iterations reaches setting Tmax, from optimal adaptation degree food sourceIt extracts needed for extreme learning machine
Connection weight wI, jWith threshold value bj, complete the building of location model.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910046012.0A CN109598320A (en) | 2019-01-16 | 2019-01-16 | A kind of RFID indoor orientation method based on locust algorithm and extreme learning machine |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910046012.0A CN109598320A (en) | 2019-01-16 | 2019-01-16 | A kind of RFID indoor orientation method based on locust algorithm and extreme learning machine |
Publications (1)
Publication Number | Publication Date |
---|---|
CN109598320A true CN109598320A (en) | 2019-04-09 |
Family
ID=65966365
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201910046012.0A Pending CN109598320A (en) | 2019-01-16 | 2019-01-16 | A kind of RFID indoor orientation method based on locust algorithm and extreme learning machine |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN109598320A (en) |
Cited By (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110069866A (en) * | 2019-04-26 | 2019-07-30 | 福州大学 | A kind of design of pressure vessels method based on Swarm Intelligence Algorithm |
CN110072192A (en) * | 2019-04-26 | 2019-07-30 | 山东科技大学 | A kind of smart phone WiFi indoor orientation method |
CN110348631A (en) * | 2019-07-09 | 2019-10-18 | 武汉四创自动控制技术有限责任公司 | A kind of regionality Methods of electric load forecasting and system |
CN111199162A (en) * | 2020-01-11 | 2020-05-26 | 华南理工大学 | RFID reader fault self-adaptive positioning method |
CN111707895A (en) * | 2020-06-22 | 2020-09-25 | 合肥博雷电气有限公司 | Electromagnetic environment complexity evaluation method and system based on machine learning |
CN112040405A (en) * | 2020-09-15 | 2020-12-04 | 河南工业大学 | Indoor positioning method based on kernel limit learning machine and particle filtering |
Citations (8)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN106022465A (en) * | 2016-05-19 | 2016-10-12 | 江南大学 | Extreme learning machine method for improving artificial bee colony optimization |
CN106803769A (en) * | 2016-12-08 | 2017-06-06 | 浙江工业大学 | Passive-equipment-free human body orientation detection method based on WIFI |
CN107490783A (en) * | 2016-06-10 | 2017-12-19 | 天津力芯伟业科技有限公司 | A kind of RFID localization methods based on ELM |
CN107798383A (en) * | 2017-10-27 | 2018-03-13 | 天津大学 | Improved core extreme learning machine localization method |
US20180240018A1 (en) * | 2016-05-19 | 2018-08-23 | Jiangnan University | Improved extreme learning machine method based on artificial bee colony optimization |
CN108871332A (en) * | 2018-04-26 | 2018-11-23 | 广西大学 | A kind of RFID indoor locating system and method based on XGBoost |
CN109116300A (en) * | 2018-06-28 | 2019-01-01 | 江南大学 | A kind of limit learning position method based on non-abundant finger print information |
CN109195110A (en) * | 2018-08-23 | 2019-01-11 | 南京邮电大学 | Indoor orientation method based on hierarchical clustering technology and online extreme learning machine |
-
2019
- 2019-01-16 CN CN201910046012.0A patent/CN109598320A/en active Pending
Patent Citations (8)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN106022465A (en) * | 2016-05-19 | 2016-10-12 | 江南大学 | Extreme learning machine method for improving artificial bee colony optimization |
US20180240018A1 (en) * | 2016-05-19 | 2018-08-23 | Jiangnan University | Improved extreme learning machine method based on artificial bee colony optimization |
CN107490783A (en) * | 2016-06-10 | 2017-12-19 | 天津力芯伟业科技有限公司 | A kind of RFID localization methods based on ELM |
CN106803769A (en) * | 2016-12-08 | 2017-06-06 | 浙江工业大学 | Passive-equipment-free human body orientation detection method based on WIFI |
CN107798383A (en) * | 2017-10-27 | 2018-03-13 | 天津大学 | Improved core extreme learning machine localization method |
CN108871332A (en) * | 2018-04-26 | 2018-11-23 | 广西大学 | A kind of RFID indoor locating system and method based on XGBoost |
CN109116300A (en) * | 2018-06-28 | 2019-01-01 | 江南大学 | A kind of limit learning position method based on non-abundant finger print information |
CN109195110A (en) * | 2018-08-23 | 2019-01-11 | 南京邮电大学 | Indoor orientation method based on hierarchical clustering technology and online extreme learning machine |
Cited By (8)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110069866A (en) * | 2019-04-26 | 2019-07-30 | 福州大学 | A kind of design of pressure vessels method based on Swarm Intelligence Algorithm |
CN110072192A (en) * | 2019-04-26 | 2019-07-30 | 山东科技大学 | A kind of smart phone WiFi indoor orientation method |
CN110069866B (en) * | 2019-04-26 | 2022-06-14 | 福州大学 | Pressure container design method based on group intelligent algorithm |
CN110348631A (en) * | 2019-07-09 | 2019-10-18 | 武汉四创自动控制技术有限责任公司 | A kind of regionality Methods of electric load forecasting and system |
CN111199162A (en) * | 2020-01-11 | 2020-05-26 | 华南理工大学 | RFID reader fault self-adaptive positioning method |
CN111199162B (en) * | 2020-01-11 | 2021-10-26 | 华南理工大学 | RFID reader fault self-adaptive positioning method |
CN111707895A (en) * | 2020-06-22 | 2020-09-25 | 合肥博雷电气有限公司 | Electromagnetic environment complexity evaluation method and system based on machine learning |
CN112040405A (en) * | 2020-09-15 | 2020-12-04 | 河南工业大学 | Indoor positioning method based on kernel limit learning machine and particle filtering |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN109598320A (en) | A kind of RFID indoor orientation method based on locust algorithm and extreme learning machine | |
CN108871332A (en) | A kind of RFID indoor locating system and method based on XGBoost | |
JP6045747B2 (en) | Scalable real-time position detection based on overlapping neural networks | |
CN103996097B (en) | Laboratory equipment dynamic management system based on passive RFID | |
CN109212476B (en) | RFID indoor positioning algorithm based on DDPG | |
Zou et al. | An RFID indoor positioning system by using weighted path loss and extreme learning machine | |
CN110866528B (en) | Model training method, energy consumption use efficiency prediction method, device and medium | |
CN109115205A (en) | A kind of indoor fingerprint positioning method and system based on geomagnetic sensor array | |
CN109191455A (en) | A kind of field crop pest and disease disasters detection method based on SSD convolutional network | |
CN108984785A (en) | A kind of update method and device of the fingerprint base based on historical data and increment | |
CN106714302A (en) | Indoor positioning device based on BP-Landmarc neural network and control method | |
CN102054166B (en) | A kind of scene recognition method for Outdoor Augmented Reality System newly | |
CN106612495B (en) | A kind of indoor orientation method and system based on propagation loss study | |
CN109165540A (en) | A kind of pedestrian's searching method and device based on priori candidate frame selection strategy | |
CN110072183B (en) | Passive positioning fingerprint database construction method based on crowd sensing | |
Wang et al. | An RFID indoor positioning system by using particle swarm optimization-based artificial neural network | |
CN103220777A (en) | Mobile device positioning system | |
CN109327797A (en) | Mobile robot indoor locating system based on WiFi network signal | |
Kuo et al. | The application of an artificial immune system-based back-propagation neural network with feature selection to an RFID positioning system | |
CN113915840B (en) | Refrigerator and food material positioning method thereof | |
CN107490783A (en) | A kind of RFID localization methods based on ELM | |
Li et al. | RBF neural network based RFID indoor localization method using artificial immune system | |
CN109889981A (en) | A kind of localization method and system based on two sorting techniques | |
TWI411805B (en) | Method of positioning a rfid tag using geographic information system (gis) | |
Zhang et al. | RFID indoor localization algorithm based on dynamic netting |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
WD01 | Invention patent application deemed withdrawn after publication |
Application publication date: 20190409 |
|
WD01 | Invention patent application deemed withdrawn after publication |