CN107403195B - A kind of meteorologic parameter Intelligent Fusion processing method of carrying robot identification floor - Google Patents

A kind of meteorologic parameter Intelligent Fusion processing method of carrying robot identification floor Download PDF

Info

Publication number
CN107403195B
CN107403195B CN201710631218.0A CN201710631218A CN107403195B CN 107403195 B CN107403195 B CN 107403195B CN 201710631218 A CN201710631218 A CN 201710631218A CN 107403195 B CN107403195 B CN 107403195B
Authority
CN
China
Prior art keywords
floor
frog
model
air pressure
drosophila
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
Application number
CN201710631218.0A
Other languages
Chinese (zh)
Other versions
CN107403195A (en
Inventor
李燕飞
刘辉
米希伟
金楷荣
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Central South University
Original Assignee
Central South University
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by Central South University filed Critical Central South University
Priority to CN201710631218.0A priority Critical patent/CN107403195B/en
Publication of CN107403195A publication Critical patent/CN107403195A/en
Application granted granted Critical
Publication of CN107403195B publication Critical patent/CN107403195B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/24—Classification techniques
    • G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
    • G06F18/2411—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on the proximity to a decision surface, e.g. support vector machines
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
    • G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
    • 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]
    • 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
    • G06N3/008—Artificial life, i.e. computing arrangements simulating life based on physical entities controlled by simulated intelligence so as to replicate intelligent life forms, e.g. based on robots replicating pets or humans in their appearance or behaviour
    • G—PHYSICS
    • G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
    • G16Z—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS, NOT OTHERWISE PROVIDED FOR
    • G16Z99/00—Subject matter not provided for in other main groups of this subclass

Abstract

The invention provides a kind of meteorologic parameter Intelligent Fusion processing method of carrying robot identification floor, utilize lower-cost temperature, humidity, baroceptor gather data, establish database, characteristic for air pressure with height change, using big data treatment technology, the foundation of forecast model is completed using the MKSVM models of SFLA optimizations, the elevator floor Adaptive Identification of carrying robot under circumstances is realized, sustained height atmospheric pressure value under varying environment is efficiently solved and changes the situation for causing floor identification inaccurate.In addition, the present invention has high universality without transforming elevator interior or outside.

Description

A kind of meteorologic parameter Intelligent Fusion processing method of carrying robot identification floor
Technical field
The invention belongs to robot control field, more particularly to a kind of meteorologic parameter intelligence of carrying robot identification floor Method for amalgamation processing.
Background technology
With automatically control, the rapid development of the technology such as machinery, sensor, computer, production, manufacture, the control of robot Technology processed has obtained great lifting, and Robot industry enters the fast-developing epoch.Improve logistics because carrying robot has Management, realizes data analysis and remote control, prevents goods from damaging, and the advantages that realizing long-distance transportation, carrying robot is gradual As the important component in the systems such as modern mechanical manufacturing, logistics transportation, scientific experiment.Modern complexity transport task It is required that carrying robot can automatically complete the delivery operation of different floors.But how accurately, stably carrying robot Identify that elevator is still a problem so far to floor.
At present, carrying robot identification elevator mainly has 3 kinds of schemes to floor.1st kind of scheme is to pass through elevator interior Reading of the carrying robot to elevator floor is realized in transformation, i.e., the information directly established between elevator device and carrying robot is handed over Mutually, elevator floor information is such as read by single-chip microcomputer, this method can be stablized, be reliably achieved elevator floor identification, but be based on Security, convenience consider, elevator arrangement carrying robot floor signal interactive device is difficult under current conditions.2nd kind of side Case is to realize reading of the carrying robot to elevator floor by installing easy device additional outside elevator, such as in all floor elevator doors Floor label is sticked in face position, and label is identified by image procossing mode for carrying robot, and this method is without strong light Under the conditions of irradiation, visual occlusion etc., discrimination is higher, but is difficult to tackle complex environment.3rd kind of scheme is robot autonomous realization Elevator floor identifies that this method is more beneficial for the extensive use of carrying robot, but the existing side of this method without external modifications Elevator interior floor label is identified by image procossing mode for case such as robot, is equally difficult in adapt to complex environment.
The content of the invention
The invention provides a kind of meteorologic parameter Intelligent Fusion processing method of carrying robot identification floor, its purpose exists It is accurate to obtain with reference to the barometric information detected in real time in overcome the deficiencies in the prior art, the MKSVM models optimized using SFLA The level number of floor where robot.
A kind of meteorologic parameter Intelligent Fusion processing method of carrying robot identification floor, comprises the following steps：
Step 1：Floor history information data is gathered, builds floor information database；
The floor history information data includes the day each floor under various weather conditions in different time interval section Gas observation, the synoptic weather observation value include temperature, humidity and air pressure；
Step 2：Sample of each floor in different time interval in floor information database is clustered, obtained every The synoptic model set of individual floor, each synoptic model corresponding one group of temperature range, humidity section and air pressure section；
The sample refers to the synoptic weather observation value average gathered in a time interval, and time interval is by historical data Acquisition time is equidistantly divided as a continuous period, is set as 1 hour；
Step 3：Obtain the training set for building the air pressure floor forecast model based on synoptic model；
After air pressure average in all samples is divided according to synoptic model, to all floors under identical synoptic model Air pressure average merge, obtain the training of all air pressure averages of full floor under same synoptic model and corresponding floor level number Subset, the full floor training subset composing training set under all synoptic models；
Step 4：Build the air pressure floor forecast model based on synoptic model；
Using the air pressure average in all air pressure average training subsets of full floor under each synoptic model as input data, Each air pressure average corresponds to floor level number as output data, trains the classification mould of the multi-kernel support vector machine MKSVM based on SFLA Type, obtain the air pressure floor forecast model based on synoptic model；
Step 5：Floor level number and synoptic weather observation value are currently located using robot, determines the synoptic model of current floor, Call the air pressure floor forecast model of corresponding synoptic model；
Step 6：The real-time gas of floor where being taken a lift using the baroceptor collection robot loaded in robot Pressure, in input air pressure floor forecast model, output device people takes a lift the floor level number at place.
Further, the disaggregated model of the multi-kernel support vector machine MKSVM based on SFLA, is with same synoptic model Lower different air pressure mean datas and corresponding floor level number are established MKSVM models and entered respectively as inputting and exporting training data Row classification based training obtains；
Wherein, the kernel function for the MKSVM models established is Radial basis kernel function and Polynomial kernel function weighted sum Multinuclear kernel function, the parameter c and g of the MKSVM models established are in optimized selection using SFLA algorithms.
The multinuclear kernel function is Kmix=dKrbf+(1-d)Kpoly, d ∈ [0,1], wherein, KrbfRepresent gaussian radial basis function core Function, KpolyPolynomial kernel function is represented, d represents kernel function weights.
Further, the process that the parameter c and g of the MKSVM models are in optimized selection using SFLA algorithms is as follows：
(1) by the penalty coefficient c, nuclear parameter g and kernel function weights d of the MKSVM models as frog individual, and at random Initialize frog population；
(2) all historical samples of each floor under same synoptic model are inputted into the MKSVM models, obtains each sample Classification results, the average absolute value error between the floor classification results of all samples and true floor result is fitness Function, successively to every frog, calculate its fitness value；
(3) all frogs are arranged by fitness size descending, makes the 1st frog enter subgroup 1, the 2nd frog enters Subgroup 2, NmFrog enters subgroup Nm, Nm+1Frog enters subgroup 1, by that analogy, completes all frogs distribution；
(4) group that fitness is best preferably with fitness in worst frog individual and whole population in each subgroup is determined The optimal frog of body, after frog individual worst in each subgroup is eliminated, according to SFLA algorithms and given evolution number, to each The frog of subgroup carries out first evolution；
(5) frog of each subgroup is merged, arranges individual by fitness function value descending, re-mix and form new colony, And record the optimal frog of colony now；
(6) fitness of the optimal frog of colony is calculated, judges whether that the fitness maximum for reaching the optimal frog of colony is equal Number, if reaching, penalty coefficient c corresponding to optimal frog, nuclear parameter g and kernel function weights d are exported, not up to, then return to (3) Continue iteration.
Further, frog sum span is [50,500], and subcluster number span is [5,50], and evolution number takes It is [100-1000] to be worth scope, and fitness maximum equal times span is [5-30].
The difference of fitness is less than a setting value, it is believed that equal, setting value span is [0.01-0.1]；
Further, GMM methods are clustered using Gaussian Mixture, to each floor in floor information database in different time Sample in interval is clustered, and obtains the synoptic model set of each floor.
Further, it is as follows using FOA algorithms structure initial Gaussian mixing cluster GMM model, detailed process：
Step A：Randomly selected from single floor sample sets Ai q sample and single floor pattern count identical n Sample is as original cluster centre, corresponding n Gauss model, then randomly selects s sample as renewal drosophila；
Using n+s selected sample as the initial position of drosophila, population is all samples in single floor sample set, if Determine maximum iteration；
Step B：The flavor concentration of each drosophila is calculated, drosophila fitness is used as using the flavor concentration of drosophila；
Fitness function is Euclidean distance sum of nearest [q/ (n+s)]+1 sample point of separating fruit fly to drosophila；
Step C：Each drosophila position is compared, obtains n optimal drosophila position of flavor concentration in per generation drosophila population Put；
Step D：The s group drosophila worst to position re-starts random generation, and by the optimal n group drosophilas in last position Position is preserved；
Step E：Judge whether now optimal n group drosophilas position reaches required precision, or meet iterations, if reached Arrive, then export optimal n group drosophilas position, otherwise return to step C continues iteration renewal, until meeting end condition；
Step F：Using n sample point of optimal n group drosophila position correspondences as initial cluster center, and in population Remaining drosophila is classified according to apart from the distance between initial cluster center, obtains classification results；
Step G：Weight, average and the variance for obeying every a kind of sample probability-distribution function are calculated, obtains initial Gaussian Mix Clustering Model.
Beneficial effect
The invention provides a kind of meteorologic parameter Intelligent Fusion processing method of carrying robot identification floor, cost is utilized Relatively low temperature, humidity, baroceptor gather data, establish database, the characteristic for air pressure with height change, using big Data processing technique, the foundation of forecast model is completed using the MKSVM models of SFLA optimizations, realizes carrying robot in various rings Elevator floor Adaptive Identification under border, efficiently solving sustained height atmospheric pressure value under varying environment and changing causes floor identification not Accurate situation.In addition, the present invention has high universality without transforming elevator interior or outside.
In cluster process, floor information data are clustered using the FOA GMM clustering algorithms optimized, FOA algorithms are used In optimum choice initial cluster center, so that the weight, average, variance of each Gauss model are calculated respectively, after being optimized Initial Gaussian mixes Clustering Model, the method by changing the initial value for assigning EM, to optimize EM algorithms, GMM is reached more preferable Clustering Effect, the accuracy of GMM clusters is improved, and GMM clustering algorithms have excellent adaptive classification ability to data, will FOA algorithms are combined with GMM clustering algorithms, can be with significant increase Clustering Effect.
The foundation of forecast model is completed using the MKSVM models of SFLA optimizations, SFLA algorithms are used to optimize MKSVM algorithms Penalty coefficient c, nuclear parameter g and kernel function weights d, the ability of searching optimum of MKSVM algorithms is improved, and MKSVM models have Excellent classification estimated performance so that forecast model is more accurate, reliable.
Brief description of the drawings
Fig. 1 is the principle flow chart of the method for the invention；
Fig. 2 is that the GMM model that FOA optimizes in the present invention builds flow chart；
Fig. 3 is the MKSVM model construction flow charts that SFLA optimizes in the present invention；
Fig. 4 is the accuracy rate schematic diagram that floor is identified using the method for the invention.
Embodiment
As shown in figure 1, a kind of meteorologic parameter Intelligent Fusion processing method of carrying robot identification floor, including following step Suddenly：
Step 1, before robot comes into operation, collect the temperature of each floor, humidity, air pressure number under building different time According to establishing database and database remain in that renewal after robot comes into operation.By taking 10 floors as an example.
Step 2, by floor number it is floor 1- floors 10, data is classified according to different floors.
Step 3, setting adjacent two integral point be time interval per hour, extract 10 floors per hour interior temperature, humidity, The average of air pressure.
Step 4, to each single floor, using temperature, humidity, the average of air pressure as observation, by between each time Observation between septal area forms sample sets A as sample, then floor 1- floors 101-A10If each sample sets include 300 samples Product.
GMM (Gaussian Mixture clustering algorithm) is established to each sample collection, and uses the initial of FOA (drosophila algorithm) optimizations GMM Parameter, under same synoptic model, the observation at different time interval relatively, if suitable pattern has 15, such as certain building Layer pattern 1 be：15-20 DEG C of temperature, 70 ± 2%rh of humidity, air pressure 995-1000hPa；Pattern 2 is：20-25 DEG C of temperature, it is wet Spend 70 ± 2%rh, air pressure 985-990hPa etc..
The pattern 1- patterns 20 of single floor can be then obtained, establish the set of patterns of 10 floors respectively.
The GMM cluster process calculation procedures of FOA optimizations are as follows：
(1) randomly selected from single floor sample sets Ai with single 20 samples of floor pattern count identical as initial Cluster centre, corresponding 20 Gauss models, then randomly select 10 samples as renewal drosophilas, using 30 selected samples as The initial position of drosophila, population are the 300 all samples in single floor sample set, set maximum iteration, and mixing is high This model definition is as follows：
Wherein, N is Number of Models (being here 20), πjFor the weight of j-th of Gauss model, p (xj) is j-th of Gaussian mode The probability density function of type, its average are μj, variance δj, x is sample；
(2) it is Euclidean distance sum of the 11 nearest sample spots of separating fruit fly to drosophila to calculate flavor concentration decision content S, S, Using S as fitness function, fitness calculating is done to each drosophila position；
(3) each drosophila position is compared, obtains 20 optimal drosophila positions of flavor concentration in per generation drosophila population；
(4) when screening every time, worst to position 10 groups of drosophilas re-start random generation, and last position is optimal 20 groups of drosophila positions preserved；
(5) judge whether 20 groups of now optimal drosophila positions reach required precision, or meet iterations, if reached Arrive, then export 20 groups of optimal drosophila positions, otherwise continue iteration renewal, until meeting end condition；
(6) optimal 20 groups of drosophila position correspondences, 20 sample points, to 280 samples in data set in addition to this 20 points This point, the Euclidean distance with this 20 points is calculated respectively, according to the most short principle of distance, 280 sample points are put into 20 classes Certain is a kind of；
(7) to Various types of data, weight, average, variance are calculated respectively, the initial Gaussian mixing cluster mould after being optimized Type；
(8) gauss hybrid models are calculated using EM algorithms, completes the cluster of sample, obtain the pattern of single floor 1- patterns 20.
The air pressure mean data of each sample data, is obtained complete under step 5, all same patterns of floor (same to period) of extraction 1 '-pattern of floor pattern 20 ', establish full floor set of patterns.
Step 6, as shown in Fig. 2 for 1 '-pattern of full floor pattern 20 ', with different air pressure mean datas under same pattern For input, using corresponding number of floor levels as output, the MKSVM (multi-kernel support vector machine) for establishing SFLA (leapfrog algorithm) optimizations is carried out Training, obtains training pattern 1 '-model 20 ', establishes Models Sets.
The MKSVM calculation procedures of SFLA optimizations are as follows：
(1) MKSVM uses the multinuclear kernel function of Radial basis kernel function and Polynomial kernel function weighted sum, consisting of： Kmix=dKrbf+(1-d)Kpoly, d ∈ [0,1], wherein KrbfRepresent gaussian radial basis function, KpolyRepresent Polynomial kernel function, D represents kernel function weights.
(2) by MKSVM penalty coefficient c, nuclear parameter g and kernel function weights d as frog individual, set frog sum as 200, subcluster number 15, evolution number is 300, and fitness maximum equal times were 10 (difference of fitness is less than 0.01)；
(3) the average absolute value error being calculated using MKSVM models, to every frog, calculates it as fitness function Fitness value；
(4) all frogs are arranged by fitness size descending, makes the 1st frog enter subgroup 1, the 2nd frog enters Subgroup 2, the 15th frog enter subgroup 15, and the 16th frog enters subgroup 1, by that analogy, complete all frog distribution；
(5) each subgroup fitness is determined preferably with worst frog individual and the optimal frog of colony, it is public according to SFLA algorithms Formula and given evolution number, carry out first evolution in each subgroup；
(6) after member evolution is performed in each subgroup, the frog of each subgroup is merged, arranged by fitness function value descending Row individual, re-mix and form new colony, and record the optimal frog of colony now；
(7) each optimal frog of colony is contrasted, judges whether to reach fitness maximum equal times 10 times, if reaching, Optimal solution is exported, not up to then continues iteration；
(8) MKSVM is built with the optimized parameter tried to achieve, using different air pressure mean datas under same pattern as input, with phase The number of floor levels answered is output, and MKSVM models are trained, obtain training pattern.
Step 7, set robot start working when, in floor 5.Robot obtains the temperature, humidity, air pressure number of floor 5 According to (instantaneous value or the average in the short time), according to each floor set of patterns, pattern-recognition is carried out with corresponding floor 5, is judged now Pattern is the pattern 8 of floor 5；According to full floor set of patterns, now corresponding full floor pattern 8 ' is found out；Similarly, according to model Collection, finds out now corresponding model 8 '.
Step 8, robot obtain instruction and go to floor 9, now initialize microprocessor and COM port.
Step 9, robot arrival floor 7, elevator stop, and current gas pressure data (instantaneous value or short time are collected by robot Interior average).
Current gas pressure data are brought into the model 8 ' trained and judged by step 10, robot, and now output is floor 7, robot is remained in elevator, waits elevator to stop next time；
Step 11, robot arrival floor 9, elevator stop, and current gas pressure data (instantaneous value or short time are collected by robot Interior average), and current gas pressure data are brought into the model 8 ' trained and judged, now output is floor 9, then robot Elevator is left after elevator door is fully opened.
Identify that the accuracy rate of floor carries out experimental examination, experiment to certain carrying robot using method proposed by the invention Operating mode is as follows：(1) it is separately operable discrimination method 100 times in not same date, adds up to operation 10 days, altogether 1000 times；(2) robot Vehicle-mounted notebook read elevator PLC floor information automatically, while but after robot runs to different floors, with this patent The floor information of the method identification current layer proposed is simultaneously stored into the vehicle-mounted notebook of robot；(3) comparative analysis sheet is special The recognition accuracy for the method that profit is proposed.Result of the test as shown in Figure 4, this 1000 times experiment in, what this patent was proposed Method successfully identifies current robot floor time 985 times (output " 1 " is represented and identified successfully), 21 (output " 0 " of wrong identification Represent identification mistake), therefore recognition success rate is 98.5%, shows that the floor recognition accuracy of the method for the invention is high.
Specific embodiment described herein is only to spirit explanation for example of the invention.Technology belonging to the present invention is led The technical staff in domain can be made various modifications or supplement to described specific embodiment or be replaced using similar mode Generation, but without departing from the spiritual of the present invention or surmount scope defined in appended claims.

Claims (6)

1. a kind of meteorologic parameter Intelligent Fusion processing method of carrying robot identification floor, it is characterised in that including following step Suddenly：
Step 1：Floor history information data is gathered, builds floor information database；
The floor history information data includes weather of each floor under various weather conditions in different time interval section and seen Measured value, the synoptic weather observation value include temperature, humidity and air pressure；
Step 2：Sample of each floor in different time interval in floor information database is clustered, obtains each building The synoptic model set of layer, each synoptic model corresponding one group of temperature range, humidity section and air pressure section；
The sample refers to the synoptic weather observation value average gathered in a time interval, and time interval is to gather historical data Time as a continuous period, is equidistantly divided, is set as 1 hour；
Step 3：Obtain the training set for building the air pressure floor forecast model based on synoptic model；
After air pressure average in all samples is divided according to synoptic model, to the gas of all floors under identical synoptic model Pressure average merges, and obtains training of all air pressure averages of full floor and corresponding floor level number under same synoptic model Collect, the full floor training subset composing training set under all synoptic models；
Step 4：Build the air pressure floor forecast model based on synoptic model；
Using the air pressure average in all air pressure average training subsets of full floor under each synoptic model as input data, each gas Pressure average corresponds to floor level number as output data, trains the disaggregated model of the multi-kernel support vector machine MKSVM based on SFLA, obtains Obtain the air pressure floor forecast model based on synoptic model；
Step 5：Floor level number and synoptic weather observation value are currently located using robot, determines the synoptic model of current floor, is called The air pressure floor forecast model of corresponding synoptic model；
Step 6：The real-time air pressure of floor, defeated where being taken a lift using the baroceptor collection robot loaded in robot Enter in air pressure floor forecast model, output device people take a lift where floor level number.
2. according to the method for claim 1, it is characterised in that the multi-kernel support vector machine MKSVM's based on SFLA Disaggregated model, be using same synoptic model under different air pressure mean datas and corresponding floor level number as input and output Training data, establish MKSVM models and carry out classification based training acquisition；
Wherein, the kernel function for the MKSVM models established is Radial basis kernel function and the multinuclear of Polynomial kernel function weighted sum Kernel function, the parameter c and g of the MKSVM models established are in optimized selection using SFLA algorithms；
The multinuclear kernel function is Kmix=dKrbf+(1-d)Kpoly, d ∈ [0,1], wherein, KrbfRepresent gaussian radial basis function, KpolyPolynomial kernel function is represented, d represents kernel function weights.
3. according to the method for claim 2, it is characterised in that the parameter c and g of the MKSVM models use SFLA algorithms The process being in optimized selection is as follows：
(1) by the penalty coefficient c, nuclear parameter g and kernel function weights d of the MKSVM models as frog individual, and it is initial at random Change frog population；
(2) all historical samples of each floor under same synoptic model are inputted into the MKSVM models, obtains point of each sample Class result, the average absolute value error between the floor classification results of all samples and true floor result is fitness letter Number, successively to every frog, calculates its fitness value；
(3) all frogs are arranged by fitness size descending, makes the 1st frog enter subgroup 1, the 2nd frog enters subgroup 2, NmFrog enters subgroup Nm, Nm+1Frog enters subgroup 1, by that analogy, completes all frogs distribution；
(4) determine that the best colony best with fitness in worst frog individual and whole population of fitness is most in each subgroup Excellent frog, after frog individual worst in each subgroup is eliminated, according to SFLA algorithms and given evolution number, to each subgroup Frog carry out first evolution；
(5) frog of each subgroup is merged, arranges individual by fitness function value descending, re-mix and form new colony, and remember The optimal frog of colony of record now；
(6) fitness of the optimal frog of colony is calculated, judges whether the fitness maximum equal times for reaching the optimal frog of colony, If reaching, penalty coefficient c corresponding to optimal frog, nuclear parameter g and kernel function weights d are exported, not up to, then (3) is returned and continues Iteration.
4. according to the method for claim 3, it is characterised in that frog sum span is [50,500], and subcluster number takes It is [5,50] to be worth scope, and evolution number span is [100-1000], and fitness maximum equal times span is [5- 30]。
5. according to the method described in claim any one of 1-4, it is characterised in that GMM methods are clustered using Gaussian Mixture, to building Sample of each floor in different time interval in layer information database is clustered, and obtains the synoptic model collection of each floor Close.
6. according to the method for claim 5, it is characterised in that using FOA algorithms structure initial Gaussian mixing cluster GMM moulds Type, detailed process are as follows：
Step A：Randomly selected from single floor sample sets Ai q sample and n sample of single floor pattern count identical As original cluster centre, corresponding n Gauss model, then s sample is randomly selected as renewal drosophila；
Using n+s selected sample as the initial position of drosophila, population is all samples in single floor sample set, and setting is most Big iterations；
Step B：The flavor concentration of each drosophila is calculated, drosophila fitness is used as using the flavor concentration of drosophila；
Fitness function is Euclidean distance sum of nearest [q/ (n+s)]+1 sample point of separating fruit fly to drosophila；
Step C：Each drosophila position is compared, obtains n optimal drosophila position of flavor concentration in per generation drosophila population；
Step D：The s group drosophila worst to position re-starts random generation, and by the optimal n group drosophilas position in last position Preserved；
Step E：Judge whether now optimal n group drosophilas position reaches required precision, or meet iterations, if reached, Optimal n group drosophilas position is then exported, otherwise return to step C continues iteration renewal, until meeting end condition；
Step F：Using n sample point of optimal n group drosophila position correspondences as initial cluster center, and to remaining in population Drosophila is classified according to apart from the distance between initial cluster center, obtains classification results；
Step G：Weight, average and the variance for obeying every a kind of sample probability-distribution function are calculated, obtains initial Gaussian mixing Clustering Model.
CN201710631218.0A 2017-07-28 2017-07-28 A kind of meteorologic parameter Intelligent Fusion processing method of carrying robot identification floor Active CN107403195B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201710631218.0A CN107403195B (en) 2017-07-28 2017-07-28 A kind of meteorologic parameter Intelligent Fusion processing method of carrying robot identification floor

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201710631218.0A CN107403195B (en) 2017-07-28 2017-07-28 A kind of meteorologic parameter Intelligent Fusion processing method of carrying robot identification floor

Publications (2)

Publication Number Publication Date
CN107403195A CN107403195A (en) 2017-11-28
CN107403195B true CN107403195B (en) 2018-03-27

Family

ID=60401645

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201710631218.0A Active CN107403195B (en) 2017-07-28 2017-07-28 A kind of meteorologic parameter Intelligent Fusion processing method of carrying robot identification floor

Country Status (1)

Country Link
CN (1) CN107403195B (en)

Families Citing this family (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN110929775B (en) * 2019-11-18 2022-08-12 南通大学 Convolutional neural network weight optimization method for retinopathy classification

Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN106793067A (en) * 2016-11-29 2017-05-31 上海斐讯数据通信技术有限公司 A kind of many floor indoor orientation methods and server based on joint network
CN106792553A (en) * 2016-11-22 2017-05-31 上海斐讯数据通信技术有限公司 A kind of many floor location methods and server based on wifi
CN106851585A (en) * 2017-01-12 2017-06-13 杭州电子科技大学 A kind of mixing floor location method based on barometer and WiFi

Family Cites Families (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US9078102B2 (en) * 2012-11-12 2015-07-07 Qualcomm Incorporated Techniques for generating environment and reference data reports for particular environments on behalf of mobile devices
CN103874200B (en) * 2014-03-19 2018-01-09 北京邮电大学 A kind of floor recognition methods and system
FR3036821B1 (en) * 2015-05-27 2017-06-23 Airbus Helicopters DECISION-ASSISTING METHOD AND DEVICE FOR DETECTING AND SIGNALING LOCAL WEATHER RISKS RELATED TO THE SURFACE OF THE AIRCRAFT SURFACE

Patent Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN106792553A (en) * 2016-11-22 2017-05-31 上海斐讯数据通信技术有限公司 A kind of many floor location methods and server based on wifi
CN106793067A (en) * 2016-11-29 2017-05-31 上海斐讯数据通信技术有限公司 A kind of many floor indoor orientation methods and server based on joint network
CN106851585A (en) * 2017-01-12 2017-06-13 杭州电子科技大学 A kind of mixing floor location method based on barometer and WiFi

Also Published As

Publication number Publication date
CN107403195A (en) 2017-11-28

Similar Documents

Publication Publication Date Title
CN107368926B (en) A kind of more natural parameter sensing method for amalgamation processing of intelligent environment carrying robot identification floor
CN107423412B (en) A kind of method of the carrying robot Intelligent Recognition floor based on meteorological sensing time series pattern
CN107368858A (en) A kind of parametrization measurement multi-model intelligent method for fusing of intelligent environment carrying robot identification floor
CN114997051B (en) Aero-engine life prediction and health assessment method based on transfer learning
CN111523603B (en) Ship power equipment fault identification method
CN108197648A (en) A kind of Fault Diagnosis Method of Hydro-generating Unit and system based on LSTM deep learning models
CN109063594A (en) Remote sensing images fast target detection method based on YOLOv2
CN109000645A (en) Complex environment target classics track extracting method
CN109146162B (en) A kind of probability wind speed forecasting method based on integrated Recognition with Recurrent Neural Network
CN109191922A (en) A kind of extensive four-dimensional dynamic track prediction method and device
CN108280397A (en) Human body image hair detection method based on depth convolutional neural networks
CN107808375B (en) Merge the rice disease image detecting method of a variety of context deep learning models
CN107341447A (en) A kind of face verification mechanism based on depth convolutional neural networks and evidence k nearest neighbor
CN104992223A (en) Dense population estimation method based on deep learning
CN111145042A (en) Power distribution network voltage abnormity diagnosis method adopting full-connection neural network
CN110210569A (en) Chemical storage tank Outlier Detection Algorithm research based on FCM-LSTM
CN106778501A (en) Video human face ONLINE RECOGNITION method based on compression tracking with IHDR incremental learnings
CN105930792A (en) Human action classification method based on video local feature dictionary
CN111563549B (en) Medical image clustering method based on multitasking evolutionary algorithm
CN107403195B (en) A kind of meteorologic parameter Intelligent Fusion processing method of carrying robot identification floor
CN111031585A (en) Dormancy control method, device, storage medium and server
CN109002928A (en) A kind of electric load peak value prediction technique and device based on Bayesian network model
CN111967326B (en) Gait recognition method based on lightweight multi-scale feature extraction
CN107195297A (en) A kind of normalized TSP question flock of birds speech recognition system of fused data
CN109815887B (en) Multi-agent cooperation-based face image classification method under complex illumination

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
GR01 Patent grant
GR01 Patent grant