CN108319572A - A kind of mixed self-adapting particle filter method of mobile robot fault diagnosis - Google Patents

A kind of mixed self-adapting particle filter method of mobile robot fault diagnosis Download PDF

Info

Publication number
CN108319572A
CN108319572A CN201711095126.1A CN201711095126A CN108319572A CN 108319572 A CN108319572 A CN 108319572A CN 201711095126 A CN201711095126 A CN 201711095126A CN 108319572 A CN108319572 A CN 108319572A
Authority
CN
China
Prior art keywords
particle
mobile robot
state
particles
particle collection
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
Application number
CN201711095126.1A
Other languages
Chinese (zh)
Inventor
段琢华
杨亮
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.)
University of Electronic Science and Technology of China Zhongshan Institute
Original Assignee
University of Electronic Science and Technology of China Zhongshan Institute
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 University of Electronic Science and Technology of China Zhongshan Institute filed Critical University of Electronic Science and Technology of China Zhongshan Institute
Priority to CN201711095126.1A priority Critical patent/CN108319572A/en
Publication of CN108319572A publication Critical patent/CN108319572A/en
Pending legal-status Critical Current

Links

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F17/00—Digital computing or data processing equipment or methods, specially adapted for specific functions
    • G06F17/10—Complex mathematical operations
    • G06F17/18—Complex mathematical operations for evaluating statistical data, e.g. average values, frequency distributions, probability functions, regression analysis
    • B—PERFORMING OPERATIONS; TRANSPORTING
    • B25—HAND TOOLS; PORTABLE POWER-DRIVEN TOOLS; MANIPULATORS
    • B25J—MANIPULATORS; CHAMBERS PROVIDED WITH MANIPULATION DEVICES
    • B25J19/00—Accessories fitted to manipulators, e.g. for monitoring, for viewing; Safety devices combined with or specially adapted for use in connection with manipulators
    • B25J19/0095—Means or methods for testing manipulators

Abstract

The present invention relates to a kind of mixed self-adapting particle filter methods of mobile robot fault diagnosis, and state space is adaptively combined with the adaptive two kinds of mechanism of population.The adaptive core concept of state space is that sample space is tied to a certain fuzzy subset in " total state space " according to the domain knowledge of mobile robot fault diagnosis, number of particles is then adjusted by the KL distances between the APPROXIMATE DISTRIBUTION of the particle set representations of two different number of particles of synchronization, domain knowledge is used to portray the mobile robot probability that various failures occur under different motion pattern, since domain knowledge uncertainty caused by driving and executing agency is indicated by fuzzy set.

Description

A kind of mixed self-adapting particle filter method of mobile robot fault diagnosis
Technical field
The present invention relates to a kind of mixed self-adapting particle filter methods of mobile robot fault diagnosis, belong to mobile machine People's fault diagnosis field.
Background technology
Fault diagnosis is most important for localization for Mobile Robot, modeling, navigation or even safety.Fault diagnosis is an allusion quotation The random hybrid system state estimation problem of type.Particle filter is the Monte Carlo method for monitoring dynamical system, passes through cum rights Sample (i.e. particle) with collecting imparametrization approximation probability distribution.Particle filter is in the complicated higher-dimension problem of processing, main problem If overcoming the contradiction between precision and efficiency：In order to improve precision, it is necessary to increase number of particles；And in order to improve efficiency, Number of particles must be reduced.
Document is (referring to " remaining cleverness, Cai Zixing, Tan Ping, Duan Zhuohua：《Based on multi-modal evolution Rao-Blackwellized The fault diagnosis of the mobile robot dead reckoning system of particle filter》, control and decision, the 12nd phase in 2010 ") be directed to one The troubleshooting issue of class mobile robot dead reckoning system proposes a kind of multi-modal evolution Rao-Blackwellized Subfilter, to solve by the poor caused problem of inconsistency of particle, using intersection and variation population policy optimization, according to grain Discontinuous Factors are added in sub- diversity, and using the MERBPF corresponding to Expert Rules judgement robot motion state, construction complexity is patrolled Collect table.The defect of this method is to use multiple particles filter for different modalities so that turns between particle filter More time and memory headroom will be expended by changing；In addition, this method can not dynamically adjust number of particles.
Invention content
The present invention relates to a kind of mixed self-adapting particle filter methods of mobile robot fault diagnosis, certainly by state space It adapts to and the adaptive two kinds of mechanism of population combines.The adaptive core concept of state space is according to mobile machine Sample space is tied to a certain fuzzy subset in " total state space " by the domain knowledge of people's fault diagnosis, and number of particles then passes through KL distances between the APPROXIMATE DISTRIBUTION of the particle set representations of the different number of particles of synchronization two adjust, and domain knowledge is used for The mobile robot probability that various failures occur under different motion pattern is portrayed, due to being led caused by driving and executing agency Domain knowledge uncertainty is indicated by fuzzy set.It is as follows：
Step 1 initialization step:
Defective space S, model space M, number of particles are set as N0, arrange parameter c, γ, α, β initialize particle collectionWherein
Step 2 determines the characteristic function C of each mode m in model space M according to domain knowledgem:S → { 0,1 }, it is each The membership function μ of a mode mm(uL,uR), wherein uL,uRMobile robot left and right wheels actuating speed is indicated respectively；
Step 3 is made iteratively step 4 to step 9 for each time step t,
It is respectively u that step 4, which is calculated according to formula (1) in mobile robot left and right wheels actuating speed,L,uRUnder conditions of failure shape The conditional probability p that state s occursa(s|uL,uR)
WhereinFor normalization factors；
Step 5 is according to formula (2) design conditions transition probability p (st|st-1,uL,uR)
p(st|st-1,uL,uR)=η2pa(st|uL,uR)pb(st|st-1)(2)
Wherein st、st-1The malfunction of t moment and t-1 moment, p are indicated respectivelyb(st|st-1) indicate malfunction transfer Equation,For constant of standardizing；Step 6 calculates importance sampling
Conditional transition probability p (the s obtained are calculated according to formula (2)t|st-1,uL,uR) sampling generation particle collection Wherein WhereinT moment is indicated respectively Malfunction, continuous state and the weight that i-th particle indicates, Nt-1Indicate the number of particles at t-1 moment, ztWhen indicating t The measurement of sensor is carved,Expression is nonserviceabledUnder from continuous state metastasis model；It indicates in event Barrier stateUnder measurement model；Step 7 calculates maximum a-posteriori estimation
Step 7.1 calculates edge distribution
Step 7.2 state estimation
Wherein δ indicates dirac functions
Step 8 number of particles adjusts
Step 8.1 is from particle collectionMiddle extraction Nt-1A particle generates new particle collectionSo thatIt is equal toProbability beWherein
Step 8.2 enablesGenerating population as stated above isParticle collectionWherein
Step 8.3 above-mentioned two particle collection shows respectively two kinds of approximations of posterior probability density, about discrete state Edge distribution be expressed as：
Particle collection is indicated respectivelyWithThe determining edge distribution about discrete state；
Step 8.4 calculates its KL distance：
Step 8.5if ρ ＞ γ Nt=α Nt-1else Nt=β Nt-1
Step 9 resampling：FromResampling NtA particle generates new particle collection
From particle collectionMiddle extraction NtA particle generates new particle collectionSo thatIt is equal toProbability beWherein
For each particle i, enable
Specific implementation mode
The specific implementation method of the present invention is described in detail below.The present invention relates to a kind of mobile robot failures to examine Disconnected mixed self-adapting particle filter method adaptively organically combines state space with the adaptive two kinds of mechanism of population Come.The adaptive core concept of state space is to be tied to sample space according to the domain knowledge of mobile robot fault diagnosis The a certain fuzzy subset in " total state space ", number of particles then pass through the particle set representations of two different number of particles of synchronization APPROXIMATE DISTRIBUTION between KL distances adjust, domain knowledge for portray mobile robot under the different motion pattern it is various therefore Hinder the probability occurred, since domain knowledge uncertainty caused by driving and executing agency is indicated by fuzzy set.
It is as follows：
Step 1 initialization step:
Defective space S, model space M, number of particles are set as N0=100, arrange parameter c=2, γ=0.1, α= 1.2, β=0.9, initialize particle collectionWherein
Step 2 determines the characteristic function C of each mode m in model space M according to domain knowledgem:S → { 0,1 }, it is each The membership function μ of a mode mm(uL,uR), wherein uL,uRMobile robot left and right wheels actuating speed is indicated respectively；
Step 3 is made iteratively step 4 to step 9 for each time step t,
It is respectively u that step 4, which is calculated according to formula (1) in mobile robot left and right wheels actuating speed,L,uRUnder conditions of failure shape The conditional probability p that state s occursa(s|uL,uR)
WhereinFor normalization factors；
Step 5 is according to formula (2) design conditions transition probability p (st|st-1,uL,uR)
p(st|st-1,uL,uR)=η2pa(st|uL,uR)pb(st|st-1) (2)
Wherein st、st-1The malfunction of t moment and t-1 moment, p are indicated respectivelyb(st|st-1) indicate malfunction transfer Equation,For constant of standardizing；
Step 6 calculates importance sampling
Conditional transition probability p (the s obtained are calculated according to formula (2)t|st-1,uL,uR) sampling generation particle collection Wherein WhereinT moment the is indicated respectively Malfunction, continuous state and the weight that i particle indicates, Nt-1Indicate the number of particles at t-1 moment, ztIndicate t moment The measurement of sensor,Expression is nonserviceabledUnder from continuous state metastasis model；It indicates in failure StateUnder measurement model；
Step 7 calculates maximum a-posteriori estimation
Step 7.1 calculates edge distribution
Step 7.2 state estimation
Wherein δ indicates dirac functions
Step 8 number of particles adjusts
Step 8.1 is from particle collectionMiddle extraction Nt-1A particle generates new particle collectionSo thatIt is equal toProbability beWherein
Step 8.2 enablesGenerating population as stated above isParticle collectionWherein
Step 8.3 above-mentioned two particle collection shows respectively two kinds of approximations of posterior probability density, about discrete state Edge distribution be expressed as：
Particle collection is indicated respectivelyWithThe determining edge distribution about discrete state；
Step 8.4 calculates its KL distance：
Step 8.5if ρ ＞ γ Nt=α Nt-1else Nt=β Nt-1
Step 9 resampling：FromResampling NtA particle generates new particle collection
From particle collectionMiddle extraction NtA particle generates new particle collectionSo thatIt is equal toProbability beWherein
For each particle i, enable
The above, only best mode for carrying out the invention, any one skilled in the art is in the present invention In the technical scope of disclosure, the simple change or equivalence replacement of the technical solution that can be become apparent to each fall within the present invention's In protection domain.

Claims (1)

1. a kind of mixed self-adapting particle filter method of mobile robot fault diagnosis, by state space is adaptive and population Adaptive two kinds of mechanism combines；The adaptive core concept of state space is according to mobile robot fault diagnosis Sample space is tied to a certain fuzzy subset in " total state space " by domain knowledge, and number of particles is then by synchronization two KL distances between the APPROXIMATE DISTRIBUTION of the particle set representations of different number of particles adjust, and domain knowledge is for portraying mobile machine People's probability that various failures occur under different motion pattern, since domain knowledge caused by driving and executing agency is uncertain Property is indicated by fuzzy set；It is as follows：
Step 1：Initialization step:Defective space S, model space M, number of particles are set as, arrange parameter c, , initialize particle collection, wherein,,；
Step 2：The characteristic function of each mode m in model space M is determined according to domain knowledge, often The membership function of one mode m, whereinMobile robot left and right wheels actuating speed is indicated respectively；
Step 3：For each time stept, it is made iteratively step 4 to step 9,
Step 4：According to formula（1）It calculates in mobile robot left and right wheels actuating speed and is respectivelyUnder conditions of malfunction s The conditional probability of generation
（1）
WhereinFor normalization factors;
Step 5：According to formula（2）Design conditions transition probabilityp(s t |s t-1 ,u L,u R)
p(s t |s t-1 ,u L,u R)= p a(s t |u L,u R)p b(s t |s t-1 ) (2)
Whereins t 、s t-1 It indicates respectivelytMoment andtThe malfunction at -1 moment,p b(s t |s t-1 ) indicate malfunction equation of transfer,For constant of standardizing;
Step 6：Calculate importance sampling：According to formula（2）Calculate the conditional transition probability obtainedp(s t | s t-1,u L,u R) sampling life Granulating subset, wherein,,,；WhereinThe failure that i-th of particle of t moment indicates is indicated respectively State, continuous state and weight,Indicate the number of particles at t-1 moment,Indicate the measurement of t moment sensor,Expression is nonserviceabledUnder from continuous state metastasis model；It indicates in failure shape StateUnder measurement model；
Step 7：Calculate maximum a-posteriori estimation：
Step 7.1 calculates edge distribution；
Step 7.2 state estimation；
WhereinIndicate dirac functions
Step 8：Number of particles adjusts
Step 8.1 is from particle collectionMiddle extractionA particle generates new particle collection, So thatIt is equal toProbability be, wherein。
Step 8.2 enables, generating population as stated above isParticle collection, Wherein；
Step 8.3 above-mentioned two particle collection shows respectively two kinds of approximations of posterior probability density, the side about discrete state Fate cloth is expressed as：
Particle collection is indicated respectivelyWithIt is determining about discrete state Edge distribution；
Step 8.4 calculates its KL distance：
Step 8.5 if else
Step 9：Resampling：FromResamplingA particle generates new particle collection
From particle collectionMiddle extractionA particle generates new particle collectionSo thatIt is equal toProbability be, wherein
For each particlei, enable。
CN201711095126.1A 2017-11-09 2017-11-09 A kind of mixed self-adapting particle filter method of mobile robot fault diagnosis Pending CN108319572A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201711095126.1A CN108319572A (en) 2017-11-09 2017-11-09 A kind of mixed self-adapting particle filter method of mobile robot fault diagnosis

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201711095126.1A CN108319572A (en) 2017-11-09 2017-11-09 A kind of mixed self-adapting particle filter method of mobile robot fault diagnosis

Publications (1)

Publication Number Publication Date
CN108319572A true CN108319572A (en) 2018-07-24

Family

ID=62893053

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201711095126.1A Pending CN108319572A (en) 2017-11-09 2017-11-09 A kind of mixed self-adapting particle filter method of mobile robot fault diagnosis

Country Status (1)

Country Link
CN (1) CN108319572A (en)

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN108972553A (en) * 2018-08-06 2018-12-11 北京邮电大学 A kind of space manipulator fault detection method based on particle filter algorithm
CN113375658A (en) * 2021-06-15 2021-09-10 电子科技大学中山学院 Method and system for simultaneously FDD and SLAM under mobile robot fault

Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20110054689A1 (en) * 2009-09-03 2011-03-03 Battelle Energy Alliance, Llc Robots, systems, and methods for hazard evaluation and visualization
CN103795373A (en) * 2013-11-29 2014-05-14 电子科技大学中山学院 Particle filter generating method for incomplete system fault diagnosis
CN107038459A (en) * 2017-04-07 2017-08-11 电子科技大学中山学院 A kind of incomplete system method for diagnosing faults merged based on model and data-driven
CN107246873A (en) * 2017-07-03 2017-10-13 哈尔滨工程大学 A kind of method of the mobile robot simultaneous localization and mapping based on improved particle filter

Patent Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20110054689A1 (en) * 2009-09-03 2011-03-03 Battelle Energy Alliance, Llc Robots, systems, and methods for hazard evaluation and visualization
CN103795373A (en) * 2013-11-29 2014-05-14 电子科技大学中山学院 Particle filter generating method for incomplete system fault diagnosis
CN107038459A (en) * 2017-04-07 2017-08-11 电子科技大学中山学院 A kind of incomplete system method for diagnosing faults merged based on model and data-driven
CN107246873A (en) * 2017-07-03 2017-10-13 哈尔滨工程大学 A kind of method of the mobile robot simultaneous localization and mapping based on improved particle filter

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
段琢华: "基于自适应粒子滤波器的移动机器人故障诊断理论与方法研究", 《中国博士学位论文全文数据库 信息科技辑》 *

Cited By (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN108972553A (en) * 2018-08-06 2018-12-11 北京邮电大学 A kind of space manipulator fault detection method based on particle filter algorithm
CN108972553B (en) * 2018-08-06 2021-08-13 北京邮电大学 Space manipulator fault detection method based on particle filter algorithm
CN113375658A (en) * 2021-06-15 2021-09-10 电子科技大学中山学院 Method and system for simultaneously FDD and SLAM under mobile robot fault
CN113375658B (en) * 2021-06-15 2023-05-09 电子科技大学中山学院 Method and system for simultaneously FDD and SLAM under fault of mobile robot

Similar Documents

Publication Publication Date Title
US11693763B2 (en) Resilient estimation for grid situational awareness
US20190221187A1 (en) System, apparatus and methods for adaptive data transport and optimization of application execution
Triguero et al. SEG-SSC: A framework based on synthetic examples generation for self-labeled semi-supervised classification
Gu et al. Bi-parameter space partition for cost-sensitive SVM
US20200097651A1 (en) Systems and methods to achieve robustness and security in medical devices
CN108319572A (en) A kind of mixed self-adapting particle filter method of mobile robot fault diagnosis
Zając Online fault detection of a mobile robot with a parallelized particle filter
Verbiest et al. Improving SMOTE with fuzzy rough prototype selection to detect noise in imbalanced classification data
Chen et al. Multi-granular mining for boundary regions in three-way decision theory
Wang et al. A modified generative adversarial network for fault diagnosis in high-speed train components with imbalanced and heterogeneous monitoring data
Manukyan et al. Classification of imbalanced data with a geometric digraph family
Prachuabsupakij CLUS: A new hybrid sampling classification for imbalanced data
Ryu et al. APPFL: open-source software framework for privacy-preserving federated learning
CN115993781A (en) Network attack resistant unmanned cluster system cooperative control method, terminal and storage medium
Han et al. GAN-SAE based fault diagnosis method for electrically driven feed pumps
Kan et al. Bi-level doubly variational learning for energy-based latent variable models
Hernández-Orallo et al. Training on the test set: Mapping the system-problem space in AI
Yu et al. A particle filter and SVM integration framework for fault-proneness prediction in robot dead reckoning system
US20210209489A1 (en) Processing a classifier
Rada-Vilela et al. Evaporation mechanisms for particle swarm optimization
Wu et al. Rule-based fuzzy classifier based on quantum ant optimization algorithm
Li et al. Distributed sensor analysis for fault detection in tightly-coupled multi-robot team tasks
Mastriani Optimal estimation of states in quantum image processing
CN105450751B (en) A kind of system improving cloud computing environment stability
Kamapantula et al. Feature ranking in transcriptional networks: packet receipt as a dynamical metric

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
RJ01 Rejection of invention patent application after publication

Application publication date: 20180724

RJ01 Rejection of invention patent application after publication