CN103616021A - Global localization method and device - Google Patents

Global localization method and device Download PDF

Info

Publication number
CN103616021A
CN103616021A CN201310649076.2A CN201310649076A CN103616021A CN 103616021 A CN103616021 A CN 103616021A CN 201310649076 A CN201310649076 A CN 201310649076A CN 103616021 A CN103616021 A CN 103616021A
Authority
CN
China
Prior art keywords
unit
robot
distribution
perception information
pose
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
CN201310649076.2A
Other languages
Chinese (zh)
Inventor
厉茂海
林睿
***
陈国栋
孙荣川
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.)
Zhangjiagang Institute of Industrial Technologies Soochow University
Original Assignee
Zhangjiagang Institute of Industrial Technologies Soochow 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 Zhangjiagang Institute of Industrial Technologies Soochow University filed Critical Zhangjiagang Institute of Industrial Technologies Soochow University
Priority to CN201310649076.2A priority Critical patent/CN103616021A/en
Publication of CN103616021A publication Critical patent/CN103616021A/en
Pending legal-status Critical Current

Links

Images

Classifications

    • GPHYSICS
    • G01MEASURING; TESTING
    • G01CMEASURING DISTANCES, LEVELS OR BEARINGS; SURVEYING; NAVIGATION; GYROSCOPIC INSTRUMENTS; PHOTOGRAMMETRY OR VIDEOGRAMMETRY
    • G01C21/00Navigation; Navigational instruments not provided for in groups G01C1/00 - G01C19/00
    • G01C21/20Instruments for performing navigational calculations

Landscapes

  • Engineering & Computer Science (AREA)
  • Radar, Positioning & Navigation (AREA)
  • Remote Sensing (AREA)
  • Automation & Control Theory (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Manipulator (AREA)

Abstract

The invention provides a global localization method and device. The global localization method comprises the following steps: acquiring the current sensing information of a robot, fusing the sensing information into Gaussian distribution to generate proposal distribution; measuring the proposal distribution, and updating to obtain pose probability distribution of the robot; finally calculating the pose of the robot by utilizing the pose probability distribution of the robot so as to realize global localization of the robot. According to the method, global localization accuracy is improved on the basis that the computation complexity is reduced.

Description

A kind of global localization method and device
Technical field
The application relates to autonomous navigation technology field, particularly relates to a kind of global localization method and device.
Background technology
Robot orientation problem is the key issue in Mobile Robotics Navigation research, and overall situation location is the most basic, the most important function of autonomous robot, and the research of localization for Mobile Robot is had to very important meaning.
Prior art is used MCL(Monte Carlo Localization conventionally, monte carlo localization) realize the overall situation location of robot, MCL is as a kind of Probabilistic Localization Methods, mainly to utilize the sampling of some Weights in state space to represent that the posterior density of robotary distributes, thereby can represent probability distribution arbitrarily, solve the state estimation problem of non-linear skewed distribution, in robot positioning field, many successes have been obtained, and be applied in actual robot system, but there are many limitation in the method: one is if do not have enough particles can cause filter divergence near correct status.Another is that the frequent Fast Convergent of particle only can obtain sub-optimal result.
How to improve accuracy and the validity of wave filter, and prevent that particle from dispersing the degeneration with wave filter is the problem that researcher is concerned about always.In order to prevent that wave filter from degenerating, frequent application sample importance resampling, but in this, method often causes the tcam-exhaustion of particle.In order to make particle better represent posterior probability density, Trun has proposed mixing MCL, and the method based on adaptively sampled, although these methods can improve the validity of wave filter, has also increased the weight of computation burden.
Therefore, need a kind of global localization method and device badly, to realize in overall position fixing process, reducing on the basis of computation complexity, improve the accuracy of overall situation location.
Summary of the invention
In view of this, the embodiment of the present application provides a kind of global localization method and device, to realize in overall position fixing process, reducing on the basis of computation complexity, improves the accuracy of overall situation location.
To achieve these goals, the technical scheme that the embodiment of the present application provides is as follows:
A global localization method, comprising:
Obtain the current perception information of robot, described perception information is fused to proposing offers in Gaussian distribution and distributes;
Described proposal distribution is measured to upgrade and obtain robot pose probability distribution;
Utilize described robot pose probability distribution to calculate the pose of described robot.
Preferably, described in obtain the current perception information of robot, described perception information is fused in Gaussian distribution to proposing offers and distributes, comprising:
The prior density setting in advance, motion artifacts and noise-aware are represented by gauss hybrid models GMM;
Utilize central difference particle filter, the prior density after representing by GMM, motion artifacts and noise-aware calculate predicted state density;
Obtain the current perception information of robot;
Described perception information is fused to proposing offers in described predicted state density to distribute.
Preferably, described proposal distribution is measured to renewal and obtains robot pose probability distribution, comprising:
In described proposal distribution, extract the primary collection corresponding with the extraction conditions setting in advance;
Calculate respectively described primary and concentrate the weight of each particle;
The weight of each particle that described primary is concentrated is carried out normalization;
The clustering algorithm of utilization based on K dimension, calculates the G-model GM M that is applicable to weighting particle collection;
Utilize described GMM to calculate the statistical property of described weighting particle collection, obtain robot pose probability distribution.
Preferably, also comprise: by self-adaption cluster, select GMM number.
Preferably, also comprise: judge whether to finish overall situation location.
An overall locating device, comprising: proposal distribution generation unit, measurement updating block and pose computing unit, wherein,
Described proposal distribution generation unit, for obtaining the current perception information of robot, is fused to proposing offers in Gaussian distribution by described perception information and distributes;
Described measurement updating block is connected with described proposal distribution generation unit, for described proposal distribution being measured upgrade, obtains robot pose probability distribution;
Described pose computing unit is connected with described measurement updating block, for utilizing described robot pose probability distribution to calculate the pose of described robot.
Preferably, described proposal distribution generation unit comprises: Gaussian distribution represents that unit, predicted state density calculation unit, perception information acquiring unit and proposal distribution generate subelement, wherein,
Described Gaussian distribution represents that unit is used for the prior density setting in advance, motion artifacts and noise-aware to represent by gauss hybrid models GMM;
Described predicted state density calculation unit represents that with described Gaussian distribution unit is connected, and prior density, motion artifacts and noise-aware for utilizing central difference particle filter, after representing by GMM are calculated predicted state density;
Described perception information acquiring unit is for obtaining the current perception information of robot;
One end that described proposal distribution generates subelement is connected with described predicted state density calculation unit, and the other end is connected with described perception information acquiring unit, for described perception information being fused to described predicted state density proposing offers, distributes.
Preferably, described measurement updating block comprises: extraction unit, the first computing unit, normalization unit, the second computing unit and pose probability distribution computing unit, wherein,
Described extraction unit is connected with described proposal distribution generation unit, for extracting the primary collection corresponding with the extraction conditions setting in advance at described proposal distribution;
Described the first computing unit is connected with described extraction unit, concentrates the weight of each particle for calculating respectively described primary;
Described normalization unit is connected with described the first computing unit, for the weight of each concentrated particle of described primary is carried out to normalization;
Described the second computing unit is connected with described normalization unit, for utilizing the clustering algorithm based on K dimension, calculates the G-model GM M that is applicable to weighting particle collection;
Described pose probability distribution computing unit is connected with described the second computing unit, for utilizing described GMM to calculate the statistical property of described weighting particle collection, obtains robot pose probability distribution.
Preferably, also comprise: selected cell, one end of described selected cell is connected with described the second computing unit, and the other end is connected with described pose probability distribution computing unit, for select GMM number by self-adaption cluster.
Preferably, also comprise: judging unit, one end of described judging unit is connected with described pose computing unit, and the other end is connected with described measurement updating block, for judging whether to finish overall situation location.
The application provides a kind of global localization method and device, by obtaining the current perception information of robot, perception information is fused to proposing offers in Gaussian distribution to distribute, then proposal distribution is measured to upgrade and obtain robot pose probability distribution, finally utilize the pose probability distribution calculating robot's of robot pose, to realize the overall situation location to robot, guaranteed, reducing on the basis of computation complexity, to improve the accuracy of overall situation location.
Accompanying drawing explanation
In order to be illustrated more clearly in the embodiment of the present application or technical scheme of the prior art, to the accompanying drawing of required use in embodiment or description of the Prior Art be briefly described below, apparently, the accompanying drawing the following describes is only some embodiment that record in the application, for those of ordinary skills, do not paying under the prerequisite of creative work, can also obtain according to these accompanying drawings other accompanying drawing.
A kind of global localization method process flow diagram that Fig. 1 provides for the embodiment of the present application one;
A kind of overall positioning device structure schematic diagram that Fig. 2 provides for the embodiment of the present application two;
A kind of proposal distribution generation unit structural representation that Fig. 3 provides for the embodiment of the present application two;
A kind of measurement updating block structural representation that Fig. 4 provides for the embodiment of the present application two.
Embodiment
In order to make those skilled in the art person understand better the technical scheme in the application, below in conjunction with the accompanying drawing in the embodiment of the present application, technical scheme in the embodiment of the present application is clearly and completely described, obviously, described embodiment is only the application's part embodiment, rather than whole embodiment.Embodiment based in the application, those of ordinary skills are not making the every other embodiment obtaining under creative work prerequisite, all should belong to the scope of the application's protection.
Embodiment mono-:
A kind of global localization method process flow diagram that Fig. 1 provides for the embodiment of the present application one.
As shown in Figure 1, the method comprises:
S101, obtain the current perception information of robot, perception information is fused to proposing offers in Gaussian distribution and distributes.
In the embodiment of the present application, this process is mainly: first the prior density setting in advance, motion artifacts and noise-aware are passed through to GMM (Gaussian mixture model, gauss hybrid models) represent, then utilize central difference particle filter, the prior density after representing by GMM, motion artifacts and noise-aware calculate predicted state density, and obtain the current perception information of robot, finally perception information is fused to proposing offers in predicted state density and distributes, the embodiment of this process that the application provides is as follows:
In the embodiment of the present application, first obtain the current perception information z of robot, at moment t-1 hypothesis prior density p (x t-1| z t-1), motion artifacts p (ε t-1) and noise-aware p (η t) known, and represent with GMM:
p ~ ( x t - 1 | z t - 1 ) = Σ g = 1 G α t - 1 g N ( μ t - 1 g , P t - 1 g ) , p ~ ( ϵ t - 1 ) = Σ i = 1 I β t - 1 i N ( μ ϵ t - 1 i , Q t - 1 i ) ,
p ~ ( η t ) = Σ j = 1 J γ t j N ( μ η t - 1 i , R t i )
"=g '+(j-1) GI calculates p (x according to step below with Central Difference Filter now for definition g '=g+ (i-1) G, g t| z t-1) and p (x t| z t) GMM approximate:
Figure BDA0000430126110000053
Now, predicted state density can be approximately with GMM:
p ~ ( x t | z t - 1 ) = Σ g ′ = 1 G ′ α t g ′ N ( μ ~ t g ′ , P ~ t g ′ )
Proposal distribution q (x t| z t) posteriority state density with GMM, be approximately:
q ( x t | z t ) = p ~ ( x t | z t ) = Σ g ′ ′ - 1 G ′ ′ α t g ′ ′ N ( μ t g ′ ′ , P t g ′ ′ )
The embodiment of the present application, at moment t-1 hypothesis prior density p (x t-1| z t-1), motion artifacts p (ε t-1) and noise-aware p (η t) in known situation, first by the motion of Central Difference Filter, upgrade and draw predicted state density p (x t| z t-1), then by the perception of Central Difference Filter, upgrade and draw proposal distribution p (x t| z t), and proposal distribution p (x t| z t) representing t constantly, robot perception is the probability distribution of the position and posture x of z Shi, robot.
S102, proposal distribution is measured to upgrade obtain robot pose probability distribution.
In the embodiment of the present application, this process is mainly: in proposal distribution, extract the primary collection corresponding with the extraction conditions setting in advance, calculate respectively primary and concentrate the weight of each particle, the weight of each particle of then primary being concentrated is carried out normalization, and the clustering algorithm of utilization based on K dimension, calculate the G-model GM M that is applicable to weighting particle collection, finally utilize GMM to calculate the statistical property of weighting particle collection, obtain robot pose probability distribution, the embodiment of this process that the application provides is as follows:
First, from proposal distribution q (x t| z t) extracting N particle as primary collection, this primary collection is { x t ( i ) ; i = 1 . . . N } .
Then, calculate the weight that this primary is concentrated each particle.
w ~ t ( i ) = p ( z t | χ t ( i ) ) p ~ ( χ t ( i ) | z t - 1 ) / q ( χ t ( i ) | z t )
Normalization weight: w t ( i ) = w ~ t ( i ) / Σ i = 1 N w ~ t ( i )
Use the clustering algorithm based on K Wei Shu to substitute resampling, find and be applicable to weighting particle collection
Figure BDA0000430126110000064
g-model GM M.Pattern number G is adaptively selected and calculate robot pose probability distribution by clustering algorithm, and this process is mainly as follows:
On particle collection, build K Wei Shu, each node of tree is a subset of particle collection.Root node has all particles, and each non-leaf node has two children, and child is according to value Node splitvaldivide father.
&chi; i &Element; Node left &DoubleLeftRightArrow; &chi; i < Node splitval , And χ i∈ Node parent
&chi; i &Element; Node right &DoubleLeftRightArrow; &chi; i &GreaterEqual; Node splitval , And χ i∈ Node parent
If a node can not divide again, just think leaf.Creating in the process of tree from top to bottom, by beta pruning, determine whether certain node is leaf, and needn't search for all child nodes, if GMM is M gany g-model meet following condition with regard to beta pruning:
max(w ig)-min(w ig)<τ(∑w ig+Node numparitcle×min(w ig))
Wherein, w ig=p (M g| χ i, θ) represent model M ghave particle χ iprobability:
w ig = p ( M g | &chi; i , &theta; ) = a ig c g / &Sigma; k = 1 G a ik c k
a ig = p ( &chi; i | M g , &theta; ) &ap; ( 2 &pi; | | P g | | ) - 1 / 2 exp ( - 1 2 ( &chi; i - &mu; g ) T P g - 1 ( &chi; i - &mu; g ) )
c g=p(M g|θ),θ=(c 1,...,c G1,...,μ G;P 1,...,P G)
Wherein, θ represents all parameters of GMM, μ gand P git is respectively model M gmean value and covariance.
K Wei Shu is carried out to EM (Expectation Maximization, expectation is maximum) circulation, the root call function MakeStates () of tree, MakeStates (Node, θ t) return to 3G value:
Sw g = &Sigma; &chi; i &Element; Node w ig , Sw&chi; g = &Sigma; &chi; i &Element; Node w ig &chi; i , Sw&chi;&chi; g = &Sigma; &chi; i &Element; Node w ig &chi; i &chi; i T
The result of MakeStates (Root) is used for building parameter θ t+1:
c g = Sw g / N , &mu; g = Sx &chi; g / Sw g , P g = ( Sw&chi; &chi; g / Sw g ) - &mu; g &mu; g T
If leaf node calls MakeStates (), to each M gsimple computation is:
w &OverBar; g = p ( M g | x &OverBar; , &theta; t ) = p ( &chi; &OverBar; | M g , &theta; t ) p ( M g | &theta; t ) / &Sigma; k = 1 G p ( &chi; &OverBar; | M k , &theta; t ) p ( M k | &theta; t )
Wherein, return so
Sw g = w &OverBar; g &times; Node NumParticles , Sw&chi; g = w &OverBar; g &times; Node NumParticles &times; &chi; &OverBar; ,
Sw&chi;&chi; g = w &OverBar; g &times; Node NumParticles &times; Node COV
If nonleaf node calls MakeStates (), its two child nodes recursive call MakeStates (), finally return to the result set of two child nodes.
Now, we can distribute with the approximate state posteriority of the G-GMM upgrading:
p ~ ( x t | z t ) = &Sigma; g = 1 G &alpha; t g N ( &mu; t g , P t g )
S103, utilize the pose probability distribution calculating robot's of robot pose.
In the embodiment of the present application, this process is as follows:
State estimation mean value
Figure BDA00004301261100000710
with corresponding error covariance
Figure BDA00004301261100000711
before weighting EM cluster, be calculated as:
x ^ t = &Sigma; i = 1 N w t ( i ) &chi; t ( i ) , P ^ t = &Sigma; i = 1 N w t ( i ) ( &chi; t - x ^ t ) ( &chi; t - x ^ t ) T
After GMM is approximate, be calculated as:
x ^ t = &Sigma; g = 1 G &alpha; t ( g ) &chi; g ( g ) , P ^ t = &Sigma; g = 1 G &alpha; t ( g ) ( P t ( g ) + ( &mu; t ( g ) - x ^ t ) ( &mu; t ( g ) - x ^ t ) T )
Further, the method also comprises: by self-adaption cluster, select GMM number.
Further, the method also comprises: judge whether to finish overall situation location.
In the embodiment of the present application, after utilizing the pose probability distribution calculating robot's of robot pose, also can judge whether to finish overall situation location, when being, export the pose of this robot, when no, return to execution step S102.
The application provides a kind of global localization method, by obtaining the current perception information of robot, perception information is fused to proposing offers in Gaussian distribution to distribute, then proposal distribution is measured to upgrade and obtain robot pose probability distribution, finally utilize the pose probability distribution calculating robot's of robot pose, to realize the overall situation location to robot, guaranteed, reducing on the basis of computation complexity, to improve the accuracy of overall situation location.
Embodiment bis-:
A kind of overall positioning device structure schematic diagram that Fig. 2 provides for the embodiment of the present application two.
As shown in Figure 2, this device comprises: proposal distribution generation unit 1, measurement updating block 2 and pose computing unit 3, wherein,
Proposal distribution generation unit 1, for obtaining the current perception information of robot, is fused to proposing offers in Gaussian distribution by perception information and distributes.
Measure updating block 2 and be connected with proposal distribution generation unit 1, for proposal distribution being measured upgrade, obtain robot pose probability distribution.
Pose computing unit 3 is connected with measurement updating block 2, for utilizing the pose probability distribution calculating robot's of robot pose.
A kind of proposal distribution generation unit structural representation that Fig. 3 provides for the embodiment of the present application two.
As shown in Figure 3, this proposal distribution generation unit comprises: Gaussian distribution represents that unit 11, predicted state density calculation unit 12, perception information acquiring unit 13 and proposal distribution generate subelement 14.
Wherein, Gaussian distribution represents that unit 11 is for representing the prior density setting in advance, motion artifacts and noise-aware by gauss hybrid models GMM.
Predicted state density calculation unit 12 represents that with Gaussian distribution unit 11 is connected, and prior density, motion artifacts and noise-aware for utilizing central difference particle filter, after representing by GMM are calculated predicted state density.
Perception information acquiring unit 13 is for obtaining the current perception information of robot.
One end that proposal distribution generates subelement 14 is connected with predicted state density calculation unit 12, and the other end is connected with perception information acquiring unit 13, for perception information being fused to predicted state density proposing offers, distributes.
A kind of measurement updating block structural representation that Fig. 4 provides for the embodiment of the present application two.
As shown in Figure 4, this measurement updating block comprises: extraction unit 21, the first computing unit 22, normalization unit 23, the second computing unit 24 and pose probability distribution computing unit 25, wherein,
Extraction unit 21 is connected with proposal distribution generation unit 1, for extracting the primary collection corresponding with the extraction conditions setting in advance at proposal distribution.
The first computing unit 22 is connected with extraction unit 21, concentrates the weight of each particle for calculating respectively primary.
Normalization unit 23 is connected with the first computing unit 22, for the weight of each concentrated particle of primary is carried out to normalization.
The second computing unit 24 is connected with normalization unit 23, for utilizing the clustering algorithm based on K dimension, calculates the G-model GM M that is applicable to weighting particle collection.
Pose probability distribution computing unit 25 is connected with the second computing unit 24, for utilizing GMM to calculate the statistical property of weighting particle collection, obtains robot pose probability distribution.
Further, the overall locating device that the embodiment of the present application provides also comprises: selected cell, and one end of selected cell is connected with the second computing unit, and the other end is connected with pose probability distribution computing unit, for select GMM number by self-adaption cluster.
Further, the overall locating device that the embodiment of the present application provides also comprises: judging unit, and one end of judging unit is connected with pose computing unit, and the other end is connected with measurement updating block, for judging whether to finish overall situation location.
The application provides a kind of overall locating device, comprise: proposal distribution generation unit, measure updating block and pose computing unit, by proposal distribution generation unit, obtain the current perception information of robot, perception information is fused to proposing offers in Gaussian distribution to distribute, then by measurement updating block, proposal distribution is measured to upgrade and obtain robot pose probability distribution, finally by pose computing unit, utilize the pose probability distribution calculating robot's of robot pose, to realize the overall situation location to robot, guaranteed reducing on the basis of computation complexity, improve the accuracy of overall situation location.
In this instructions, each embodiment adopts the mode of going forward one by one to describe, and each embodiment stresses is the difference with other embodiment, between each embodiment identical similar part mutually referring to.For the disclosed device of embodiment, because it corresponds to the method disclosed in Example, so description is fairly simple, relevant part partly illustrates referring to method.
Below be only the application's preferred implementation, make those skilled in the art can understand or realize the application.To the multiple modification of these embodiment, will be apparent to one skilled in the art, General Principle as defined herein can be in the situation that do not depart from the application's spirit or scope, realization in other embodiments.Therefore, the application will can not be restricted to these embodiment shown in this article, but will meet the widest scope consistent with principle disclosed herein and features of novelty.

Claims (10)

1. a global localization method, is characterized in that, comprising:
Obtain the current perception information of robot, described perception information is fused to proposing offers in Gaussian distribution and distributes;
Described proposal distribution is measured to upgrade and obtain robot pose probability distribution;
Utilize described robot pose probability distribution to calculate the pose of described robot.
2. method according to claim 1, is characterized in that, described in obtain the current perception information of robot, described perception information is fused in Gaussian distribution to proposing offers and distributes, comprising:
The prior density setting in advance, motion artifacts and noise-aware are represented by gauss hybrid models GMM;
Utilize central difference particle filter, the prior density after representing by GMM, motion artifacts and noise-aware calculate predicted state density;
Obtain the current perception information of robot;
Described perception information is fused to proposing offers in described predicted state density to distribute.
3. method according to claim 1, is characterized in that, described proposal distribution is measured to upgrade obtain robot pose probability distribution, comprising:
In described proposal distribution, extract the primary collection corresponding with the extraction conditions setting in advance;
Calculate respectively described primary and concentrate the weight of each particle;
The weight of each particle that described primary is concentrated is carried out normalization;
The clustering algorithm of utilization based on K dimension, calculates the G-model GM M that is applicable to weighting particle collection;
Utilize described GMM to calculate the statistical property of described weighting particle collection, obtain robot pose probability distribution.
4. method according to claim 3, is characterized in that, also comprises: by self-adaption cluster, select GMM number.
5. method according to claim 1, is characterized in that, also comprises: judge whether to finish overall situation location.
6. an overall locating device, is characterized in that, comprising: proposal distribution generation unit, measurement updating block and pose computing unit, wherein,
Described proposal distribution generation unit, for obtaining the current perception information of robot, is fused to proposing offers in Gaussian distribution by described perception information and distributes;
Described measurement updating block is connected with described proposal distribution generation unit, for described proposal distribution being measured upgrade, obtains robot pose probability distribution;
Described pose computing unit is connected with described measurement updating block, for utilizing described robot pose probability distribution to calculate the pose of described robot.
7. device according to claim 6, is characterized in that, described proposal distribution generation unit comprises: Gaussian distribution represents that unit, predicted state density calculation unit, perception information acquiring unit and proposal distribution generate subelement, wherein,
Described Gaussian distribution represents that unit is used for the prior density setting in advance, motion artifacts and noise-aware to represent by gauss hybrid models GMM;
Described predicted state density calculation unit represents that with described Gaussian distribution unit is connected, and prior density, motion artifacts and noise-aware for utilizing central difference particle filter, after representing by GMM are calculated predicted state density;
Described perception information acquiring unit is for obtaining the current perception information of robot;
One end that described proposal distribution generates subelement is connected with described predicted state density calculation unit, and the other end is connected with described perception information acquiring unit, for described perception information being fused to described predicted state density proposing offers, distributes.
8. device according to claim 6, is characterized in that, described measurement updating block comprises: extraction unit, the first computing unit, normalization unit, the second computing unit and pose probability distribution computing unit, wherein,
Described extraction unit is connected with described proposal distribution generation unit, for extracting the primary collection corresponding with the extraction conditions setting in advance at described proposal distribution;
Described the first computing unit is connected with described extraction unit, concentrates the weight of each particle for calculating respectively described primary;
Described normalization unit is connected with described the first computing unit, for the weight of each concentrated particle of described primary is carried out to normalization;
Described the second computing unit is connected with described normalization unit, for utilizing the clustering algorithm based on K dimension, calculates the G-model GM M that is applicable to weighting particle collection;
Described pose probability distribution computing unit is connected with described the second computing unit, for utilizing described GMM to calculate the statistical property of described weighting particle collection, obtains robot pose probability distribution.
9. device according to claim 8, it is characterized in that, also comprise: selected cell, one end of described selected cell is connected with described the second computing unit, the other end is connected with described pose probability distribution computing unit, for select GMM number by self-adaption cluster.
10. device according to claim 6, is characterized in that, also comprises: judging unit, and one end of described judging unit is connected with described pose computing unit, and the other end is connected with described measurement updating block, for judging whether to finish overall situation location.
CN201310649076.2A 2013-12-04 2013-12-04 Global localization method and device Pending CN103616021A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201310649076.2A CN103616021A (en) 2013-12-04 2013-12-04 Global localization method and device

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201310649076.2A CN103616021A (en) 2013-12-04 2013-12-04 Global localization method and device

Publications (1)

Publication Number Publication Date
CN103616021A true CN103616021A (en) 2014-03-05

Family

ID=50166728

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201310649076.2A Pending CN103616021A (en) 2013-12-04 2013-12-04 Global localization method and device

Country Status (1)

Country Link
CN (1) CN103616021A (en)

Cited By (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN105333879A (en) * 2015-12-14 2016-02-17 重庆邮电大学 Synchronous positioning and map building method
CN106052693A (en) * 2016-05-23 2016-10-26 上海钛米机器人科技有限公司 Service robot indoor global localization method
WO2018119607A1 (en) * 2016-12-26 2018-07-05 Bayerische Motoren Werke Aktiengesellschaft Method and apparatus for uncertainty modeling of point cloud
CN109506641A (en) * 2017-09-14 2019-03-22 深圳乐动机器人有限公司 The pose loss detection and relocation system and robot of mobile robot
CN110567441A (en) * 2019-07-29 2019-12-13 广东星舆科技有限公司 Particle filter-based positioning method, positioning device, mapping and positioning method
CN113008245A (en) * 2019-12-20 2021-06-22 北京图森智途科技有限公司 Positioning information fusion method and device and computer server

Cited By (11)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN105333879A (en) * 2015-12-14 2016-02-17 重庆邮电大学 Synchronous positioning and map building method
CN105333879B (en) * 2015-12-14 2017-11-07 重庆邮电大学 Synchronous superposition method
CN106052693A (en) * 2016-05-23 2016-10-26 上海钛米机器人科技有限公司 Service robot indoor global localization method
CN106052693B (en) * 2016-05-23 2019-05-24 上海钛米机器人科技有限公司 Global localization method in service robot room
WO2018119607A1 (en) * 2016-12-26 2018-07-05 Bayerische Motoren Werke Aktiengesellschaft Method and apparatus for uncertainty modeling of point cloud
US10740962B2 (en) 2016-12-26 2020-08-11 Bayerische Motoren Werke Aktiengesellschaft Method and apparatus for uncertainty modeling of point cloud
CN109506641A (en) * 2017-09-14 2019-03-22 深圳乐动机器人有限公司 The pose loss detection and relocation system and robot of mobile robot
CN110567441A (en) * 2019-07-29 2019-12-13 广东星舆科技有限公司 Particle filter-based positioning method, positioning device, mapping and positioning method
CN110567441B (en) * 2019-07-29 2021-09-28 广东星舆科技有限公司 Particle filter-based positioning method, positioning device, mapping and positioning method
CN113008245A (en) * 2019-12-20 2021-06-22 北京图森智途科技有限公司 Positioning information fusion method and device and computer server
CN113008245B (en) * 2019-12-20 2022-12-27 北京图森智途科技有限公司 Positioning information fusion method and device and computer server

Similar Documents

Publication Publication Date Title
CN103616021A (en) Global localization method and device
Feppon et al. A geometric approach to dynamical model order reduction
Zhong et al. A cross-entropy method and probabilistic sensitivity analysis framework for calibrating microscopic traffic models
CN103326903B (en) Based on the Internet network latency prediction method of Hidden Markov
CN103034778B (en) Be applicable to the individual brain function network extraction method of how tested brain function data analysis
CN102567719B (en) Human age automatic estimation method based on posterior probability neural network
WO2022077767A1 (en) Traffic flow prediction method and apparatus, computer device, and readable storage medium
CN103902812B (en) A kind of particle filter method, device and method for tracking target, device
CN104778508A (en) Public bicycle renting forecasting method based on multi-source data fusion
Liu et al. Simulation optimization based on Taylor Kriging and evolutionary algorithm
Harlim Model error in data assimilation
CN105354860A (en) Box particle filtering based extension target CBMeMBer tracking method
CN104048676A (en) MEMS (Micro Electro Mechanical System) gyroscope random error compensating method based on improved particle filter
Yuan et al. Calibration of stochastic computer models using stochastic approximation methods
Wang et al. Fusing heterogeneous traffic data by Kalman filters and Gaussian mixture models
Leifsson et al. Aerodynamic design optimization: physics-based surrogate approaches for airfoil and wing design
CN104795063A (en) Acoustic model building method based on nonlinear manifold structure of acoustic space
Billé Computational Issues in the Estima tion of the Spatial Probit Model: A Comparison of Various Estimators
Tanskanen et al. Non-linearities in Gaussian processes with integral observations
Jin et al. An efficient variational Bayesian algorithm for calibrating fundamental diagrams and its probabilistic sensitivity analysis
Zhang et al. Offline state estimation for hybrid systems via nonsmooth variable projection
CN109712389B (en) Path travel time estimation method based on Copula and Monte-Carlo simulation
Chen et al. Research on adaptive Monte Carlo location method based on fusion posture estimation
Busatto et al. Inference of multiple high-dimensional networks with the Graphical Horseshoe prior
JP6927161B2 (en) Learning devices, predictors, methods, and programs

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
C10 Entry into substantive examination
SE01 Entry into force of request for substantive examination
RJ01 Rejection of invention patent application after publication
RJ01 Rejection of invention patent application after publication

Application publication date: 20140305