CN110068330A - Autonomic positioning method based on the robot that arma modeling is realized - Google Patents
Autonomic positioning method based on the robot that arma modeling is realized Download PDFInfo
- Publication number
- CN110068330A CN110068330A CN201910041274.8A CN201910041274A CN110068330A CN 110068330 A CN110068330 A CN 110068330A CN 201910041274 A CN201910041274 A CN 201910041274A CN 110068330 A CN110068330 A CN 110068330A
- Authority
- CN
- China
- Prior art keywords
- map
- parameter
- indicate
- model
- grid
- 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.)
- Granted
Links
Classifications
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01C—MEASURING DISTANCES, LEVELS OR BEARINGS; SURVEYING; NAVIGATION; GYROSCOPIC INSTRUMENTS; PHOTOGRAMMETRY OR VIDEOGRAMMETRY
- G01C21/00—Navigation; Navigational instruments not provided for in groups G01C1/00 - G01C19/00
- G01C21/20—Instruments 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)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The present invention provides a kind of methods of the autonomous positioning of robot based on arma modeling, carry out parameter Estimation by data under autoregressive moving-average model (ARMA) to map change in long term environment, state, and finally establish the environment variation, the cartographic model of Probabilistic Cell form；Subsequent prediction is carried out in the grid time-domain of the cartographic model of foundation, and forecast confidence is assessed, the grating map after obtaining the prediction with forecast confidence label；Real-time observation information is finally fused to by the grating map after prediction using bayes rule, is obtained as the available updated environmental map of robot, robot realizes the autonomous long-term tillage in the environment of the variation under the map, in conjunction with particle filter algorithm.To achieve the effect that there is higher positioning accuracy and positioning robustness in the environment of robot change in long term.
Description
Technical field
The present invention relates to robotic technology fields, and in particular, to it is a kind of based on arma modeling realize robot in length
Autonomic positioning method under phase changing environment.
Background technique
Orientation problem is always the core content of autonomous mobile robot area research, and the performance for positioning level also will be very big
Influence to degree the performance of the task executions such as robot path planning, independent navigation.In static environment, classics can be used
Particle filter algorithm establishes static probability grating map, and laser radar and odometer is used to carry out as sensor on this basis
Autonomous positioning.But in many actual scenes of reality, environment is frequently not unalterable, but by artificial daily row
Dynamic influence is being changed with certain rule.At this moment classical particle filter algorithm due to establish static map it
On, often fail because environment changes the inaccuracy for leading to priori cartographic information.Such as the vehicle passed in and out of parking lot and
Up and-down people, the cargo etc. for being picked and placed and being carried in industrial plant, will influence particle filter by the variation of environment
Locating effect.
There are the algorithm positioned based on dynamic positioning ability matrix in foreign countries, and in flows of the people such as subway station, campus dining rooms
It is tested under biggish dynamic environment, demonstrates the validity of the algorithm.But above method is still based on robot
Local environment will be static constant it is assumed that this hypothesis will be so that such methods will be no longer applicable under the situation that environment changes for a long time.
It is positioned for there is the mode that provisional map and static map combine, when environmental information and cartographic information
When mismatch, understand according to Current observation information creating provisional map, and the foundation as position matching；And when environmental information and ground
When figure information matches, provisional map can be released, and carry out position matching using static map.Although the method for this modeling is examined
Considered environmental change to influence caused by positioning and be modified by provisional map, but when environment recurring structure and
When wide variation, static map will be entirely ineffective, the long-term long-term Shandong that cannot be guaranteed location algorithm by provisional map positioning
Stick.
It is matched for using the model of dynamic probability grating map, and with the calculating of the matching way of scan-matching
Point, real-time map rejuvenation is carried out on the basis of static map.When matching score is greater than given threshold and detects that environment becomes
After change, the robustness of positioning is improved to obtain the map more identical with current true environment with regard to carrying out map rejuvenation.It should
There is also identical problems for algorithm, when environmental change differs larger with preparatory static map, observation information and cartographic information
Matching degree is too poor, to be unable to get the current location information feedback of robot, the update of static map is caused to be failed.
Foreign scholar thinks that many daily routines of the mankind with periodically, have certain rule that can follow, by unknown ring
Border change procedure is defined as periodic function, carries out modeling statement using change in time and space of the frequency spectrum to environment.Time domain is turned simultaneously
Turn to frequency domain, can efficiently discrimination analysis and save regularity process of environmental change.In this way robot can after model foundation,
There is a prediction to the state of environment, so that robot be helped to realize long-term positioning.The disadvantages of the method are as follows from method level
For, it is necessary to assume that environmental characteristic has stronger periodicity, and modeling process is relative complex.
Work based on forefathers, the present invention innovatively propose it is a kind of can be in the environment of change in long term, by ring
Border modeling carries out map prediction and updates, to realize the algorithm of robot long-term tillage.
Summary of the invention
For the defects in the prior art, the object of the present invention is to provide a kind of robots realized based on arma modeling
Autonomic positioning method.
A kind of autonomic positioning method of the robot realized based on arma modeling provided according to the present invention, including following step
It is rapid:
Step 1: environmental modeling step: acquiring map datum using the self-contained sensor of robot, pass through autoregression
Diagram data carries out parameter Estimation and modeling to moving average model ARMA over the ground, obtains cartographic model.
Step 2: map prediction steps: prediction of the grid in time-domain is carried out using cartographic model, to forecast confidence
It is assessed, obtains the grating map with forecast confidence label.
Step 3: map rejuvenation step:, will be real-time using bayes rule for the deviation of grating map and practical map
Observation information be fused in predictive information, carry out real-time map rejuvenation, obtain update map.
Step 4: based on map is updated, the autonomous long-term tillage of robot autonomous positioning step: is realized using particle filter.
Preferably, the environmental modeling step includes:
Model order step: determining the first parameter p and the second parameter q of arma modeling according to environmental characteristics, described first
Parameter p characterizes the state relation of physical environment significant condition in time, and the second parameter q characterization map predicted state is missed
Serial correlation of the poor item in time series；
Parametric estimation step: the first parameter p and the second parameter q based on arma modeling estimate third parameter (α, β)
Meter, state relation before and after the third parameter (α, β) characterizes trellis states could in time-domain, wherein α ∈ RpAnd β ∈ Rq。
Preferably, the environmental modeling step further include:
Modeling procedure: the public upper bound P value of selected first parameter p and the second parameter q, i.e. 0≤p≤P and 0≤q≤P；By
Grid sample numberIteration finds out σ2Least-squares estimation, i.e.,
It willIt substitutes intoIt obtains A (0,0), A (0,1), A (1,1) ... A (P, P) then hasWherein i indicates that the i-th row of grid in map datum, j indicate the jth column of grid in map datum, under
Marking T indicates number of samples,Indicate the T sample arranged in the i-th row jth, σ2Indicate the value of least-squares estimation, k table
The row k of grid in pictorial map data,Indicate the least-squares estimation value of the grid sample of row k jth column,It indicates
The mean value of grid sample,Indicate the regression coefficient a of grid sample least-squares estimation,Indicate grid sample minimum two
The regression coefficient b, n for multiplying estimation indicate the sum of time series,Indicate that the AIC of the grid sample of row k jth column is quasi-
Then it is worth, A (p, q) indicates AIC criterion value corresponding to the value of the p that model order determines and q,Indicate row k jth column
The AIC criterion value of grid sample.
Preferably, the parametric estimation step includes:
Fit procedure: grid sample number is utilizedMake high-order autoregression sliding die
The fitting of type AR (p)；
Recursion step: recurrence calculation residual sequenceI.e.
T=P+1, P+2 ... T
Separate regression steps: by residual sequenceAs independent sequence, linear regression model (LRM) is utilized:
Obtaining third parameter indicates as follows:
Subscript t indicates sampling time sequence, t=P+1, P+2 ... T；Indicate state error；Indicate that p moment grid exists
Regression parameter under arma modeling；αiIndicate the regression parameter of arma modeling；βjIndicate that the error of arma modeling slides parameter；Z′
Indicate the transposition of Z,It indicatesTransposition,Indicate the state value of Probabilistic Cell map；
Model replaces step: setting model parameter detects, and carries out model replacement when Parameters variation is greater than given threshold, i.e.,
S=| p1-p2|+|q1-q2| > Sth
It is wherein designated as 1 and 2 two groups of parameter models for respectively representing calculating down；When its order difference or parameter are greater than setting
Think that model changes after value, model parameter before being replaced with "current" model parameter.
Preferably, the map prediction steps include:
Status predication step: prediction of the grid in time-domain is carried out using existing map datum, obtains each grid
In the trellis states could of t moment, whole prediction grating map is obtained.
Confidence level estimation step: through predicted state compared with true observation state, the pre- of each trellis states could is determined
Confidence level is surveyed, assigns initial confidence level, then the true map state obtained by database update before each confidence level estimation
It is compared, obtains final forecast confidence.Final output has the grating map of forecast confidence label.
Preferably, the map rejuvenation step includes:
It introduces step: introducing stationkeeping ability matrix to measure the confidence level at different location；
Normalization step: to stationkeeping ability matrix ask determinant and normalize after obtain m2, m2Map difference position is reacted
Set the size for locating corresponding data reliability；Discrimination standard of the revised matching degree as effective observation data, it may be assumed that
M=λ m1+(1-λ)m2
Wherein, λ is observation weight item, indicates that λ value is bigger to matched Feasible degree is observed, and represents observation matching entire
Shared ratio is bigger in the calculating weight of matching degree；M indicates revised map and observes the matching degree of data, m1Indicate ground
The matching degree of figure and observation data, m2Indicate corresponding data reliability at map different location；When m is greater than the threshold value of setting
Afterwards, i.e. m > mth, then defining the data that observation obtains is effectively, thus using m as authentic data.
Preferably, the map rejuvenation step includes:
Data fusion step: data fusion is carried out using bayes rule, by long-term forecast information and short-term information
It blends, obtains accurate map in real time；The form for using dynamic probability grating map to store as map datum, uses HMMs
As basic model, the confidence level in long-term information is incorporated as weight term, state update is carried out, obtains the state at t+1 moment
Vector, fusion formula are as follows:
Wherein QtFor the map trellis states could exported after fused data, AcIt is state-transition matrix, G is forecast confidence,
X(t)It is predicted state, BzIt is observation information, η is normalization factor, and I is unit matrix；Subscript t indicates time series t, subscript
K indicates the row k of map grid, subscript c representing matrix, and subscript z indicates observation
The autonomous positioning step assesses ring in conjunction with dynamic positioning ability after obtaining the dynamic map of real-time update
The dynamic attribute in map is not updated in border, realizes the autonomous long-term of robot with the particle filter algorithm based on stationkeeping ability
Positioning.
Preferably, environmental modeling step uses off-line form, and map prediction steps and map rejuvenation step are used linear
Formula.
Preferably, model order step is determined using AIC criterion, and parametric estimation step is approached using autoregression.
Compared with prior art, the present invention have it is following the utility model has the advantages that
It can make robot that there is good positioning accuracy and positioning robustness in the environment of change in long term.
Detailed description of the invention
Upon reading the detailed description of non-limiting embodiments with reference to the following drawings, other feature of the invention,
Objects and advantages will become more apparent upon:
Fig. 1 is flow chart of the method for the present invention.
Specific embodiment
The present invention is described in detail combined with specific embodiments below.Following embodiment will be helpful to the technology of this field
Personnel further understand the present invention, but the invention is not limited in any way.It should be pointed out that the ordinary skill of this field
For personnel, without departing from the inventive concept of the premise, several changes and improvements can also be made.These belong to the present invention
Protection scope.
For the deficiency of original technology, innovatively propose herein it is a kind of based on arma modeling realize robot from
Master positioning method, suitable for the robot autonomous localization problem under change in long term environment.
Autonomic positioning method of the robot proposed by the present invention realized based on arma modeling under change in long term environment, it is main
Want the following steps are included:
Step 1: modeling acquires map datum using the self-contained sensor of robot, passes through autoregressive moving average
Diagram data carries out parameter Estimation and modeling to model (ARMA) over the ground.
Step 2: prediction carries out prediction of the grid in time-domain using established model, and to forecast confidence into
Row assessment, final output have the grating map of forecast confidence label.
Step 3: updating, for the deviation for predicting map and practical map caused by some uncertain factors, using shellfish
Real-time observation information is fused in predictive information by this rule of leaf, real-time map rejuvenation is carried out, as particle filtering
Priori map foundation.The autonomous long-term tillage of robot is realized using particle filter algorithm.
Specifically, the modeling process of the step 1 mainly comprises the steps that
Step 1.1: model order will determine the parameter p and q of model before arma modeling foundation.Parameter p characterization be
The state relation of physical environment significant condition in time, parameter q characterization is map predicted state error term in time sequence
Correlation on column.For the i-th row in map, the grid of jth column, modeling process is as follows:
Step 1.1.1 selectes the public upper bound P value of order p and q, i.e. 0≤p, q≤P
Step 1.1.2 is by sample numberIteration finds out σ2Least-squares estimation, i.e.,
Step 1.1.3 willIt substitutes intoIt obtains A (0,0), A (0,1), A (1,1),
... A (P, P) then has
Then p, q are the model order needed at this time.
Step 1.2: parameter Estimation needs after having the parameter p and q of autoregressive moving-average model to its model
Parameter is estimated.Needing exist for determining parameter is α and β, wherein α ∈ Rp, β ∈ Rq, the physical significance of characterization is grid
State in time-domain before and after state relation.Parameter Estimation uses the method that autoregression approaches, specific steps are as follows institute
Show:
Step 1.2.1 is first with grid initial dataMake high-order autoregression
The fitting of gliding model AR (p)
Step 1.2.2 by above-mentioned estimation AR (p) model recurrence calculation residual sequenceI.e.
T=P+1, P+2 ... T
Step 1.2.3 arranges residual errorFor independent sequence, linear regression model (LRM) is utilized:
Wherein t=P+1, P+2 ... T
Indicate as follows to obtain estimation parameter:
Wherein,
In view of the variation of environmental model, need to be arranged model parameter detection, when Parameters variation is carried out greater than certain threshold value
Model replacement, i.e.,
S=| p1-p2|+|q1-q2| > Sth
It is wherein designated as 1 and 2 two groups of parameter models for respectively representing calculating down.When its order difference or parameter are greater than setting
All think that model changes after value, model parameter before being replaced with "current" model parameter.
Specifically, the prediction process of the step 2 mainly comprises the steps that
Step 2.1: status predication carries out prediction of the grid in time-domain using existing map datum, so as to
Each grid is obtained in the trellis states could of t moment, so as to obtain whole prediction grating map.
Step 2.2: confidence level estimation, after the prediction of trellis states could, it is also necessary to assess forecast confidence, be used to
Measure the accuracy of each grid prediction.Through predicted state compared with true observation state, to determine each trellis states could
Forecast confidence, initial confidence level can be all assigned for it before each confidence level estimation, is obtained in conjunction with by database update
To true map state be compared, obtain final forecast confidence.Final output has the grid of forecast confidence label
Lattice map.
Specifically, the renewal process of the step 3 mainly comprises the steps that
Step 3.1: introducing stationkeeping ability matrix to measure the confidence level at the position that do not exist together
Step 3.2: obtaining m to stationkeeping ability Matrix Calculating determinant and after normalizing2, reacted at map different location
The size of corresponding data reliability.Discrimination standard of the revised matching degree as effective observation data, it may be assumed that
M=λ m1+(1-λ)m2
Wherein, λ is observation weight item, and value represents to matched Feasible degree is observed, and the value is bigger, represents observation matching
Shared ratio is bigger in the calculating weight of entire matching degree.Only when m is after the threshold value of setting, i.e. m > mthJust think
Observing obtained data is effectively, thus as authentic data.
Step 3.3 application bayes rule carries out data fusion, and long-term forecast information and short-term information are blended,
Obtain real-time, accurate map.The form for using dynamic probability grating map to store as map datum, and made using HMMs
For its basic model.The confidence level in long-term information is incorporated simultaneously as weight term, is carried out state update, is obtained the t+1 moment
State vector, fusion formula are as follows:
Wherein QtFor the map trellis states could exported after fused data, AcIt is state-transition matrix, G is forecast confidence,
X(t)It is predicted state, BzIt is observation information, η is normalization factor, and I is unit matrix.
After obtaining the dynamic map of real-time update, not update map in Evaluation Environment in conjunction with dynamic positioning ability
In dynamic attribute, with based on stationkeeping ability particle filter algorithm realize robot autonomous long-term tillage.
Specifically, step 1 modeling process uses off-line form, and step 2 predicts process and step 3 renewal process using online
Form, the sensor are laser radar, and the environmental characteristic data are stored in map data base.
Determine rank process to determine using AIC criterion, the method that parameter Estimation uses autoregression to approach, it is pre- used in status predication
Survey expression formula such as following formula
Specifically, only just think when the matching degree of agreement of laser data and map reaches certain threshold value current
Positioning under the observation laser data that obtains be it is effective, just can be by measurement data fusion into predictive information.
Specifically, stationkeeping ability matrix is introduced to measure the confidence level at the position that do not exist together, and is shown below:
Wherein, p=(x, y, θ) is the pose of robot.It is the i in LRF modelthDesired distance in laser beam.
N0It is the total quantity of LRF model laser beam.And for The laser beam of LRF model returns after the mobile Δ x of expression robot
The change of value.
Specifically, cartographic information fusion formula is as follows:
Wherein QtFor the map trellis states could exported after fused data, AcIt is state-transition matrix, G is forecast confidence,
X(t)It is predicted state, BzIt is observation information, η is normalization factor, and I is unit matrix.
Specifically, the coordinate of final robot can be expressed as following formula
In the specific implementation process, as shown in Figure 1, the present invention, which is divided into (1) modeling (2) prediction (3), updates three parts
It carries out, mainly there is following steps:
Step 1: being acquired by the self-contained sensors towards ambient data of robot, the ring in settling time dimension
Border property data base.Parameter Estimation and modeling are carried out by autoregressive moving-average model (ARMA), to the ring on time dimension
The relevance of border significant condition is assessed, and then predicts the change in future and trend of ambient condition.
Step 1.1: model order before the foundation for carrying out arma modeling, first has to parameter p and q it is confirmed that model.Ginseng
Number p characterization is the state relation of physical environment significant condition in time, and parameter q characterization is map predicted state error
Correlation of the item in time series.Determine rank process to be determined using AIC criterion here, for the i-th row in map, jth column
Grid, modeling process is as follows:
Step 1.1.1 selectes the public upper bound P value of order p and q, i.e. 0≤p, q≤P
Step 1.1.2 is by sample numberIteration finds out σ2Least-squares estimation, i.e.,
Step 1.1.3 willIt substitutes intoIt obtains A (0,0), A (0,1), A (1,1),
... A (P, P) then has
Then p, q are the model order needed at this time.
Step 1.2: parameter Estimation needs after having the parameter p and q of autoregressive moving-average model to its model
Parameter is estimated.Needing exist for determining parameter is α and β, wherein α ∈ Rp, β ∈ Rq, the physical significance of characterization is grid
State in time-domain before and after state relation.Parameter Estimation uses the method that autoregression approaches, specific steps are as follows institute
Show:
Step 1.2.1 is first with grid initial dataMake high-order autoregression
The fitting of gliding model AR (p)
Step 1.2.2 by above-mentioned estimation AR (p) model recurrence calculation residual sequenceI.e.
T=P+1, P+2 ... T (3)
Step 1.2.3 arranges residual errorFor independent sequence, linear regression model (LRM) is utilized:
Wherein t=P+1, P+2 ... T
Indicate as follows to obtain estimation parameter:
Wherein,
In view of the variation of environmental model, need to be arranged model parameter detection, when Parameters variation is carried out greater than certain threshold value
Model replacement, i.e.,
S=| p1-p2|+|q1-q2| > Sth
It is wherein designated as 1 and 2 two groups of parameter models for respectively representing calculating down.When its order difference or parameter are greater than setting
All think that model changes after value, model parameter before being replaced with "current" model parameter.
Step 2: carrying out prediction of the grid in time-domain using established model, and forecast confidence is commented
Estimate, final output has the grating map of forecast confidence label.
Step 2.1: status predication carries out prediction of the grid in time-domain using existing map datum, so as to
Each grid is obtained in the trellis states could of t moment, so as to obtain whole prediction grating map.Prediction expression is as follows
Step 2.2: confidence level estimation, after the prediction of trellis states could, it is also necessary to assess forecast confidence, be used to
Measure the accuracy of each grid prediction.Through predicted state compared with true observation state, to determine each trellis states could
Forecast confidence, initial confidence level can be all assigned for it before each confidence level estimation, is obtained in conjunction with by database update
To true map state be compared, obtain final forecast confidence.Final output has the grid of forecast confidence label
Lattice map.The specific implementation process is as follows shown:
Step 3: for the deviation for predicting map and practical map caused by some uncertain factors, using Bayesian Method
Then real-time observation information is fused in predictive information, carries out real-time map rejuvenation, the priori as particle filtering
Map foundation.Specific step is as follows
Step 3.1 introduces stationkeeping ability matrix to measure the confidence level at the position that do not exist together, as follows:
Wherein, p=(x, y, θ) is the pose of robot.It is the i in LRF modelthDesired distance in laser beam.
N0It is the total quantity of LRF model laser beam.And for The laser beam of LRF model returns after the mobile Δ x of expression robot
The change of value.
Step 3.2 obtains m to stationkeeping ability Matrix Calculating determinant and after normalizing2, reacted at map different location
The size of corresponding data reliability.Work as m2When larger, i.e., cartographic information feature rich, data reliability are high；And work as m2It is smaller
When, i.e. cartographic information feature is few, and data reliability is low.Discrimination standard of the revised matching degree as effective observation data, it may be assumed that
M=λ m1+(1-λ)m2 (9)
Wherein, λ is observation weight item, and value represents to matched Feasible degree is observed, and the value is bigger, represents observation matching
Shared ratio is bigger in the calculating weight of entire matching degree.Only when m is after the threshold value of setting, i.e. m > mthJust think
Observing obtained data is effectively, thus as authentic data.
Step 3.3 application bayes rule carries out data fusion, and long-term forecast information and short-term information are blended,
Obtain real-time, accurate map.The form for using dynamic probability grating map to store as map datum, and made using HMMs
For its basic model.The confidence level in long-term information is incorporated simultaneously as weight term, is carried out state update, is obtained the t+1 moment
State vector, fusion formula are as follows:
Wherein QtFor the map trellis states could exported after fused data, AcIt is state-transition matrix, G is forecast confidence,
X(t)It is predicted state, BzIt is observation information, η is normalization factor, and I is unit matrix.
Final form of the state as map of prediction will be relied more heavily in the higher grid of confidence level based on this；
And grid lower for confidence level, observation data in real time will be relied more heavily on and be iterated update.It is more accorded with this
Close the map of true environment.
Step 4: after obtaining the dynamic map of real-time update, coming in Evaluation Environment not update in conjunction with dynamic positioning ability
Dynamic attribute into map realizes the autonomous long-term tillage of robot with the particle filter algorithm based on stationkeeping ability.Specifically
It is accomplished by
The dynamic positioning ability matrix that stationkeeping ability is embodied from initial map is expressed as follows:
Wherein siIt is the probability that laser beam sweeps to unknown barrier.It is observation information variance.P=(x, y, θ) is machine
The pose of people.It is the i in LRF modelthDesired distance in laser beam.N0It is the total quantity of LRF model laser beam.And it is right
In The change of the laser beam return value of LRF model after the mobile Δ x of expression robot.
Observation information is merged by using classical data anastomosing algorithmWith odometer informationTo correct
It is recommended that distribution function.Later, the Δ O and p of each particle can be by Δ O(k)WithReplacement.Therefore, the odometer after merging
Increment will become
Wherein h is scale factor,It can be obtained by classical particle filter algorithm, andIt indicates in only odometer
IncrementUnder counted robot coordinate.
After obtaining modified odometer increment, modified sampling particle can be expressed as
The coordinate of last robot can be expressed as
One skilled in the art will appreciate that in addition to realizing system provided by the invention in a manner of pure computer readable program code
It, completely can be by the way that method and step be carried out programming in logic come so that provided by the invention other than system, device and its modules
System, device and its modules are declined with logic gate, switch, specific integrated circuit, programmable logic controller (PLC) and insertion
The form of controller etc. realizes identical program.So system provided by the invention, device and its modules may be considered that
It is a kind of hardware component, and the knot that the module for realizing various programs for including in it can also be considered as in hardware component
Structure；It can also will be considered as realizing the module of various functions either the software program of implementation method can be Hardware Subdivision again
Structure in part.
Specific embodiments of the present invention are described above.It is to be appreciated that the invention is not limited to above-mentioned
Particular implementation, those skilled in the art can make a variety of changes or modify within the scope of the claims, this not shadow
Ring substantive content of the invention.In the absence of conflict, the feature in embodiments herein and embodiment can any phase
Mutually combination.
Claims (9)
1. a kind of autonomic positioning method of the robot realized based on arma modeling, which comprises the following steps:
Step 1: environmental modeling step: acquiring map datum using the self-contained sensor of robot, slided by autoregression
Diagram data carries out parameter Estimation and modeling to averaging model ARMA over the ground, obtains cartographic model.
Step 2: map prediction steps: carrying out prediction of the grid in time-domain using cartographic model, forecast confidence is carried out
Assessment obtains the grating map with forecast confidence label.
Step 3: map rejuvenation step: for the deviation of grating map and practical map, will be seen in real time using bayes rule
Measurement information is fused in predictive information, carries out real-time map rejuvenation, obtains updating map.
Step 4: based on map is updated, the autonomous long-term tillage of robot autonomous positioning step: is realized using particle filter.
2. the autonomic positioning method of the robot according to claim 1 realized based on arma modeling, which is characterized in that institute
Stating environmental modeling step includes:
Model order step: the first parameter p and the second parameter q, the first parameter p of arma modeling are determined according to environmental characteristics
The state relation of physical environment significant condition in time is characterized, the second parameter q characterizes map predicted state error term
Serial correlation in time series；
Parametric estimation step: the first parameter p and the second parameter q based on arma modeling estimate third parameter (α, β),
State relation before and after the third parameter (α, β) characterizes trellis states could in time-domain, wherein α ∈ RpAnd β ∈ Rq。
3. the autonomic positioning method of the robot according to claim 2 realized based on arma modeling, which is characterized in that institute
State environmental modeling step further include:
Modeling procedure: the public upper bound P value of selected first parameter p and the second parameter q, i.e. 0≤p≤P and 0≤q≤P；By grid
Sample numberIteration finds out σ2Least-squares estimation, i.e.,It willIt substitutes intoIt obtains A (0,0), A (0,1), A (1,1) ... A (P, P) then hasWherein i indicates that the i-th row of grid in map datum, j indicate the jth column of grid in map datum, under
Marking T indicates number of samples,Indicate the T sample arranged in the i-th row jth, σ2Indicate the value of least-squares estimation, k is indicated
The row k of grid in map datum,Indicate the least-squares estimation value of the grid sample of row k jth column,Indicate grid
The mean value of lattice sample,Indicate the regression coefficient a of grid sample least-squares estimation,Indicate grid sample least square
The regression coefficient b, n of estimation indicate the sum of time series,Indicate the AIC criterion of the grid sample of row k jth column
Value, A (p, q) indicate AIC criterion value corresponding to the value of the p that model order determines and q,Indicate the grid of row k jth column
The AIC criterion value of lattice sample.
4. the autonomic positioning method of the robot according to claim 3 realized based on arma modeling, which is characterized in that institute
Stating parametric estimation step includes:
Fit procedure: grid sample number is utilizedMake high-order autoregressive sliding model AR
(p) fitting；
Recursion step: recurrence calculation residual sequenceI.e.
Separate regression steps: by residual sequenceAs independent sequence, linear regression model (LRM) is utilized:
Obtaining third parameter indicates as follows:
Wherein, subscript t indicates sampling time sequence, t=P+1, P+2 ... T；Indicate state error；Indicate p moment grid
Regression parameter under arma modeling；αiIndicate the regression parameter of arma modeling；βjIndicate that the error of arma modeling slides parameter；
The transposition of Z ' expression Z,It indicatesTransposition,Indicate the state value of Probabilistic Cell map；
Model replaces step: setting model parameter detects, and carries out model replacement when Parameters variation is greater than given threshold, i.e.,
S=| p1-p2|+|q1-q2| > Sth
It is wherein designated as 1 and 2 two groups of parameter models for respectively representing calculating down；After its order difference or parameter are greater than the set value
Think that model changes, model parameter before being replaced with "current" model parameter.
5. the autonomic positioning method of the robot according to claim 1 realized based on arma modeling, which is characterized in that institute
Stating map prediction steps includes:
Status predication step: prediction of the grid in time-domain is carried out using existing map datum, obtains each grid in t
The trellis states could at quarter obtains whole prediction grating map.
Confidence level estimation step: through predicted state compared with true observation state, determine that the prediction of each trellis states could is set
Reliability assigns initial confidence level before each confidence level estimation, then is carried out by the true map state that database update obtains
Compare, obtains final forecast confidence.Final output has the grating map of forecast confidence label.
6. the autonomic positioning method of the robot according to claim 1 realized based on arma modeling, which is characterized in that institute
Stating map rejuvenation step includes:
It introduces step: introducing stationkeeping ability matrix to measure the confidence level at different location；
Normalization step: to stationkeeping ability matrix ask determinant and normalize after obtain m2, m2It has reacted at map different location
The size of corresponding data reliability；Discrimination standard of the revised matching degree as effective observation data, it may be assumed that
M=λ m1+(1-λ)m2
Wherein, λ is observation weight item, indicates that λ value is bigger to matched Feasible degree is observed, and represents observation matching and is entirely matching
Shared ratio is bigger in the calculating weight of degree；M indicates revised map and observes the matching degree of data, m1Indicate map with
Observe the matching degree of data, m2Indicate corresponding data reliability at map different location；After m is greater than the threshold value of setting, i.e. m
> mth, then defining the data that observation obtains is effectively, thus using m as authentic data.
7. the autonomic positioning method of the robot according to claim 1 realized based on arma modeling, which is characterized in that institute
Stating map rejuvenation step includes:
Data fusion step: data fusion is carried out using bayes rule, long-term forecast information and short-term information are mutually melted
It closes, obtains accurate map in real time；The form for using dynamic probability grating map to store as map datum, use HMMs as
Basic model incorporates the confidence level in long-term information as weight term, carries out state update, obtain the state vector at t+1 moment,
Fusion formula is as follows:
Wherein QtFor the map trellis states could exported after fused data, AcIt is state-transition matrix, G is forecast confidence, X(t)It is
Predicted state, BzIt is observation information, η is normalization factor, and I is unit matrix；Subscript t indicates that time series t, subscript k are indicated
The row k of map grid, subscript c representing matrix, subscript z indicate observation.
The autonomous positioning step is come in Evaluation Environment after obtaining the dynamic map of real-time update in conjunction with dynamic positioning ability
The dynamic attribute in map is not updated, realizes the autonomous long-term fixed of robot with the particle filter algorithm based on stationkeeping ability
Position.
8. the autonomic positioning method of the robot according to claim 1 realized based on arma modeling, which is characterized in that ring
Border modeling procedure uses off-line form, and map prediction steps and map rejuvenation step use online form.
9. the autonomic positioning method of the robot according to claim 2 realized based on arma modeling, which is characterized in that mould
Type determines step and AIC criterion is used to determine suddenly, and parametric estimation step is approached using autoregression.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910041274.8A CN110068330B (en) | 2019-01-16 | 2019-01-16 | Autonomous positioning method of robot based on ARMA model |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910041274.8A CN110068330B (en) | 2019-01-16 | 2019-01-16 | Autonomous positioning method of robot based on ARMA model |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN110068330A true CN110068330A (en) | 2019-07-30 |
| CN110068330B CN110068330B (en) | 2021-04-13 |
Family
ID=67365918
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910041274.8A Active CN110068330B (en) | 2019-01-16 | 2019-01-16 | Autonomous positioning method of robot based on ARMA model |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110068330B (en) |
Cited By (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111207754A (en) * | 2020-02-28 | 2020-05-29 | 上海交通大学 | Particle filter-based multi-robot formation positioning method and robot equipment |
| CN111928860A (en) * | 2020-07-24 | 2020-11-13 | 上海交通大学烟台信息技术研究院 | Autonomous vehicle active positioning method based on three-dimensional curved surface positioning capability |
| CN111947672A (en) * | 2020-03-02 | 2020-11-17 | 北京百度网讯科技有限公司 | Method, apparatus, device and medium for detecting environmental changes |
| CN112116656A (en) * | 2020-08-03 | 2020-12-22 | 歌尔股份有限公司 | Incremental mapping method and device in synchronous positioning and mapping slam |
| CN112710312A (en) * | 2020-12-24 | 2021-04-27 | 哈尔滨工业大学芜湖机器人产业技术研究院 | Laser SLAM front-end matching method fusing distance residual and probability residual |
| CN113920137A (en) * | 2021-10-14 | 2022-01-11 | 平安科技（深圳）有限公司 | Lymph node metastasis prediction method, apparatus, device and storage medium |
| CN113970754A (en) * | 2021-10-25 | 2022-01-25 | 北京京东乾石科技有限公司 | Positioning method and device of autonomous travelable equipment |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2018048897A1 (en) * | 2016-09-06 | 2018-03-15 | The Charles Stark Draper Laboratory, Inc. | Fused sensor ensemble for navigation and calibration process therefor |
| CN108732932A (en) * | 2018-06-01 | 2018-11-02 | 荷塘智能科技(固安)有限公司 | A kind of quadrotor drone Accurate Position Control method based on minimum variance regulator |
| CN108896050A (en) * | 2018-06-26 | 2018-11-27 | 上海交通大学 | A kind of mobile robot long-term tillage system and method based on laser sensor |
-
2019
- 2019-01-16 CN CN201910041274.8A patent/CN110068330B/en active Active
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2018048897A1 (en) * | 2016-09-06 | 2018-03-15 | The Charles Stark Draper Laboratory, Inc. | Fused sensor ensemble for navigation and calibration process therefor |
| CN108732932A (en) * | 2018-06-01 | 2018-11-02 | 荷塘智能科技(固安)有限公司 | A kind of quadrotor drone Accurate Position Control method based on minimum variance regulator |
| CN108896050A (en) * | 2018-06-26 | 2018-11-27 | 上海交通大学 | A kind of mobile robot long-term tillage system and method based on laser sensor |
Non-Patent Citations (3)
| Title |
|---|
| 付小宁，等: "《数据处理及模型化方法》", 30 June 2016, 西北农林科技大学出版社 * |
| 唐恒博，等: "高精度高鲁棒性的轨道机器人全局定位方法", 《机器人ROBOT》 * |
| 邢远见，等: "现代时间序列分析在SLAM问题中的应用", 《中国科技论文在线》 * |
Cited By (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111207754A (en) * | 2020-02-28 | 2020-05-29 | 上海交通大学 | Particle filter-based multi-robot formation positioning method and robot equipment |
| CN111947672A (en) * | 2020-03-02 | 2020-11-17 | 北京百度网讯科技有限公司 | Method, apparatus, device and medium for detecting environmental changes |
| US11466992B2 (en) | 2020-03-02 | 2022-10-11 | Beijing Baidu Netcom Science And Technology Co., Ltd. | Method, apparatus, device and medium for detecting environmental change |
| CN111928860A (en) * | 2020-07-24 | 2020-11-13 | 上海交通大学烟台信息技术研究院 | Autonomous vehicle active positioning method based on three-dimensional curved surface positioning capability |
| CN112116656A (en) * | 2020-08-03 | 2020-12-22 | 歌尔股份有限公司 | Incremental mapping method and device in synchronous positioning and mapping slam |
| CN112116656B (en) * | 2020-08-03 | 2024-05-31 | 歌尔股份有限公司 | Incremental mapping method and device in synchronous positioning and map construction slam |
| CN112710312A (en) * | 2020-12-24 | 2021-04-27 | 哈尔滨工业大学芜湖机器人产业技术研究院 | Laser SLAM front-end matching method fusing distance residual and probability residual |
| CN112710312B (en) * | 2020-12-24 | 2023-12-01 | 长三角哈特机器人产业技术研究院 | Laser SLAM front end matching method integrating distance residual error and probability residual error |
| CN113920137A (en) * | 2021-10-14 | 2022-01-11 | 平安科技（深圳）有限公司 | Lymph node metastasis prediction method, apparatus, device and storage medium |
| CN113970754A (en) * | 2021-10-25 | 2022-01-25 | 北京京东乾石科技有限公司 | Positioning method and device of autonomous travelable equipment |
Also Published As
| Publication number | Publication date |
|---|---|
| CN110068330B (en) | 2021-04-13 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110068330A (en) | Autonomic positioning method based on the robot that arma modeling is realized | |
| Sun et al. | Using Bayesian deep learning to capture uncertainty for residential net load forecasting | |
| CN104021430B (en) | Method for analyzing uncertainty of passenger flow of urban mass transit terminal | |
| CN108955679A (en) | A kind of robot used for intelligent substation patrol high-precision locating method | |
| Aguado et al. | Study of self-calibration and multilateration in machine tool volumetric verification for laser tracker error reduction | |
| CN106403953B (en) | A method of for underwater independent navigation and positioning | |
| CN103810266A (en) | Semantic network object identification and judgment method | |
| Sileshi et al. | Particle filter slam on fpga: A case study on robot@ factory competition | |
| Antoniou et al. | Calibration methods for simulation-based dynamic traffic assignment systems | |
| CN102609790A (en) | Online state estimation method for electric power system by means of hybrid measurement | |
| Naiman et al. | Generative modeling of regular and irregular time series data via koopman VAEs | |
| Blanco et al. | An optimal filtering algorithm for non-parametric observation models in robot localization | |
| Pei et al. | An improved FastSLAM system based on distributed structure for autonomous robot navigation | |
| Mohedano et al. | Camera localization usingtrajectories and maps | |
| US20220383231A1 (en) | Generating status of construction site based on hierarchical modeling that standardizes physical relationships of elements of a structure | |
| Patel et al. | CNN based Variation and Prediction Analysis of 2m Air Temperature for Different Zones of the Indian Region | |
| Zhao et al. | Uncertainty-aware gaussian mixture model for UWB time difference of arrival localization in cluttered environments | |
| Khadka et al. | Performance analysis of hybrid forecasting model in stock market forecasting | |
| Ping et al. | Particle filter based time series prediction of daily sales of an online retailer | |
| Beltran et al. | Oracle-based optimization applied to climate model calibration | |
| Özkan et al. | Non-parametric bayesian measurement noise density estimation in non-linear filtering | |
| Hasberg et al. | Online-estimation of road map elements using spline curves | |
| Dehbi et al. | Room shapes and functional uses predicted from sparse data | |
| JP2021012539A (en) | Data prediction system and method thereof | |
| Dyvak et al. | The structural identification of interval models of static systems |
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 |