CN106774314A - A kind of home-services robot paths planning method based on run trace - Google Patents
A kind of home-services robot paths planning method based on run trace Download PDFInfo
- Publication number
- CN106774314A CN106774314A CN201611134903.4A CN201611134903A CN106774314A CN 106774314 A CN106774314 A CN 106774314A CN 201611134903 A CN201611134903 A CN 201611134903A CN 106774314 A CN106774314 A CN 106774314A
- Authority
- CN
- China
- Prior art keywords
- track
- similarity
- point
- motion
- run trace
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 238000000034 method Methods 0.000 title claims abstract description 41
- 230000033001 locomotion Effects 0.000 claims abstract description 86
- 238000004422 calculation algorithm Methods 0.000 claims abstract description 39
- 238000005259 measurement Methods 0.000 claims abstract description 14
- 238000013459 approach Methods 0.000 claims description 6
- 239000011159 matrix material Substances 0.000 claims description 6
- 230000008569 process Effects 0.000 claims description 6
- 238000003064 k means clustering Methods 0.000 claims description 5
- 238000000205 computational method Methods 0.000 claims description 3
- 230000007613 environmental effect Effects 0.000 abstract description 3
- 230000006872 improvement Effects 0.000 abstract description 3
- 238000010845 search algorithm Methods 0.000 abstract 1
- 238000005070 sampling Methods 0.000 description 8
- 230000006399 behavior Effects 0.000 description 7
- 230000000694 effects Effects 0.000 description 6
- 230000004888 barrier function Effects 0.000 description 4
- 238000004364 calculation method Methods 0.000 description 4
- 230000011218 segmentation Effects 0.000 description 4
- 238000002790 cross-validation Methods 0.000 description 3
- 238000011160 research Methods 0.000 description 3
- 238000004458 analytical method Methods 0.000 description 2
- 230000015572 biosynthetic process Effects 0.000 description 2
- 230000008859 change Effects 0.000 description 2
- 230000006866 deterioration Effects 0.000 description 2
- 238000010586 diagram Methods 0.000 description 2
- 239000000284 extract Substances 0.000 description 2
- 235000012054 meals Nutrition 0.000 description 2
- 230000007246 mechanism Effects 0.000 description 2
- 241000894007 species Species 0.000 description 2
- XLYOFNOQVPJJNP-UHFFFAOYSA-N water Substances O XLYOFNOQVPJJNP-UHFFFAOYSA-N 0.000 description 2
- 241001465754 Metazoa Species 0.000 description 1
- 230000002776 aggregation Effects 0.000 description 1
- 238000004220 aggregation Methods 0.000 description 1
- 230000003542 behavioural effect Effects 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000002457 bidirectional effect Effects 0.000 description 1
- 230000004087 circulation Effects 0.000 description 1
- 238000003181 co-melting Methods 0.000 description 1
- 238000010411 cooking Methods 0.000 description 1
- 230000001186 cumulative effect Effects 0.000 description 1
- 238000007405 data analysis Methods 0.000 description 1
- 230000003247 decreasing effect Effects 0.000 description 1
- 230000007812 deficiency Effects 0.000 description 1
- 230000001419 dependent effect Effects 0.000 description 1
- 238000001514 detection method Methods 0.000 description 1
- 230000002708 enhancing effect Effects 0.000 description 1
- 238000000605 extraction Methods 0.000 description 1
- 238000007306 functionalization reaction Methods 0.000 description 1
- 238000009499 grossing Methods 0.000 description 1
- 239000003102 growth factor Substances 0.000 description 1
- 238000007689 inspection Methods 0.000 description 1
- 238000012544 monitoring process Methods 0.000 description 1
- 230000007935 neutral effect Effects 0.000 description 1
- 238000005457 optimization Methods 0.000 description 1
- 238000004088 simulation Methods 0.000 description 1
- 230000007958 sleep Effects 0.000 description 1
- 230000003595 spectral effect Effects 0.000 description 1
- 238000012360 testing method Methods 0.000 description 1
- 230000000007 visual effect Effects 0.000 description 1
Classifications
-
- G—PHYSICS
- G05—CONTROLLING; REGULATING
- G05D—SYSTEMS FOR CONTROLLING OR REGULATING NON-ELECTRIC VARIABLES
- G05D1/00—Control of position, course, altitude or attitude of land, water, air or space vehicles, e.g. using automatic pilots
- G05D1/02—Control of position or course in two dimensions
- G05D1/021—Control of position or course in two dimensions specially adapted to land vehicles
- G05D1/0212—Control of position or course in two dimensions specially adapted to land vehicles with means for defining a desired trajectory
- G05D1/0223—Control of position or course in two dimensions specially adapted to land vehicles with means for defining a desired trajectory involving speed control of the vehicle
-
- G—PHYSICS
- G05—CONTROLLING; REGULATING
- G05D—SYSTEMS FOR CONTROLLING OR REGULATING NON-ELECTRIC VARIABLES
- G05D1/00—Control of position, course, altitude or attitude of land, water, air or space vehicles, e.g. using automatic pilots
- G05D1/02—Control of position or course in two dimensions
- G05D1/021—Control of position or course in two dimensions specially adapted to land vehicles
- G05D1/0212—Control of position or course in two dimensions specially adapted to land vehicles with means for defining a desired trajectory
- G05D1/0214—Control of position or course in two dimensions specially adapted to land vehicles with means for defining a desired trajectory in accordance with safety or protection criteria, e.g. avoiding hazardous areas
Landscapes
- Engineering & Computer Science (AREA)
- Aviation & Aerospace Engineering (AREA)
- Radar, Positioning & Navigation (AREA)
- Remote Sensing (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Automation & Control Theory (AREA)
- Control Of Position, Course, Altitude, Or Attitude Of Moving Bodies (AREA)
- Manipulator (AREA)
Abstract
A kind of home-services robot paths planning method based on run trace, the present invention discloses a kind of service robot air navigation aid of the run trace based on people.First, the describing mode of pedestrian movement track is inquired into, propose a kind of key point based on Density Estimator to extract and track expression, the classification of pedestrian movement track is realized by similarity measurement of the track in space length, the direction of motion, movement velocity.Secondly, to avoid extension of the path search algorithm in global space, improve the real-time of algorithm, it is proposed that a kind of improvement RRT Connect algorithms based on track gravitation function, realize the path search scheme for tending to pedestrian's run trace.Finally, realize based on topology grid environmental model bilayer path planning, on the basis of safety navigation, improve navigation efficiency.
Description
Technical field
The invention belongs to robot navigation and positioning field, more particularly to a kind of home services machine based on run trace
People's paths planning method.
Technical background：
Navigation is one of basic function of service robot, is also the basis that robot completes service role.Traditional machine
Device people navigation mode is generally based on most short time, shortest path or the most low optimizing index of power consumption and realizes, it is right to have ignored
The requirement of the service quality such as robot motion's security, stationarity and people's friendly, the motion path cooked up is generally segmentation
Broken line.Service robot is accomplished by continually slowing down, stop, turning and accelerating when path trace is carried out, so as to cause machine
Discontinuous, the stationarity and deterioration of safety of robot of people's motion.It is traditional particularly with the stronger home environment of dynamic
Air navigation aid only considers the optimization of Path selection, does not account for the constraint of the motor behavior and robot motion region of pedestrian
Property, so that being had difficulty in taking a step in path trace, the objective of pleasantization service is deviate from, let alone it is co-melting with people.
However, obtaining knowledge from the behavior of people, the daily exercise rail of people is extracted in motion of the observer in home environment
Mark, and guided robot walks along the moving line of people as people, helps to provide the service of hommization.The walking of people
Track is directly perceived expression of the human motion behavior in time-space domain, reflects daily behavior and the habits and customs of people, is contained
Abundant information.Pedestrian movement track under home environment is just as one straightway " highway " in real world, tool
There is the accessibility in realistic meaning.
In recent years, by monitoring people's moving line under environment indoors, movement locus, guided robot navigation and control are extracted
Research in terms of system has gradually obtained the concern [141] of people.For example, Simon etc. [142] is using the inspection of distribution type laser radar
Pedestrian position is surveyed, its movement locus is extracted, every track is modeled by Gaussian process, while a plurality of track is directly averagely made
It is desired trajectory.The method have ignored amphicheirality and the complexity of pedestrian movement track, merely can not be complete to track mean deviation
Reflect the actual run trace of people, the track route of the path planning and people that are easily caused robot is clashed.Bennewitz
The movement locus of people being extracted using visual apparatus and being analyzed, the movement locus based on people predicts the motion road of people Deng [143]
A kind of line, it is proposed that robot navigation method of compliance target person motor behavior.The method excessively considers the motor behavior of people, sacrificial
The domestic animal flexibility of robot autonomous motion.
The content of the invention
In view of the regular daily routines of people show similitude and repeatability higher on movement locus, contain
Abundant operable information, quickly navigates for guide service robot, and the present invention provides a kind of family based on run trace
Service robot paths planning method, comprises the following steps：
S1, the describing mode according to pedestrian movement track, are extracted using Density Estimator to track key point；
The similarity measurement of S2, definition track in space length, the direction of motion, movement velocity；
S3, carry out the classification of pedestrian movement track with k-means algorithms；
S4, path planning is carried out to pedestrian's run trace in RRT-Connect algorithms.
Preferably, step 2 is specially：
One track of moving target is described as:In two-dimensional space, what is positioned based on target and obtained (is shifted from key point
Node) point set that is constituted to the oriented moving target anchor point between key point (transfering node), thus, moving target certain
Bar movement locus j can be described as：
Tj={ xi=(xi,yi),ai,vi, i=1 ..., N }
Wherein, (xi,yi) describe the coordinate position of target trajectory point i, aiAnd viTarget when respectively producing tracing point i
The direction of motion and locomotion speed value,
When then carrying out similarity measurement to track, trajectory range Distance conformability degree S is introduced respectivelyd, movement velocity is similar
Degree Sv, direction of motion similarity Sa, it is similar in space length, movement velocity, the direction of motion with the movement locus that this weighs people
Degree, if Tm, TnIt is two tracks of moving target, then similarity degree can be represented with a three-dimensional matrice between them:
S(Tm,Tn)=[Sd(Tm,Tn),Sv(Tm,Tn),Sa(Tm,Tn)]
The computational methods of this several similarity measurement are described below, wherein, trace space Distance conformability degree SdOnly with tracing point
Locus it is relevant, for track TmOn any point Xi, in track TnOn closest approach be represented by:
Then track TmWith track TnBetween space length can be expressed as:
Wherein, NmIt is track TmThe number of upper point, track TmWith track TnBetween similarity can be expressed as:
With reference to trajectory range Distance conformability degree expression way, track TmWith track TnBetween direction of motion similarity can table
Up to for:
Track TmWith track TnBetween movement velocity similarity can be expressed as:
Preferably, step 3 is specially：The pedestrian's run trace extracted is clustered using K-means clustering methods,
It is as follows that it implements process:
Step 1:Prediction cluster number, classifies according to interstitial content, track key point number journey meter is shifted in home environment
The species k of track；
Step 2:Track similarity matrix is set up, if the collection of the L bar run trace that Ω is extracted by Target Tracking System
Close, wherein, Ω={ T1,...,Ti,...,TL, TiIt is not i-th track.Any two tracks are carried out to save institute based on 4.2.3
Corresponding similarity three-dimensional vector is calculated by the track measuring similarity stated, then all L bars tracks are carried out similar
Degree measurement, can obtain the similarity matrix of L × L × 3:DI, j=d (S (Ti,Tj))；
Step 3:Initialization cluster centre, randomly selects certain track of track concentration as the initial clustering of the first kind
Center Cn1；Secondly, the initial cluster center C for choosing an other track as Equations of The Second Kind is concentrated in remaining L-1 bars trackn2；
In order to prevent two selected tracks to belong to same class, similarity threshold value ρ is set, make the distance between two class trajectory clustering centers
Meet following formula, by that analogy, until finding out k initial cluster center,
DN1, n2=d (S (Tn1,Tn2))≥ρ
Wherein, the selection of similarity threshold ρ can obtain relevant parameter by being trained to known similar track；
Step 4:Sample trace is sorted out, relatively more all track sample TiWith each initialization cluster centre Cnj's
Similarity, the class where track sample is referred into the initialization cluster centre most like with it,
Step 5:Adjustment cluster centre, the initial category of track and the rail of each class of correspondence are can obtain according to Step 4
Mark number of samples Li, for each class track, in affiliated such all track samples, certain sample trace is found out,
It is arrived the minimum apart from sum of such all track sample, that is, elect new cluster centre as,
Step 6:Step 4 and step Step 5 is repeated, until the adjustment iteration of double cluster centre no longer occurs
Untill change.
The service robot air navigation aid of the run trace based on people of the invention, first, has inquired into pedestrian movement track
Describing mode, it is proposed that a kind of key point based on Density Estimator is extracted and track expression, by track in space
Similarity measurement in distance, the direction of motion, movement velocity realizes the classification of pedestrian movement track.Secondly, to avoid path
Extension of the searching algorithm in global space, improves the real-time of algorithm, it is proposed that a kind of improvement based on track gravitation function
RRT-Connect algorithms, realize the path search scheme for tending to pedestrian's run trace.Finally, realize based on topology-grid
Environmental model bilayer path planning, on the basis of safety navigation, improves navigation efficiency.
Brief description of the drawings
Fig. 1 is schematic flow sheet of the invention；
Fig. 2 is the schematic diagram that Density Estimator is realized based on Gaussian kernel functions；
Fig. 3 is the path planning figure based on single step RRT algorithms；
Fig. 4 is the path planning figure based on RRT-Connect algorithms.
Specific embodiment
As shown in figure 1, the embodiment of the present invention provides a kind of home-services robot path planning side based on run trace
Method, comprises the following steps：
Step one：Track key point is extracted using Density Estimator
Target trajectory is through connection based on a series of oriented data point extracted to the positioning of moving target
The curve of formation, reflects the historical law of target motion.Analysis target movement locus before, first should combing understand
Which type of one section of motion path is referred to as a track.There is very strong purpose and repetition in view of the activity of home environment servant
Property, the motion of people would generally be ordered about by certain purpose sexuality, such as have a meal, sleeps, learns.Therefore, the movement locus of people
Certain aggregation can be shown, its generally area distribution or zonal distribution.For it is accurate, vividly describe the motion rail of people
Mark, introduces the concept of " key point ".Key point refers to the motor behavior dwell point of people, such as water dispenser, desk, the head of a bed.It
Can be extracted from track.Because the segmentation of room area function is obvious in home environment, generally connected by door between region
Connect.Although this kind of node location with very strong identification is difficult to have been embodied in the daily routines of people, they are families
Service robot needs the key node often passed by under the environment of front yard.Therefore, introducing " transfering node ", that is, indicate each function in room
The link position in region.Their position distribution can be manually set.In sum, target motion can be produced a track
It is described as two directing curves connected between key point or transfering node.
Because people's daily life has stronger purpose, such as, had a meal before reading, dining table before desk.Closing
The activity or operation of key point position can cause that the residence time of people is more long, and substantial amounts of track can be produced in object tracking process
Point.Therefore, extract key point and be accomplished by traveling through all tracks, count the probability density of tracing point, moving target residence time compared with
Position long, its probability density will necessarily be of a relatively high.
It is different to the degree of dependence of sample parameter according to object to be estimated during sample probability density estimation, can divide
It is Parameterization estimate method and imparametrization method of estimation.Compared with Parameterization estimate method, nonparametric Multilayer networks are not
Need to make the distribution form of sample prior it is assumed that the probability density of Arbitrary distribution can be estimated, rely solely on sample data
Achieve that and probability density function is more accurately estimated, its superperformance is that the Multilayer networks of tracing point and modeling are provided
One new resolving ideas.Conventional Non-parameter density estimation method mainly has histogram density estimation, kernel function estimation
Method, k-nearest neighbor and base function expansion method etc., the wherein application of first two method are the most universal.With histogram density estimation method
Compare, Density Estimator precision is higher and continuous, it has also become a kind of main stream approach in current sample rate estimation.Therefore, this
Distribution of the Zhang Caiyong Density Estimators to tracing point is counted, and extracts key point.
If { X1,X2,...,XNIt is N number of sample of stochastic variable x, f (x) is the density function of stochastic variable x, then sample
The Density Estimator of data is defined as：
Wherein, h is bandwidth, and D is the dimension of sample data, and K is kernel function, the shape and codomain of kernel function constrain for
The producing level of the number of data point and each data point used by estimated probability density.
Conventional kernel function has the functions such as Uniform, Epanechikov, Biweight and Gaussian, and upper table gives
Their a dimensional expression.They are on origin symmetry and meet following property:
∫ K (u) du=1, ∫ uK (u) du=0, ∫ u2K (u) du=μ2(K)＞ 0
By taking the Gaussian kernel functions described in upper table as an example, find out from the expression formula of Density Estimator, sample point XiWith x
Distance it is nearer, Xi- x close to zero, its probability density K ((x-Xi)/h) it is bigger.I.e. for all data points, its away from
The nearlyer influence to estimation is bigger with a distance from x points, and when bandwidth h is smaller, only range points x especially near data point can just rise
Larger effect, with the increase of bandwidth, the effect of the remote data point of some is consequently increased.Therefore, Density Estimator can be seen
Work is the process cumulative to the core window centered on each sample point, and its Kernel Function determines the type of smooth window, bandwidth h
The width of smooth window is described, Fig. 2 gives the schematic diagram that Density Estimator is realized based on Gaussian kernel functions, wherein counting
More intensive according to collecting, probability density function values are bigger at the position.
In general, the Density Estimator of sample object is relevant with sample data, the selection of kernel function and bandwidth setting.It is right
In given sample data, Density Estimator is dependent only on the selection of the type and bandwidth of kernel function.It is symmetrical with regard to Density Distribution
For kernel function, Silverman and Pracasa Rao etc. once point out the Density Estimator influence of different IPs function pair sample data
It is little, and much smaller than the influence of bandwidth h.It is intended to select optimal nucleus band wide, sample data probability Estimation is calculated firstPhase
Hope:
Wherein, above formula assumes sample number strong point xiIt is the stochastic variable extracted from the sample real density f for being estimated object.
It can thus be seen that the convolution for being desired for kernel function and real density f of Density Estimator, bandwidth h smoothing kernel of kernel function
The effect of density estimation.When h → 0, kernel function enters close to unit impulse function, then sample data probability EstimationIt is close
Real data distribution density.Due to limited sample size in practical application, h is too small to cause density estimation to be excessively sharp, no
Beneficial to data analysis.Density estimation can be then caused excessively to smooth using excessive bandwidth, the true probability distribution of obfuscated data.
Assuming that the second dervative of real density function f is continuous and ∫ [f " (x)]2Dx is present, then f (x-yh) can Taylor at x
Launch:
Density estimationDeviation can Taylor expansion：
Density estimationVariance can be expressed as：
Estimate for sample, density is generally estimated using mean square error (Mean Square Error, MSE) measurement
With the similarity degree between real density f, its definition is as shown in formula (4-7):
Above formula can also be expressed as:
As can be seen that deviation and h2In direct ratio, variance is inversely proportional with bandwidth h, and a less bandwidth can be given
One less estimation of deviation, while variance can be caused to increase.Therefore, when kernel function bandwidth is selected, it is necessary to density estimation exists
Estimated and weigh between the sample bias of object and sample variance.In theory, by minimizing " average integral square error
(Mean Integrated Square Error, MISE) [153], are obtained in that optimum bandwidth h, and its definition is such as formula (4-9) institute
Show:
Make R (f)=∫ f2X () dx, bandwidth h can be expressed as:
Optimal bandwidth h depends on the unknown R (f "), and to obtain optimum bandwidth, researcher proposes several data and drives
Dynamic method, such as Plug-in methods, Bootstrap methods and cross-validation method, but estimated object for different types of
Say, the selection of optimum bandwidth is still a problem for needing to be continued to solve.In view of bandwidth selection be not this chapter research
Emphasis, this chapter has inclined cross validation (Biased Cross-Validation, BCV) method using with higher robustness, visits
Suitable bandwidth is sought, its main thought is using estimationInstead of the unknown R (f "),It is defined as follows:
Object function in above formula is expressed as:
By minimizing object function, can obtain optimum bandwidth is:
hBCV=argminhBCV(h)
The positioning result for moving people can be intuitively expressed as a series of two-dimemsional number strong points, and their sequential combination is constituted
The run trace of people, uses the spatial distribution of Density Estimator analyze data point to explore the life that people's run trace is contained
Rule specifies direction.Resident or operation of the target person at key point can cause that the position produces substantial amounts of tracing point, compare
Track dot density at other positions, key point region is relatively large.This chapter is based on this priori, inversely seeks probability
The larger location point of density is designated as key point, realizes that the key point in run trace is extracted.
Step 2：Define the measurement of track likeness in form degree
Generally, target trajectory can be expressed using the space coordinates of target positioning, be shown below:
T={ Xi=(xi,yi), i=1 ..., N }
But it is not fine enough that this method for expressing can make track classify, and can not embody the direction of target motion.In general, no
The same direction of motion, the target of different motion speed can include the behavioural habits of certain feature, such as, the walking of keeping right of pedestrian.Cause
This, to improve the accuracy of target trajectory classification, by the direction of motion of target, produces movement velocity during track to include target track
The description content of mark.
Therefore be described as a track of moving target by the present invention:In two-dimensional space, based on target position and obtain from
The point set that key point (transfering node) is constituted to the oriented moving target anchor point between key point (transfering node).Thus,
Certain movement locus j of moving target can be described as
Tj={ xi=(xi,yi),ai,vi, i=1 ..., N }
Wherein, (xi,yi) describe the coordinate position of target trajectory point i, aiAnd viTarget when respectively producing tracing point i
The direction of motion and locomotion speed value.
Because the motion of people in home environment has very strong repeatability, target fortune may be have recorded between two key points
Dynamic many bar tracks, robot is generally difficult to judge optimal along which bar track walking effect.It is the phase between statistics track
Mutual relation, finds representative track, it is necessary to carry out track classification.According to the describing mode of track, enter to track
During row similarity measurement, trajectory range Distance conformability degree S is introduced respectivelyd, movement velocity similarity Sv, direction of motion similarity Sa,
Similarity degree of the movement locus of people in space length, movement velocity, the direction of motion is weighed with this.If Tm, TnIt is moving target
Two tracks, then similarity degree can be represented with a three-dimensional matrice between them:
S(Tm,Tn)=(Sd(Tm,Tn),Sv(Tm,Tn),Sa(Tm,Tn)]
The computational methods of this several similarity measurement, wherein trajectory range Distance conformability degree S are described belowdOnly with tracing point
Locus it is relevant, for track TmOn any point Xi, in track TnOn closest approach be represented by:
Then track TmWith track TnBetween space length can be expressed as:
Wherein, NmIt is track TmThe number of upper point.Then track TmWith track TnBetween similarity can be expressed as:
With reference to trajectory range Distance conformability degree expression way, track TmWith track TnBetween direction of motion similarity can table
Up to for:
Track TmWith track TnBetween movement velocity similarity can be expressed as:
Step 3：Track classification is carried out using k-means algorithms
Trajectory clustering is the basis of trajectory analysis and application, it is contemplated that seeking family based on the cluster to run trace
The highway of robot ambulation under environment.Current conventional clustering method mainly has K-means clusters, spectral clustering, level to gather
The method such as class and the cluster based on neutral net, wherein, K-means clustering methods are simple to operate, and amount of calculation is small, are widely used in
In various types of characteristic classification.For semi-structured home environment, because the motion of pedestrian target is usually focused on
Certain route and direction, data characteristics are relatively clear, and being based on Density Estimator can estimate the number of key point on track,
So as to the classification for movement locus specifies direction.Therefore, pedestrian walking of the present invention using K-means clustering methods to extracting
Track is clustered, and it is as follows that it implements process:
Step 1:Prediction cluster number.Classify according to interstitial content, track key point number journey meter is shifted in home environment
The species k of track；
Step 2:Set up track similarity matrix.If the collection of the L bar run trace that Ω is extracted by Target Tracking System
Close, wherein, Ω={ T1,...,Ti,...,TL, TiIt is not i-th track.Any two tracks are carried out to save institute based on 4.2.3
Corresponding similarity three-dimensional vector is calculated by the track measuring similarity stated, then all L bars tracks are carried out similar
Degree measurement, can obtain the similarity matrix of L × L × 3:DI, j=d (S (Ti,Tj))。
Step 3:Initialization cluster centre.Certain track of track concentration is randomly selected as the initial clustering of the first kind
Center Cn1；Secondly, the initial cluster center C for choosing an other track as Equations of The Second Kind is concentrated in remaining L-1 bars trackn2；
In order to prevent two selected tracks to belong to same class, similarity threshold value ρ is set, make the distance between two class trajectory clustering centers
Meet following formula.By that analogy, until finding out k initial cluster center.
DN1, n2=d (S (Tn1,Tn2))≥ρ
Wherein, the selection of similarity threshold ρ can obtain relevant parameter by being trained to known similar track.
Step 4:Sample trace is sorted out.Compare all track sample TiWith each initialization cluster centre Cnj's
Similarity, the class where track sample is referred into the initialization cluster centre most like with it.
Step 5:Adjustment cluster centre.The initial category of track and the rail of each class of correspondence are can obtain according to Step 4
Mark number of samples Li.For each class track, in affiliated such all track samples, certain sample trace is found out,
It is arrived the minimum apart from sum of such all track sample, that is, elect new cluster centre as.
Step 6:Step 4 and step Step 5 is repeated, until the adjustment iteration of double cluster centre no longer occurs
Untill change.
The movement locus of people is extracted under the home environment of simulation, the basis of key point is being extracted based on Density Estimator
On, classify using K-means clustering methods and to it, 75 fortune of the extracted people before desk, water dispenser, bed can be shown
Dynamic rail mark.With reference to the number K of key pointp=3, the amphicheirality of run trace, set the initial clustering number of track as k=2 ×
Kp, trajectory clustering is more clearly shown the statistical law of pedestrian movement, such as walking of keeping right.
Step 4：Path planning is carried out based on RRT-Connect algorithms
After pedestrian's run trace of extraction is classified, to make service robot as far as possible along the height in home environment
Highway is walked, it is necessary to corresponding path planning algorithm.Due to article distribution diversity, randomness, server in home environment
The motion planning of device people would generally be faced with the complex environment constraint that a large amount of barriers and slype are formed.Traditional path
Planing method is absorbed in from geometry angle calculation robot motion route and orientation, have ignored robot joint angles limitation and
" incomplete property " is constrained, and in the path planning problem of solving complexity environmental constraints, the motion path cooked up is generally segmentation
Broken line, planning efficiency can then be decreased obviously, or even be difficult to try to achieve effectively solution.And robot when path trace is carried out, it is necessary to frequency
Slow down numerously, stop, turning, accelerate, cause that robot motion's is discontinuous, its stationarity and deterioration of safety.For example, being in
In the environment of front yard, the segmentation of room area functionalization is obvious, when service-delivery machine human desires is by narrow suitable with car body size of width
During kissing gate mouthful, stronger constraint can be caused with steering to the movement velocity of robot.The present invention regarding to the issue above, is proposed a kind of
Improve two-way Quick Extended random tree (Rapidly-exploring Random Tree-Connect, RRT Connect) calculation
Method, purposefully guided robot carry out path planning along the track route of people.
RRT algorithms were proposed first by S.M.LaValle in 1998, were that calculation is planned in a kind of single query path based on sampling
Method.It does not need specific heuristic function, and by the constantly stochastical sampling in state space, most region of search is directed at last
The guiding of property is to target area.Distinctive stochastical sampling mechanism makes the algorithm solve higher dimensional space multi-freedom robot complexity
During constrained path planning problem, efficient search attribute is presented.
RRT algorithm main tasks are that one from initial position x is found in state space Qinit, set out to terminal xend
Continuous path, it is desirable to the path cut-through region Q for cooking upobst, it is ensured that robot is in free state region QfreeMotion.
If robot pose changes upstate equation dx/dt=f (x, u) expression, wherein x is robot original position, and u is defeated for control
Enter, time interval is Δ t, then the new position and posture of robot can be expressed as xnew=x+f (x, u) Δ t.Following table gives unidirectionally
The false code description of RRT algorithms.First, initialization input parameter:；xinit, N, u, Δ t；Secondly, into Line3-8 circulations:Adjust
With subfunction RandomState (), state x is randomly selected in state spacerand；Call subfunction NearestNeighbor
(), the detection range x in Tree is setrandNearest node xnear；Subfunction SelectInput () is called, control rate u is selected,
Propulsion Δ t obtains new state xnew；By xnewIt is added in tree Tree with control rate u；Each state in last reverse connection tree Tree
Point can obtain path planning.
RRT Connect algorithms, also known as two-way (double trees) RRT, algorithm, as on the basis of unidirectional RRT algorithms, in mesh
Punctuate is re-introduced into one tree, while using greedy strategy, the extension of random tree is by the single step life for becoming many step-lengths, accelerating tree long
Speed long, improves planning efficiency.
As shown in figure 3, in unidirectional RRT algorithms, calling subfunction Selectinput (), control rate u, propulsion Δ t are selected
Obtain new state xnew, the formation speed of random tree is limited to a certain extent.And in random two-way tree, first to new section
Point xnewGenerating mode improved, that is, be directly toward stochastical sampling collocation point xrandOne step-length δ of extension, produces distance
Nearest new node xnew, and to the node x of new extensionnewTest, if collided with barrier, according to this principle,
Following table gives the Implementation of pseudocode of extension random tree (Extend RRT).
As can be seen from the above table, from xnearSet out and only extend a step-length, the expansion rate of algorithm is relatively slow.
Kuffner [158] is pointed out along xnearWith xrandLine more extend several step-lengths, can accelerate tree the speed of growth.Not only such as
This, one tree is drawn from impact point again, and using two-way multistep, RRT convergences of algorithm speed will be multiplied.Therefore, they
Two-way multistep RRT algorithms are proposed, also known as RRT-Connect algorithms.Growth random tree is extended simultaneously from starting point and target,
Extend RRT algorithms are called by loop iteration, and along sampling collocation point xrandThe multiple step-lengths of extension, until meeting barrier
Hinder thing or reach sampling collocation point xrand, new node is returned again to, when two trees meet, path produces.Table 4-4 give RRT-
The false code that Connect algorithms are realized.Input item x is set firstinit, xgoal, N, S build two from beginning and end respectively
Tree；Secondly two alternate cycles extensions of random tree are realized using Expansion () function, if the node of two trees can be right
Should connect, then return path；Otherwise, two trees are exchanged, is extended again.
Table 4-4
RRT-Connect algorithms are advised as a kind of two-way approach planning algorithm based on stochastical sampling in robot path
The field of drawing, has been obtained for in-depth study and is widely applied.The characteristics of its random sampling, determines that it can only be in probability
Complete in theory, itself also includes some shortcomings:First, its stability is not strong, i.e., repeated when to same planning tasks
During planning, different robot motion paths may be produced；Secondly, with certain deviation, that is, the motion road cooked up
Footpath may not be optimal path or sub-optimal path；3rd, without guidance quality, if not applying to be partial to the guiding of target to search tree
Constraint, its convergence rate is relatively slow.For the deficiency of above-mentioned RRT algorithms, domestic and international researcher has been carried out constantly to the algorithm
Improvement trial.For example, the stability in order to improve the algorithm search path, proposes ERRT [161], DRRT [162] in succession,
MP-RRT algorithms [163] etc.；To improve its search efficiency, researcher proposes deflection search tree [164], bidirectional research tree successively
(Bidirectional-RRT, Bi-RRT) [165] etc., they improve the performance of algorithm to a certain extent.
As shown in figure 4, present subject matter thought is the service-delivery machine that the run trace realization based on people is walked as people
People navigates, and its final goal is to guide to pedestrian movement track the path planning of robot.Therefore, from heuritic approach
Inspiration is drawn in thought, RRT Connect algorithms are improved, introduce track gravitation function, it is to avoid random tree is complete
Extension in office space, reduces operand, improves the real-time of algorithm, makes the extension of random tree as far as possible along the walking of people
Track grows.Its core concept is:Increase gravitation function G (x) at each node x, then from starting point to node x to mesh
The extension of punctuate instructs function can be expressed as:
| F (x)=R (x)+G (x) |
Wherein, R (x) is the random growth function from starting point to node x, is defined as：
Wherein, a is random growth coefficient.It is similar to therewith, impact point xgoalWith current point xcurrentBetween gravitation function can
It is defined as:
Wherein, η is gravitation coefficient of growth.Can then construct RRT-Connect algorithms from starting point to node x to impact point
Extension instructs the function to be:
It can thus be concluded that, improving the new leaf node computing formula of RRT-Connect algorithms is:
By setting random growth factor alpha and gravitational field coefficient η, you can targetedly guide the growth side of random tree
To.When random tree grows in the area of space of clear into, its direction of growth is guided the course bearing of people, transported along pedestrian
Dynamic direction growth, reduces the randomness of young leaves node selection.When random tree encounters barrier, enhancing random tree growth with
Machine, promotes random tree avoiding obstacles, is grown to depletion region.This growth mechanism not only ensure that and tend to run trace
Path planning, and make the path of planning close to optimal.
Claims (3)
1. a kind of home-services robot paths planning method based on run trace, it is characterised in that comprise the following steps：
S1, the describing mode according to pedestrian movement track, are extracted using Density Estimator to track key point；
The similarity measurement of S2, definition track in space length, the direction of motion, movement velocity；
S3, carry out the classification of pedestrian movement track with k-means algorithms；
S4, path planning is carried out to pedestrian's run trace in RRT-Connect algorithms.
2. the home-services robot paths planning method of run trace is based on as claimed in claim 1, it is characterised in that step
Rapid 2 are specially：
One track of moving target is described as:In two-dimensional space, what is positioned based on target and obtained (shifts section from key point
Point) point set that is constituted to the oriented moving target anchor point between key point (transfering node), thus, certain of moving target
Movement locus j can be described as：
Tj={ xi=(xi,yi),ai,vi, i=1 ..., N }
Wherein, (xi,yi) describe the coordinate position of target trajectory point i, aiAnd viThe motion of target when respectively producing tracing point i
Direction and locomotion speed value,
When then carrying out similarity measurement to track, trajectory range Distance conformability degree S is introduced respectivelyd, movement velocity similarity Sv,
Direction of motion similarity Sa, similar journey of the movement locus of people in space length, movement velocity, the direction of motion is weighed with this
Degree, if Tm, TnIt is two tracks of moving target, then similarity degree can be represented with a three-dimensional matrice between them:
S(Tm,Tn)=[Sd(Tm,Tn),Sv(Tm,Tn),Sa(Tm,Tn)]
The computational methods of this several similarity measurement are described below, wherein, trace space Distance conformability degree SdOnly with the space of tracing point
Position is relevant, for track TmOn any point Xi, in track TnOn closest approach be represented by:
Then track TmWith track TnBetween space length can be expressed as:
Wherein, NmIt is track TmThe number of upper point, track TmWith track TnBetween similarity can be expressed as:
With reference to trajectory range Distance conformability degree expression way, track TmWith track TnBetween direction of motion similarity can be expressed as:
Track TmWith track TnBetween movement velocity similarity can be expressed as:
。
3. the home-services robot paths planning method of run trace is based on as claimed in claim 1, it is characterised in that step
Rapid 3 are specially：
The pedestrian's run trace extracted is clustered using K-means clustering methods, it is as follows that it implements process:
Step 1:Prediction cluster number, according to transfer interstitial content, track key point number journey meter classification track in home environment
Species k；
Step 2:Track similarity matrix is set up, if the set of the L bar run trace that Ω is extracted by Target Tracking System, its
In, Ω={ T1,...,Ti,...,TL, TiIt is not i-th track.Any two tracks are carried out to be based on described in 4.2.3 sections
Track measuring similarity can be calculated corresponding similarity three-dimensional vector, then carry out similarity degree to all L bars tracks
Amount, can obtain the similarity matrix of L × L × 3:Di,j=d (S (Ti,Tj))；
Step 3:Initialization cluster centre, randomly selects certain track of track concentration as the initial cluster center of the first kind
Cn1；Secondly, the initial cluster center C for choosing an other track as Equations of The Second Kind is concentrated in remaining L-1 bars trackn2；In order to
Prevent two selected tracks to belong to same class, set similarity threshold value ρ, meet the distance between two class trajectory clustering centers
Following formula, by that analogy, until finding out k initial cluster center,
Dn1,n2=d (S (Tn1,Tn2))≥ρ
Wherein, the selection of similarity threshold ρ can obtain relevant parameter by being trained to known similar track；
Step 4:Sample trace is sorted out, relatively more all track sample TiWith each initialization cluster centre CnjIt is similar
Degree, the class where track sample is referred into the initialization cluster centre most like with it,
Step 5:Adjustment cluster centre, the initial category of track and the track sample of each class of correspondence are can obtain according to Step 4
This number Li, for each class track, in affiliated such all track samples, certain sample trace is found out, make it
It is minimum apart from sum to such all track sample, that is, elect new cluster centre as,
Step 6:Step 4 and step Step 5 is repeated, until the adjustment iteration of double cluster centre no longer changes
Untill.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201611134903.4A CN106774314A (en) | 2016-12-11 | 2016-12-11 | A kind of home-services robot paths planning method based on run trace |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201611134903.4A CN106774314A (en) | 2016-12-11 | 2016-12-11 | A kind of home-services robot paths planning method based on run trace |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN106774314A true CN106774314A (en) | 2017-05-31 |
Family
ID=58875300
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201611134903.4A Pending CN106774314A (en) | 2016-12-11 | 2016-12-11 | A kind of home-services robot paths planning method based on run trace |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN106774314A (en) |
Cited By (27)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107168342A (en) * | 2017-07-12 | 2017-09-15 | 哈尔滨工大智慧工厂有限公司 | A kind of pedestrian track Forecasting Methodology for robot path planning |
| CN107368076A (en) * | 2017-07-31 | 2017-11-21 | 中南大学 | Robot motion's pathdepth learns controlling planning method under a kind of intelligent environment |
| CN108871344A (en) * | 2018-07-13 | 2018-11-23 | 北京工业大学 | Soccer robot GGRRT paths planning method |
| CN108981704A (en) * | 2018-07-13 | 2018-12-11 | 昆明理工大学 | A kind of two-way RRT paths planning method of target gravitation based on dynamic step length |
| CN109033395A (en) * | 2018-08-01 | 2018-12-18 | 安徽华米信息科技有限公司 | Determine the method and device in section |
| CN109376212A (en) * | 2018-11-22 | 2019-02-22 | 上海木木聚枞机器人科技有限公司 | A kind of generation method and system of the map calculating pedestrian's probability of occurrence |
| CN109540166A (en) * | 2018-11-30 | 2019-03-29 | 电子科技大学 | A kind of Safe path planning method based on Gaussian process |
| CN109737970A (en) * | 2019-03-21 | 2019-05-10 | 集美大学 | A kind of unmanned surface vehicle paths planning method based on improvement RRT algorithm |
| CN109839110A (en) * | 2019-01-09 | 2019-06-04 | 浙江大学 | A kind of multiple target point path planning method based on quick random search tree |
| CN109855622A (en) * | 2019-01-07 | 2019-06-07 | 上海岚豹智能科技有限公司 | Method for searching path and equipment for mobile robot |
| CN110083165A (en) * | 2019-05-21 | 2019-08-02 | 大连大学 | A kind of robot paths planning method under complicated narrow environment |
| CN110347162A (en) * | 2019-07-29 | 2019-10-18 | 南京市晨枭软件技术有限公司 | A kind of industrial robot path modified method and system automatically |
| CN110354475A (en) * | 2019-07-16 | 2019-10-22 | 哈尔滨理工大学 | A kind of tennis racket swinging movement pattern training method and device |
| WO2019218526A1 (en) * | 2018-05-14 | 2019-11-21 | 平安科技（深圳）有限公司 | Method for robot to follow human, and apparatus, storage medium, and robot |
| CN110488811A (en) * | 2019-07-22 | 2019-11-22 | 上海有个机器人有限公司 | A method of the robot based on social networks model predicts pedestrian track |
| CN110490047A (en) * | 2019-07-01 | 2019-11-22 | 珠海格力电器股份有限公司 | A kind of appliances equipment control method, household appliance and computer-readable medium |
| CN110703757A (en) * | 2019-10-24 | 2020-01-17 | 北京航盛新能科技有限公司 | Energy consumption optimization-oriented high-speed train speed planning method |
| CN110765130A (en) * | 2019-09-25 | 2020-02-07 | 武汉大学 | Ripley's K function-based spatio-temporal POI data point pattern analysis method in distributed environment |
| CN111103875A (en) * | 2018-10-26 | 2020-05-05 | 科沃斯机器人股份有限公司 | Method, apparatus and storage medium for avoiding |
| CN112087488A (en) * | 2020-08-03 | 2020-12-15 | 济南浪潮高新科技投资发展有限公司 | Method, device, equipment and medium for determining important cloud robot nodes |
| CN112828889A (en) * | 2021-01-05 | 2021-05-25 | 佛山科学技术学院 | Six-axis cooperative mechanical arm path planning method and system |
| CN113110482A (en) * | 2021-04-29 | 2021-07-13 | 苏州大学 | Indoor environment robot exploration method and system based on priori information heuristic method |
| CN113240353A (en) * | 2021-07-12 | 2021-08-10 | 南京擎天全税通信息科技有限公司 | Cross-border e-commerce oriented export factory classification method and device |
| CN114511999A (en) * | 2020-11-17 | 2022-05-17 | 郑州宇通客车股份有限公司 | Pedestrian behavior prediction method and device |
| CN112363505B (en) * | 2020-11-10 | 2022-06-21 | 合肥工业大学 | Articulated sweeper speed planning method and system based on target distance |
| CN115297303A (en) * | 2022-09-29 | 2022-11-04 | 国网浙江省电力有限公司 | Image data acquisition and processing method and device suitable for power grid power transmission and transformation equipment |
| CN116909293A (en) * | 2023-09-13 | 2023-10-20 | 宁德思客琦智能装备有限公司 | Robot path planning method and device, electronic equipment and computer readable medium |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP2009211571A (en) * | 2008-03-06 | 2009-09-17 | Sony Corp | Course planning device, course planning method, and computer program |
| US20110035087A1 (en) * | 2009-08-10 | 2011-02-10 | Samsung Electronics Co., Ltd. | Method and apparatus to plan motion path of robot |
| CN102169344A (en) * | 2010-12-30 | 2011-08-31 | 南开大学 | Multi-robot cooperative exploring and mapping method with communication distances limited under unknown environments |
| CN103456192A (en) * | 2013-09-01 | 2013-12-18 | 中国民航大学 | Terminal area prevailing traffic flow recognizing method based on track spectral clusters |
| CN103558856A (en) * | 2013-11-21 | 2014-02-05 | 东南大学 | Service mobile robot navigation method in dynamic environment |
| CN105488099A (en) * | 2015-11-03 | 2016-04-13 | 杭州全实鹰科技有限公司 | Vehicle retrieval method based on similarity learning |
-
2016
- 2016-12-11 CN CN201611134903.4A patent/CN106774314A/en active Pending
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP2009211571A (en) * | 2008-03-06 | 2009-09-17 | Sony Corp | Course planning device, course planning method, and computer program |
| US20110035087A1 (en) * | 2009-08-10 | 2011-02-10 | Samsung Electronics Co., Ltd. | Method and apparatus to plan motion path of robot |
| CN102169344A (en) * | 2010-12-30 | 2011-08-31 | 南开大学 | Multi-robot cooperative exploring and mapping method with communication distances limited under unknown environments |
| CN103456192A (en) * | 2013-09-01 | 2013-12-18 | 中国民航大学 | Terminal area prevailing traffic flow recognizing method based on track spectral clusters |
| CN103558856A (en) * | 2013-11-21 | 2014-02-05 | 东南大学 | Service mobile robot navigation method in dynamic environment |
| CN105488099A (en) * | 2015-11-03 | 2016-04-13 | 杭州全实鹰科技有限公司 | Vehicle retrieval method based on similarity learning |
Non-Patent Citations (1)
| Title |
|---|
| 段朋: "与人共融的家庭服务机器人定位与导航", 《中国博士学位论文全文数据库 信息科技辑》 * |
Cited By (42)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107168342A (en) * | 2017-07-12 | 2017-09-15 | 哈尔滨工大智慧工厂有限公司 | A kind of pedestrian track Forecasting Methodology for robot path planning |
| CN107168342B (en) * | 2017-07-12 | 2020-04-07 | 哈尔滨工大智慧工厂有限公司 | Pedestrian trajectory prediction method for robot path planning |
| CN107368076A (en) * | 2017-07-31 | 2017-11-21 | 中南大学 | Robot motion's pathdepth learns controlling planning method under a kind of intelligent environment |
| CN107368076B (en) * | 2017-07-31 | 2018-03-27 | 中南大学 | Robot motion's pathdepth learns controlling planning method under a kind of intelligent environment |
| WO2019218526A1 (en) * | 2018-05-14 | 2019-11-21 | 平安科技（深圳）有限公司 | Method for robot to follow human, and apparatus, storage medium, and robot |
| CN108871344A (en) * | 2018-07-13 | 2018-11-23 | 北京工业大学 | Soccer robot GGRRT paths planning method |
| CN108981704A (en) * | 2018-07-13 | 2018-12-11 | 昆明理工大学 | A kind of two-way RRT paths planning method of target gravitation based on dynamic step length |
| CN108871344B (en) * | 2018-07-13 | 2022-02-08 | 北京工业大学 | Football robot GGRRT path planning method |
| CN109033395A (en) * | 2018-08-01 | 2018-12-18 | 安徽华米信息科技有限公司 | Determine the method and device in section |
| CN111103875A (en) * | 2018-10-26 | 2020-05-05 | 科沃斯机器人股份有限公司 | Method, apparatus and storage medium for avoiding |
| CN109376212A (en) * | 2018-11-22 | 2019-02-22 | 上海木木聚枞机器人科技有限公司 | A kind of generation method and system of the map calculating pedestrian's probability of occurrence |
| CN109540166A (en) * | 2018-11-30 | 2019-03-29 | 电子科技大学 | A kind of Safe path planning method based on Gaussian process |
| CN109855622A (en) * | 2019-01-07 | 2019-06-07 | 上海岚豹智能科技有限公司 | Method for searching path and equipment for mobile robot |
| CN109855622B (en) * | 2019-01-07 | 2021-06-11 | 上海岚豹智能科技有限公司 | Path searching method and device for mobile robot |
| CN109839110B (en) * | 2019-01-09 | 2020-07-24 | 浙江大学 | Multi-target point path planning method based on rapid random search tree |
| CN109839110A (en) * | 2019-01-09 | 2019-06-04 | 浙江大学 | A kind of multiple target point path planning method based on quick random search tree |
| CN109737970A (en) * | 2019-03-21 | 2019-05-10 | 集美大学 | A kind of unmanned surface vehicle paths planning method based on improvement RRT algorithm |
| CN110083165A (en) * | 2019-05-21 | 2019-08-02 | 大连大学 | A kind of robot paths planning method under complicated narrow environment |
| CN110083165B (en) * | 2019-05-21 | 2022-03-08 | 大连大学 | Path planning method of robot in complex narrow environment |
| CN110490047A (en) * | 2019-07-01 | 2019-11-22 | 珠海格力电器股份有限公司 | A kind of appliances equipment control method, household appliance and computer-readable medium |
| CN110354475A (en) * | 2019-07-16 | 2019-10-22 | 哈尔滨理工大学 | A kind of tennis racket swinging movement pattern training method and device |
| CN110488811B (en) * | 2019-07-22 | 2023-04-07 | 上海有个机器人有限公司 | Method for predicting pedestrian track by robot based on social network model |
| CN110488811A (en) * | 2019-07-22 | 2019-11-22 | 上海有个机器人有限公司 | A method of the robot based on social networks model predicts pedestrian track |
| CN110347162A (en) * | 2019-07-29 | 2019-10-18 | 南京市晨枭软件技术有限公司 | A kind of industrial robot path modified method and system automatically |
| CN110765130A (en) * | 2019-09-25 | 2020-02-07 | 武汉大学 | Ripley's K function-based spatio-temporal POI data point pattern analysis method in distributed environment |
| CN110765130B (en) * | 2019-09-25 | 2021-01-01 | 武汉大学 | Ripley's K function-based spatio-temporal POI data point pattern analysis method in distributed environment |
| CN110703757B (en) * | 2019-10-24 | 2022-09-23 | 海风智能科技(浙江)有限公司 | Energy consumption optimization-oriented high-speed train speed planning method |
| CN110703757A (en) * | 2019-10-24 | 2020-01-17 | 北京航盛新能科技有限公司 | Energy consumption optimization-oriented high-speed train speed planning method |
| CN112087488A (en) * | 2020-08-03 | 2020-12-15 | 济南浪潮高新科技投资发展有限公司 | Method, device, equipment and medium for determining important cloud robot nodes |
| CN112087488B (en) * | 2020-08-03 | 2023-08-25 | 山东浪潮科学研究院有限公司 | Method, device, equipment and medium for determining important cloud robot nodes |
| CN112363505B (en) * | 2020-11-10 | 2022-06-21 | 合肥工业大学 | Articulated sweeper speed planning method and system based on target distance |
| CN114511999B (en) * | 2020-11-17 | 2023-09-01 | 宇通客车股份有限公司 | Pedestrian behavior prediction method and device |
| CN114511999A (en) * | 2020-11-17 | 2022-05-17 | 郑州宇通客车股份有限公司 | Pedestrian behavior prediction method and device |
| CN112828889A (en) * | 2021-01-05 | 2021-05-25 | 佛山科学技术学院 | Six-axis cooperative mechanical arm path planning method and system |
| CN113110482B (en) * | 2021-04-29 | 2022-07-19 | 苏州大学 | Indoor environment robot exploration method and system based on priori information heuristic method |
| CN113110482A (en) * | 2021-04-29 | 2021-07-13 | 苏州大学 | Indoor environment robot exploration method and system based on priori information heuristic method |
| CN113240353A (en) * | 2021-07-12 | 2021-08-10 | 南京擎天全税通信息科技有限公司 | Cross-border e-commerce oriented export factory classification method and device |
| CN113240353B (en) * | 2021-07-12 | 2021-11-02 | 南京擎天全税通信息科技有限公司 | Cross-border e-commerce oriented export factory classification method and device |
| CN115297303A (en) * | 2022-09-29 | 2022-11-04 | 国网浙江省电力有限公司 | Image data acquisition and processing method and device suitable for power grid power transmission and transformation equipment |
| CN115297303B (en) * | 2022-09-29 | 2022-12-27 | 国网浙江省电力有限公司 | Image data acquisition and processing method and device suitable for power grid power transmission and transformation equipment |
| CN116909293A (en) * | 2023-09-13 | 2023-10-20 | 宁德思客琦智能装备有限公司 | Robot path planning method and device, electronic equipment and computer readable medium |
| CN116909293B (en) * | 2023-09-13 | 2023-12-12 | 宁德思客琦智能装备有限公司 | Robot path planning method and device, electronic equipment and computer readable medium |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN106774314A (en) | A kind of home-services robot paths planning method based on run trace | |
| CN105263113B (en) | A kind of WiFi location fingerprints map constructing method and its system based on crowdsourcing | |
| CN103278170B (en) | Based on mobile robot's cascade map creating method that remarkable scene point detects | |
| CN110083165A (en) | A kind of robot paths planning method under complicated narrow environment | |
| CN107423730A (en) | A kind of body gait behavior active detecting identifying system and method folded based on semanteme | |
| CN107272705A (en) | A kind of multiple neural network controlling planning method of robot path under intelligent environment | |
| CN106127125B (en) | Distributed DTW Human bodys' response method based on human body behavioural characteristic | |
| CN104715493A (en) | Moving body posture estimating method | |
| AU2022204569B2 (en) | Method for multi-agent dynamic path planning | |
| CN105243139A (en) | Deep learning based three-dimensional model retrieval method and retrieval device thereof | |
| CN106780484A (en) | Robot interframe position and orientation estimation method based on convolutional neural networks Feature Descriptor | |
| CN109872346A (en) | A kind of method for tracking target for supporting Recognition with Recurrent Neural Network confrontation study | |
| CN102324041B (en) | Pixel classification method, joint body gesture recognition method and mouse instruction generating method | |
| CN108680155A (en) | The robot optimum path planning method of mahalanobis distance map process is perceived based on part | |
| CN103678949B (en) | Density based is analyzed and many Extended target of spectral clustering follow the tracks of measurement collection division methods | |
| CN102722721A (en) | Human falling detection method based on machine vision | |
| CN107045519B (en) | A kind of visualization system of analysis crowd movement law | |
| Wu et al. | Spatial semantic hybrid map building and application of mobile service robot | |
| CN104462662B (en) | Underwater production facility dummy assembly method | |
| CN104777839A (en) | BP neural network and distance information-based robot autonomous obstacle avoiding method | |
| CN106296743A (en) | A kind of adaptive motion method for tracking target and unmanned plane follow the tracks of system | |
| CN109376584A (en) | A kind of poultry quantity statistics system and method for animal husbandry | |
| CN107742097A (en) | A kind of Human bodys' response method based on depth camera | |
| CN103714554A (en) | Video tracking method based on spread fusion | |
| CN107679516A (en) | Lower extremity movement recognition methods based on multiple dimensioned Gauss Markov random field model |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| WD01 | Invention patent application deemed withdrawn after publication |
Application publication date: 20170531 |
|
| WD01 | Invention patent application deemed withdrawn after publication |