AU2021266204A1

AU2021266204A1 - Mobile robot pose correction method and system for recognizing dynamic pallet terminals

Info

Publication number: AU2021266204A1
Application number: AU2021266204A
Authority: AU
Inventors: Xinbiao Gao; Panling HUANG; Kai Song; Di Wu; Xuhao Yang; Yifan Zhao; Huazhang Zhou; Jun Zhou
Original assignee: Shandong University; Shandong Alesmart Intelligent Technology Co Ltd
Current assignee: Shandong University; Shandong Alesmart Intelligent Technology Co Ltd
Priority date: 2021-03-01
Filing date: 2021-11-09
Publication date: 2022-09-15
Anticipated expiration: 2041-11-09
Also published as: CN112935703B; AU2021266204B2; CN112935703A

Abstract

The present disclosure provides a mobile robot pose correction method system for recognizing dynamic pallet terminals. The method includes: obtaining the laser point cloud cluster data of a pallet, and processing the obtained laser point cloud cluster data to obtain pallet leg feature mean characterization points; obtaining a point cloud distribution score and a point cloud shape score according to all the pallet leg feature mean characterization points, and obtaining pallet leg feature point coordinates according to a product of the two scores; and performing pose correction on a mobile robot according to the obtained pallet leg feature point coordinates. The present disclosure overcomes the defect of poor application adaptability due to the inability to correct the pose of a mobile robot in real time resulting from the pose change in a dynamic pallet in the prior art, realizes highly flexible and highly adaptable correction of the pose of the mobile robot, and ensures the efficiency and safety of the operation of the robot. Drawings I|--------------------------------------------------------------------------1 2 5 3 Fig. 9 5/5

Description

Drawings

I|--------------------------------------------------------------------------1

2 5

3

Fig. 9

5/5

Description

MOBILE ROBOT POSE CORRECTION METHOD AND SYSTEM FOR RECOGNIZING DYNAMIC PALLET TERMINALS

Field of the Invention

The present disclosure relates to the technical field of robot pose control, in particular to a mobile

robot pose correction method and system for recognizing dynamic pallet terminals.

Background of the Invention

The statements in this section merely provide a background art related to the present disclosure, but

do not necessarily constitute the prior art.

With the rapid development of computer technology and sensor technology, mobile robots have

gradually become widely used, especially in the logistics industry. When performing

logistics-related operations, a knapsack or forklift mobile robot uses a pallet as a communication

medium with goods, the robot uses a jacking device to lift the pallet to complete the transportation

of a specified path and the loading and unloading of a specified station, and the robots plays an

important role in accurate loading and unloading of the goods, and intelligent carrying.

A traditional pallet placement position is a known and static point in the path of the mobile robot,

which in turn has higher accuracy requirements for the placement pose of the pallet, once a

relatively large error is generated in the pose of the pallet, it is prone to safety issues of collisions

between cars and objects, and a greater impact is generated on the efficiency of the entire

production line. For the pose change of a dynamic pallet terminal, the traditional method cannot

perform an adaptive pose correction action of the mobile robot with the dynamic change of the

pallet, such that the system lacks a certain degree of flexibility and intelligence.

The inventor found that the prerequisite for the application of a mobile robot pose correction system

is to accurately recognize the pallet and determine its position. The existing pallet recognition

methods include: the method of using the RFID technology is simple and reliable, but has the

problems of lower positioning accuracy and flexibility; the method of using visual target

recognition is highly flexible and accurate, but it is vulnerable to environmental light interference

and has worse environmental adaptability; the method of using 2D laser radar recognition has

stronger environmental adaptability and higher position accuracy, but it is necessary to establish a

Description

pallet recognition template in advance, which has worse adaptability to different shelves in different

working scenarios, and the data size of 2D laser point clouds is relatively small, thus being prone to

the phenomenon of false detection; and the method of using 3D laser radar recognition has stronger

adaptability to scenarios and environments, and higher accuracy, but the sensor equipment is more

expensive.

Summary of the Invention

In order to solve the shortcomings of the prior art, the present disclosure provides a mobile robot

pose correction method and system for recognizing dynamic pallet terminals, which overcome the

defect of poor application adaptability due to the inability to correct the pose of a mobile robot in

real time resulting from the pose change in a dynamic pallet in the prior art, realize highly flexible

and highly adaptable correction of the pose of the mobile robot, and ensure the efficiency and safety

of the operation of the robot.

In order to achieve the above objectives, the present disclosure adopts the following technical

solutions:

The first aspect of the present disclosure provides a mobile robot pose correction method for

recognizing dynamic pallet terminals.

A mobile robot pose correction method for recognizing dynamic pallet terminals, including the

following processes:

obtaining the laser point cloud cluster data of a pallet, and processing the obtained laser point cloud

cluster data to obtain pallet leg feature mean characterization points;

obtaining a point cloud distribution score and a point cloud shape score according to all the pallet

leg feature mean characterization points, and obtaining pallet leg feature point coordinates

according to a product of the two scores; and

performing pose correction on a mobile robot according to the obtained pallet leg feature point

coordinates.

As an optional embodiment, the step of obtaining the laser point cloud cluster data of the pallet, and

processing the obtained laser point cloud cluster data includes the following processes:

after the mobile robot stops moving, obtaining at least one frame of scanning point cloud data of a

laser radar to serve as original point cloud data;

Description

eliminating and screening the original point cloud data;

performing feature point classification on the point cloud data after the eliminating and screening

processing, classifying point clouds into line feature points and corner feature points, eliminating

the line feature points, and dividing the corner feature points into breakpoints and corner points;

performing voxel filtering on the breakpoints;

performing windowed pallet feature filtering on the basis of the breakpoints; and

completing clustering search according to a preset pallet search box.

Further, the step of eliminating and screening the original point cloud data includes the following

processes:

eliminating empty points, limiting a rectangular range of a middle distance in front of the laser radar

as a pallet placement area, screening out the point clouds existing in this rectangular range in the

original data to serve as input information for recognition and detection, eliminating the remaining

unscreened data, and keeping the order of the original point cloud data unchanged during the

process.

Further, the step of performing feature point classification on the point cloud data after the

eliminating and screening processing includes the following processes:

classifying the point clouds into points with line features and corner features by establishing a

curvature threshold, traversing the point clouds to give classification information and eliminate the

line feature points, establishing a distribution variance threshold and a curvature threshold for the

screened out corner feature points, and dividing the corner feature points into breakpoints and

corner points again.

Further, the step of performing voxel filtering on the breakpoints includes the following processes:

performing filtering on the point clouds belonging to the same breakpoint, and selecting the

breakpoint with the minimum serial number within a discontinuous point cloud range as breakpoint

characterization.

Further, the step of performing windowed pallet feature filtering on the basis of the breakpoints

includes the following processes:

for the breakpoints obtained by filtering, sequentially selecting and accessing whether there are at

least two breakpoints within a rectangular range, storing breakpoint linked lists that meet the

requirements in a queue, and eliminating the unsatisfied points.

Description

Further, the step of completing clustering search according to the preset pallet search box includes

the following steps:

taking out the breakpoint linked lists in the queue in sequence, calculating a mean of all objects in

the breakpoint linked lists, and using the mean as the central point of the pallet search box, so as to

construct the pallet search box; and

performing a clustering operation on all point clouds in the search box, classifying the point clouds

into linear feature points, pallet leg feature points and invalid points again in combination with the

obtained curvature threshold, and storing the pallet leg feature mean characterization points in all

search boxes.

As an optional embodiment, a circular range is constructed with the obtained pallet leg feature point

coordinates as the central point, and the point cloud distribution score is calculated, and is

specifically the ratio of the number of the same type of clustered point clouds falling in the circular

range to the number of the same type of clustered point clouds.

As an optional embodiment, the acquisition of the point cloud shape score includes:

performing selection and calculation on all the pallet leg feature mean characterization points,

circularly selecting three points in the point clouds, and calculating slope values of connecting lines

of the three characterization points; and

calculating the degree of a right angle between the connecting lines according to the obtained slopes

of the connecting lines, and calculating the pallet leg point cloud shape score through the included

angles among the connecting lines of the three points.

As an optional embodiment, after the two scores are obtained, the pallet leg point cloud shape score

is multiplied by all the pallet leg point cloud distribution scores that constitute a quadrilateral shape

of the pallet to obtain a total score, and the group of point clouds with the highest score is used as

the pallet leg feature point coordinates.

As an optional embodiment, whether there is data meeting the feature information of a reflector is

queried from point cloud reflectance data in a detection frame, if so, and a high reflectance

characterization point is located at the front end of a total score queue, the priority of the

characterization point linked list with the information of the reflector is updated, and the first pallet

characterization point linked list in the total score queue is output after the adjustment is completed.

As an optional embodiment, the step of performing pose correction on the mobile robot according

Description

to the obtained pallet leg feature point coordinates includes the following processes:

according to the obtained pallet leg feature point coordinates, sorting the headstock orientation of

the mobile robot from small to large, and calculating a virtual cut-in cap point and a virtual

termination cap point of the mobile robot;

adjusting the posture of the mobile robot to the virtual cut-in cap point, constructing a virtual path

to cut into the lower side of the pallet, and when the position reaches the virtual termination cap

point, controlling a top end camera module of the mobile robot to recognize an identification code

at the bottom of the pallet; and

after the mobile robot operates normally, returning to a normal operation path of the mobile robot

according to the cut-in path, and deleting the virtual cut-in cap point and the virtual termination cap

point.

The second aspect of the present disclosure provides a mobile robot pose correction system for

recognizing dynamic pallet terminals.

A mobile robot pose correction system for recognizing dynamic pallet terminals, including:

a storage pallet terminal recognition and detection module, configured to: obtain the laser point

cloud cluster data of a pallet, and process the obtained laser point cloud cluster data to obtain pallet

leg feature mean characterization points;

a pallet credibility self-evaluation module, configured to obtain a point cloud distribution score and

a point cloud shape score according to all the pallet leg feature mean characterization points, and

obtain pallet leg feature point coordinates according to a product of the two scores; and

a mobile robot pose correction module, configured to perform pose correction on a mobile robot

according to the obtained pallet leg feature point coordinates.

The third aspect of the present disclosure provides a computer readable storage medium on which a

program is stored, and when executed by a processor, the program implements the steps in the

mobile robot pose correction method for recognizing dynamic pallet terminals in the first aspect of

the present disclosure.

The fourth aspect of the present disclosure provides an electronic device, including a memory, a

processor, and a program stored in the memory and capable of running on the processor, wherein

when the processor executes the program, the program implements the steps in the mobile robot

pose correction method for recognizing dynamic pallet terminals in the first aspect of the present

Description

disclosure.

Compared with the prior art, the present disclosure has beneficial effects as follows:

1. In the method, the system, the medium or the electronic device described in the present

disclosure, coupling processing is performed on machine learning and feature extraction of

scanning data of the laser radar, the laser point clouds are divided into linear features, corner

features and breakpoint features according to the obtained category and structure information, and

an approximate point cloud of the pallet is determined in the form of a sliding window by means of

the features of the pallet and the structural environment, as well as the clustering information in the

search box. Compared with the recognition of a shelf by the traditional laser point cloud, the defect

of building a pallet length template is removed. Compared with the traditional form of RFID, this

solution can intelligently detect the pallet at any dynamically changing position, and no mark point

needs to be set in the path, such that the accuracy is higher, and the degree of flexibility in terminal

operations is higher.

2. In the method, the system, the medium or the electronic device described in the present

disclosure, a pallet credibility self-evaluation function is established to increase the confidence of

recognition, safe and reliable recognition and detection works can be ensured, the system is

endowed with discrimination ability, and the intelligence level is higher.

3. In the method, the system, the medium or the electronic device described in the present

disclosure, for terminal pose correction, the pallet can be used as a reference to correct a cumulative

error during the transportation path process of the mobile robot, the cut-in pose is automatically

adjusted without human intervention, and a corresponding response can be made to the randomly

placed pallet for adjustment, thereby reducing the strictness of workers on the pallet placement on a

production line and improving the fault tolerance of the mobile robot.

Brief Description of the drawings

The drawings constituting a part of the present disclosure are used for providing a further

understanding of the present disclosure. The exemplary embodiments of the present disclosure and

the descriptions thereof are used for explaining the present disclosure, but do not constitute

improper limitations of the present disclosure.

Fig. 1 is a schematic diagram of composition of a mobile robot pose correction system for

Description

recognizing dynamic pallet terminals provided by embodiment 1 of the present disclosure.

Fig. 2 is a total flow diagram of a working method of the mobile robot pose correction system for

Fig. 3 is a flow diagram of the working method of a storage pallet terminal recognition and

detection module in the mobile robot pose correction system for recognizing dynamic pallet

terminals provided by embodiment 1 of the present disclosure.

Fig. 4 is a flow diagram of the working method of a pallet credibility self-evaluation module in the

mobile robot pose correction system for recognizing dynamic pallet terminals provided by

embodiment 1 of the present disclosure.

Fig. 5 is a flow diagram of the working method of a mobile robot pose correction module in the

embodiment 1 of the present disclosure.

Fig. 6 is a schematic structural diagram of a storage pallet used by a backpack mobile robot

provided by embodiment 1 of the present disclosure.

Fig. 7 is a planning map of a laser radar point cloud layout when the mobile robot provided by

embodiment 1 of the present disclosure recognizes the storage pallet.

Fig. 8 is a laser radar scanning point cloud diagram of the storage pallet provided by embodiment 1

of the present disclosure in a hall environment.

Fig. 9 is a schematic diagram of pose correction work after the mobile robot provided by

embodiment 1 of the present disclosure recognizes a dynamic pallet terminal.

Detailed Description of the Embodiments

The present disclosure will be further described below in conjunction with the drawings and

embodiments.

It should be pointed out that the following detailed descriptions are all illustrative, and are intended

to provide further descriptions of the present disclosure. Unless otherwise indicated, all technical

and scientific terms used herein have the same meaning as commonly understood by those of

ordinary skill in the technical field to which the present disclosure belongs.

It should be noted that the terms used here are only for describing specific embodiments, and are

not intended to limit the exemplary embodiments according to the present disclosure. As used

Description

herein, unless the context clearly indicates otherwise, the singular form is also intended to include

the plural form. In addition, it should also be understood that when the terms "comprising" and/or

"including" is used in this specification, it indicates the presence of features, steps, operations,

devices, components, and/or combinations thereof.

In the case of no conflict, the embodiments in the present disclosure and the features in the

embodiments can be combined with each other.

Embodiment 1:

Embodiment 1 of the present disclosure provides a mobile robot pose correction system for

recognizing dynamic pallet terminals. As shown in Fig. 1, the system includes a storage pallet

terminal recognition and detection module, a pallet credibility self-evaluation module and a mobile

robot pose correction module; and

the storage pallet terminal recognition and detection module mainly completes such works as

eliminating and screening key frame laser point clouds, classifying feature points, performing

machine learning and unsupervised learning analysis, and performing window search on a possible

range of the pallet; the pallet credibility self-evaluation module mainly completes such works as

performing online supervision on a pallet leg point cloud self-evaluation function, performing

reflector detection on the basis of a feature tag, and outputting a score queue of credibility priority

updating; and the mobile robot pose correction module constructs a virtual cap point, so that a

mobile robot adaptively adjusts the operation pose.

Fig. 2 shows a working method of the mobile robot pose correction system for recognizing dynamic

pallet terminals. The mobile robot will autonomously plan to move according to a path given by a

scheduling system, when the robot reaches a specified station and needs to perform a terminal

operation, the mobile robot stops the operation and sends a recognized approximate pallet point

cloud cluster queue of a window range to the pallet credibility self-evaluation module through the

storage pallet terminal recognition and detection module, calculates the score of the obtained point

cloud cluster, selects the point cloud information of a group of pallet legs with the highest score to

the mobile robot pose correction module, so as to flexibly complete the pose correction of the robot

in real time, thereby ensuring the efficiency and safety of the operation of the robot.

For the storage pallet terminal recognition and detection module, the module is composed of a laser

radar and industrial cameras on a mobile robot substrate, wherein the laser radar is installed at the

Description

middle position of the front end of the mobile robot with the maximum scanning range and outputs

a point cloud with position information, the two industrial cameras are installed in opposite

directions, the industrial camera at the top end completes the detection of two-dimensional code

information of the terminal pallet, the industrial camera at the bottom end completes the detection

of precise path position information, and the storage pallet terminal recognition and detection

module performs data processing on the point cloud within the recognition range of the laser radar

to obtain a point cloud cluster that conforms to the feature information of the pallet, and confirms

the terminal pallet information through the industrial camera, so as to complete the feedback loop of

the work scheduling system of the mobile robot.

The working method of the storage pallet terminal recognition and detection module , as shown in

Fig. 3, includes the following steps:

step Al: when the mobile robot runs to a specified lifting operation station on the specified path of

the scheduling system, the robot stops moving, and stores a frame of scanning point cloud data of

the laser radar at the same time to serve as original data for completing the subsequent processing.

Fig. 5 shows the pallet used by a knapsack mobile robot, the bottom end of the pallet is supported

by four legs, the information of the pallet in the original data of the scanning point cloud of the

mobile robot is represented by four leg point clouds, 1 in the figure represents an information

two-dimensional code of the pallet, which is used for feeding back the transportation information to

the scheduling system and feeding back that the lifting operation is running as scheduled, and 2 in

the figure represents a reflector with high reflectance, which is used as a feature tag of the pallet to

enhance the recognizability of the pallet.

Step A2: screening and eliminating the original point cloud data.

There are empty points in the original data (the x and y coordinates output by points that are not

detected are all 0), and the empty points are eliminated to reduce the subsequent calculation amount;

and the density of laser point clouds has the nature of becoming sparse as the distance increases, so

a rectangular range of a middle distance in front of the laser radar is limited as a pallet placement

area.

Fig. 7 shows a schematic diagram of a screening and eliminating range of the laser point clouds

when the mobile robot is at a recognition position, wherein the area 1 represents a pallet parking

range with a length of 28.28m and a width of 14.14m, this range is consistent with the pallet

Description

parking range in the factory, which ensures the accuracy of subsequent pallet recognition, the area 2

represents a laser point cloud removal part, since there is an inverse proportional function for the

density and distance of the laser point clouds, it is difficult to recognize the pallet features of too far

points, which is likely to cause safety problems, the area 3 represents a laser point cloud removal

part, and the pallet is often stored in front of the robot body, so the point clouds in this area are

eliminated.

The point clouds existing in this rectangular range area 2 in the original data are screened out as the

input information for recognition and detection, the remaining unscreened data is eliminated, and

the order of the original point cloud data remains unchanged during the process.

Step A3: processing the input data to complete feature point classification.

The input data is used for calculating curvatures, means and variance values of corresponding point

cloud coordinates in the range of a neighborhood 3, and the calculation formulas are as follows:

cur nj+3

n 3 .

n-j+3

i-j-3

n Ii-j-3 yvar = -- 2

wherein, cur , x, , Y and Y respectively represent the curvatures, x coordinate means, y

coordinate means, x coordinate variances and y coordinate variances of the point clouds, and the

subscript j represents the serial number of the current point cloud. By establishing a curvature

threshold 0.01m, the point clouds are classified as points with line features and corner features, the

point clouds are traversed to give classification information and eliminate the line feature points, a

distribution variance threshold and a curvature threshold are established for the screened out corner

feature points, and the corner feature points are divided into breakpoints and corner points again.

Fig. 8 shows a laser radar scanning point cloud map of the storage pallet in a hall environment, the

Description

area 1 in the figure represents the screened out linear features, and the main characterization points

of the pallet are breakpoints and corner points, so the linear features are eliminated, 2 represents the

position of the laser radar in the laser point cloud, 4 in the figure represents a corner point, the

corner points are mostly straight turning points of obstacles, such as a wall corner, 5 in the figure

represents a breakpoint, that is, a discontinuous point, which is mainly represented as that the laser

point cloud is reflected from one object to another object.

Step A4: performing voxel filtering on breakpoint feature points.

The breakpoints at discontinuities in the point cloud are repeated because of similar curvatures, the

point clouds belonging to the same breakpoint are filtered, and the breakpoint with the minimum

serial number within the range of the discontinuous point cloud is selected as the breakpoint

characterization.

Step A5: performing windowed pallet feature filtering on the basis of the breakpoints.

For the breakpoints obtained by the above filtering processing, whether there are at least two

breakpoints in a 4x4m2 rectangular range is sequentially selected and accessed, breakpoint linked

lists meeting the requirements are stored in a queue, and the unsatisfied points are eliminated. By

means of this step, structure body edge breakpoints and obstacle edge breakpoints can be effectively

distinguished.

Step A6: setting a pallet search box and completing clustering search.

The breakpoint linked lists in the queue are taken out in sequence, the mean of all objects in the

breakpoint linked lists is calculated to serve as the central point of the pallet search box, so as to

construct the pallet search box to specify the approximate range of the pallet; and KD-TREE is

established for all point clouds in the search box, so that the progress of the DBSCAN clustering

operation can be accelerated, the point clouds are divided into linear feature points, pallet leg

feature points and invalid points again in combination with the curvature threshold in step A3, the

serial number 3 in Fig. 8 shows this type of point clouds, that is, the pallet legs, and the number of

the point clouds is mostly greater than two, the pallet leg feature mean characterization points in all

search boxes are stored, and it is ensured that the number of pallet leg characterization points in

each group is at least 2.

The pallet credibility self-evaluation module constructs a pallet feature self-evaluation function, and

performs score estimation processing on the pallet leg feature mean characterization points output

Description

by the storage pallet terminal recognition and detection module. This module no longer builds a

distance template when the traditional laser radar recognizes the pallet, but calculates and sorts

pallet leg characterization point linked lists that meet the requirements in all the search boxes output

by the above module, and uses the group of point clouds with the highest score as pallet leg feature

point coordinates. In order to ensure the stability and safety of module recognition, the highest score

feature points of 10 frames of point clouds are checked and compared, and an optimal solution is

obtained according to the Gaussian distribution. This method has strong adaptability and high

degree of flexibility.

The working method of the pallet credibility self-evaluation module, as shown in Fig. 4, includes

the following steps:

Step B1: calculating a pallet leg point cloud distribution score.

A circular range is constructed with the pallet leg characterization point obtained by clustering as

the central point, a classification score is calculated, and the calculation formula is as follows:

count dis score = range clust sum

wherein, dis _ score represents the point cloud distribution score of a pallet leg characterization

point, range_ count represents the number of the same type of clustered point clouds falling into

the circular range, clust-sum represents the number of the same type of clustered point clouds,

and the higher the score is, the greater the possibility of deeming this type of point clouds as the

pallet legs is.

Step B2: calculating a pallet leg point cloud shape score.

All the pallet leg feature mean characterization points are selected and calculated, three points of the

point cloud are circularly selected, the slope values of the connecting lines of the three

characterization points are calculated, and the calculation formula is as follows:

k x-x .

wherein, k , x and Y respectively represent the slope of the connecting line, the abscissa of the

characterization point and the vertical coordinate of the characterization point, after the slope of the

connecting line is calculated, the degree of a right angle formed between the connecting lines is

Description

calculated, and the calculation formula is as follows:

(k - k, ) 180 K angle =abs arctan y(1+ k~x k,)) xkk, 1 pi

wherein, angle represents the included angle between the connecting lines, the pallet leg point

cloud shape score is calculated through the included angles among the line segments of the three

points, and the calculation formula is as follows:

ranglel, ane9g0 angle.e (0,90),i = min(relative angle) shape-score=agi-9 angle, -9 else

wherein, shape _ score represents the pallet leg point cloud shape score, i represents a connecting

line angle index value with the minimum relative angle value, relative_ angle represents the

relative angle value of the included angle between the connecting lines, and the score can be used

for representing the degree of approximation of a quadrangle formed by the point cloud

characterization points, and the higher the score is, the more square the quadrangle is.

Step B3: calculating the total score of the pallet characterization point linked lists in the queue,

multiplying the pallet leg point cloud shape score with all the pallet leg point cloud distribution

scores that form the quadrilateral shape of the pallet, so as to obtain the total score. The higher the

total score is, the greater the possibility that the characterization point queue is deemed as the pallet

feature point cloud is.

Step B4: querying the features of a reflector and updating the priority. Whether there is data

meeting the feature information of the reflector is queried from point cloud reflectance data in a

detection frame, if so, and a high reflectance characterization point is located at the front end of a

total score queue, the priority of the characterization point linked list with the information of the

reflector is updated, and the first pallet characterization point linked list in the total score queue is

output after the adjustment is completed.

For the mobile robot pose correction module, the mobile robot corrects the cut-in position of the

mobile robot according to the recognized position of the pallet, so as to ensure safe and reliable

lifting operation of the mobile robot. The mobile robot pose correction module, as shown in Fig. 5,

includes the following steps:

Description

Step Cl: calculating a virtual cut-in cap point and a virtual termination cap point of the mobile

robot. The coordinate points output by the pallet credibility self-evaluation module are sorted from

small to large according to the headstock orientation of the mobile robot, a virtual cut-in cap point

and a virtual termination cap point of the mobile robot are calculated, and the calculation formulas

are as follows:

entry cap= X1 2 0(y

) entry angle = arctan_ X1 iX 2

diagonal y end cap= (diagonal _x - 2 2

wherein, entry -cap represents the virtual cut-in cap point, which is used for adjusting the position

oftherobot, entryangle represents a virtual cut-in angle, which is used for adjusting the pose of

therobot, end cap represents a virtual pallet terminal parking point of the mobile robot, and

diagonalx and diagonal-y represent diagonal coordinate element values.

Step C2: adjusting the pose of the mobile robot and performing operations. Fig. 9 shows a

schematic diagram of the actual work of the mobile robot, the serial number 1 in the figure is

end_ cap, which is calculated from the pallet leg point cloud coordinates and is used for the

accurate parking of the mobile robot to ensure high-precision operations, the serial number 2 in the

figure is the position where the mobile robot detects and recognizes the pallet, the position contains

two-dimensional code information, which records the actions to be executed by the mobile robot,

the serial number 3 in the figure represents the cap point on the normal working path of the mobile

robot, which is used by the robot to record key positions and local navigation, the serial number 4 in

the figure is a parking area of the pallet, the serial number 5 in the figure is entry cap , the dotted

line in the figure represents the specified path of the scheduling system, and the thin-dotted line

represents a virtual path when the mobile robot lifts the pallet. An upper computer of the mobile

robot controls a lower computer to adjust the pose to entry_ cap, and constructs the virtual path to

cut into the lower side of the pallet, when the position reaches the point end cap , the industrial

Description

camera at the top end of the mobile robot recognizes a two-dimensional code at the bottom of the

pallet, feeds it back to the scheduling system of the robot, and returns to the normal operation path

of the mobile robot according to the cut-in path after the robot works normally, the two types of

virtualcappoints,thatis,entrycap andendcap, are deleted, so as to realize the flexible and

intelligent terminal pallet jacking operation of the mobile robot without human intervention, which

improves the fault tolerance and robustness of the terminal operations of the entire mobile robot.

Embodiment 2:

Embodiment 2 of the present disclosure provides a mobile robot pose correction method for

recognizing dynamic pallet terminals, including the following processes:

cluster data to obtain pallet leg feature mean characterization points;

according to a product of the two scores; and

coordinates.

SI: the obtaining the laser point cloud cluster data of the pallet, and processing the obtained laser

point cloud cluster data to obtain pallet leg feature mean characterization points includes the

following processes:

Step A2: screening and eliminating the original point cloud data.

There are empty points in the original data ( the x and y coordinates output by points that are not

area, the point clouds existing in this rectangular range area in the original data are screened out as

the input information for recognition and detection, the remaining unscreened data is eliminated,

and the order of the original point cloud data remains unchanged during the process.

Description

Step A3: processing the input data to complete feature point classification.

cur nj+3

xx n 3

. n-j+3

ni-j-3 1=n-3

xvar= x n-i-j3

I n--3 yvar = -- 2 n I j-3

wherein, cur, x, y, Y and Y respectively represent the curvatures, x coordinate means, y

threshold, the point clouds are classified as points with line features and corner features, the point

clouds are traversed to give classification information and eliminate the line feature points, a

Step A4: performing voxel filtering on breakpoint feature points.

characterization.

breakpoints in the rectangular range is sequentially selected and accessed, breakpoint linked lists

meeting the requirements are stored in a queue, and the unsatisfied points are eliminated. By means

of this step, structure body edge breakpoints and obstacle edge breakpoints can be effectively

Description

distinguished.

Step A6: setting a pallet search box and completing clustering search.

construct the pallet search box to specify the approximate range of the pallet; and a DBSCAN

clustering operation is performed on all the point clouds in the search box, the point clouds are

divided into linear feature points, pallet leg feature points and invalid points again in combination

with the curvature threshold in step A3, the pallet leg feature mean characterization points in all

each group is at least 2.

S2: the obtaining the point cloud distribution score and the point cloud shape score according to all

the pallet leg feature mean characterization points, and obtaining the pallet leg feature point

coordinates according to the product of the two scores includes the following processes:

Step B1: calculating the pallet leg point cloud distribution score.

count dis score = range - clust sum

pallet legs is.

Step B2: calculating the pallet leg point cloud shape score.

k x,-x.

Description

calculated, and the calculation formula is as follows:

(k - k, ) 180 angle =abs arctan (kk -1 K y(1+ k~x k,) x pi

points, and the calculation formula is as follows:

angle an_g0 angle.e (0,90),i = min(relative angle) shape-score=agl-9 - angle, -9 else

Step B3: calculating the total score of the pallet characterization point linked lists in the queue.

The pallet leg point cloud shape score is multiplied with all the pallet leg point cloud distribution

feature point cloud is.

Step B4: querying the features of a reflector and updating the priority.

Whether there is data meeting the feature information of the reflector is queried from point cloud

reflectance data in a detection frame, if so, and a high reflectance characterization point is located at

the front end of a total score queue, the priority of the characterization point linked list with the

information of the reflector is updated, and the first pallet characterization point linked list in the

total score queue is output after the adjustment is completed.

S3: the performing pose correction on the mobile robot according to the obtained pallet leg feature

Description

point coordinates includes the following processes:

are as follows:

entry cap= X1 2 0(y

) entry angle = arctan x1 x2 Y1 -Y 2

. _x diagonal y end cap=(diagonal ,I - 2 2

diagonalx and diagonal-y represent diagonal coordinate element values.

Step C2: adjusting the pose of the mobile robot and performing operations.

An upper computer of the mobile robot controls a lower computer to adjust the pose toentrycap

and constructs a virtual path to cut into the lower side of the pallet, when the position reaches the

point endcap, the industrial camera at the top end of the mobile robot recognizes a

two-dimensional code at the bottom of the pallet, feeds it back to the scheduling system of the robot,

and returns to the normal operation path of the mobile robot according to the cut-in path after the

robot works normally, and the two types of virtual cap points, that is, entrycap and

end_ cap ,are deleted.

Embodiment 3:

Embodiment 3 of the present disclosure provides a computer readable storage medium on which a

Description

mobile robot pose correction method for recognizing dynamic pallet terminals in embodiment 2 of

the present disclosure.

Embodiment 4:

Embodiment 4 of the present disclosure provides an electronic device, including a memory, a

pose correction method for recognizing dynamic pallet terminals in embodiment 2 of the present

disclosure.

Those skilled in the art should understand that the embodiment of the present disclosure can be

provided as a method, a system or a computer program product. Accordingly, the present disclosure

can adopt the form of a hardware embodiment, a software embodiment, or an embodiment

combining software with hardware. Moreover, the present disclosure can adopt the form of a

computer program product which is implemented on one or more computer usable storage media

(including, but not limited to, a magnetic disk memory and an optical memory and the like)

containing computer usable program codes.

The present disclosure is described in accordance with the flow diagram and/or block diagram of

the method, the device (system) and the computer program product in the embodiments of the

present disclosure. It should be understood that computer program instructions can realize each

flow and/or block in the flow diagram and/or the block diagram and the combination of the flows

and/or blocks in the flow diagram and/or the block diagram. These computer program instructions

can be provided to a general-purpose computer, a special-purpose computer, an embedded

processor or processors of other programmable data processing devices to generate a machine, such

that the instructions executed by the computers or the processors of the other programmable data

processing devices generate apparatuses used for achieving specified functions in one or more flows

of the flow diagram and/or one or more blocks of the block diagram.

These computer program instructions can also be stored in a computer readable memory that is

capable of guiding the computers or the other programmable data processing devices to work in

particular manners, such that the instructions stored in the computer readable memory generate

products including instruction apparatuses, and the instruction apparatuses achieve the specified

functions in one or more flows of the flow diagram and/or one or more blocks of the block diagram.

Description

These computer program instructions can also be loaded on the computers or the other

programmable data processing devices, so as to execute a series of operation steps on the computers

or the other programmable data processing devices to produce processing achieved by the

computers, such that the instructions executed on the computers or the other programmable data

processing devices provide steps used for achieving the specified functions in one or more flows of

the flow diagram and/or one or more blocks of the block diagram.

Those of ordinary skill in the art can understand that all or a part of flows in the above-mentioned

method embodiment can be implemented by a computer program instructing corresponding

hardware, the program can be stored in a computer readable storage medium, and when being

executed, the program can include the flows of the above-mentioned method embodiments. The

storage medium can be a magnetic disk, an optical disk, a read-only memory (Read-Only Memory,

ROM), a random access memory (Random Access Memory, RAM), etc.

The above descriptions are only preferred embodiments of the present disclosure and are not

intended to limit the present disclosure. For those skilled in the art, the present disclosure can have

various modifications and changes. Any modifications, equivalent replacements, improvements and

the like, made within the spirit and principle of the present disclosure, shall be included in the

protection scope of the present disclosure.

Claims

1. A mobile robot pose correction method for recognizing dynamic pallet terminals, comprising the

following processes:

cluster data to obtain pallet leg feature mean characterization points;

according to a product of the two scores; and

coordinates.

2. The mobile robot pose correction method for recognizing dynamic pallet terminals of claim 1,

wherein:

the step of obtaining the laser point cloud cluster data of the pallet, and processing the obtained

laser point cloud cluster data comprises the following processes:

laser radar to serve as original point cloud data;

eliminating and screening the original point cloud data;

performing voxel filtering on the breakpoints;

completing clustering search according to a preset pallet search box.

3. The mobile robot pose correction method for recognizing dynamic pallet terminals of claim 2,

wherein:

the step of eliminating and screening the original point cloud data comprises the following

processes:

Claims

process; or, the step of performing feature point classification on the point cloud data after the eliminating and

screening processing comprises the following processes:

corner points again;

or,

the step of performing voxel filtering on the breakpoints comprises the following processes:

characterization;

or,

the step of performing windowed pallet feature filtering on the basis of the breakpoints comprises

the following processes:

requirements in a queue, and eliminating the unsatisfied points;

or,

the step of completing clustering search according to the preset pallet search box comprises the

following steps:

construct the pallet search box; and

search boxes.

Claims

4. The mobile robot pose correction method for recognizing dynamic pallet terminals of claim 1,

wherein:

a circular range is constructed with the obtained pallet leg feature point coordinates as the central

point, and the point cloud distribution score is calculated, and is specifically the ratio of the number

of the same type of clustered point clouds falling in the circular range to the number of the same

type of clustered point clouds.

5. The mobile robot pose correction method for recognizing dynamic pallet terminals of claim 1,

wherein:

the acquisition of the point cloud shape score comprises:

of the three characterization points; and

angles among the connecting lines of the three points.

6. The mobile robot pose correction method for recognizing dynamic pallet terminals of claim 1,

wherein:

after the two scores are obtained, the pallet leg point cloud shape score is multiplied by all the pallet

leg point cloud distribution scores that constitute a quadrilateral shape of the pallet to obtain a total

score, and the group of point clouds with the highest score is used as the pallet leg feature point

coordinates;

or,

whether there is data meeting the feature information of a reflector is queried from point cloud

total score queue is output after the adjustment is completed.

7. The mobile robot pose correction method for recognizing dynamic pallet terminals of claim 1,

wherein:

the step of performing pose correction on the mobile robot according to the obtained pallet leg

Claims

feature point coordinates comprises the following processes:

termination cap point of the mobile robot;

at the bottom of the pallet; and

point.

8. A mobile robot pose correction system for recognizing dynamic pallet terminals, comprising:

leg feature mean characterization points;

according to the obtained pallet leg feature point coordinates.

9. A computer readable storage medium on which a program is stored, wherein when executed by a

processor, the program implements the steps in the mobile robot pose correction method for

recognizing dynamic pallet terminals of any one of claims 1-7.

10. An electronic device, comprising a memory, a processor, and a program stored in the memory

and capable of running on the processor, wherein when the processor executes the program, the

program implements the steps in the mobile robot pose correction method for recognizing dynamic

pallet terminals of any one of claims 1-7.