KR20120102998A - Positional recognition aparatus and method for autonomous mobile robot using for rfid tag and positional recognition system - Google Patents

Abstract

PURPOSE: An apparatus, a method and a system for recognizing the position of an autonomous moving robot are provided to control a route by using a sensor model of an RFID tag and a prediction model of a robot. CONSTITUTION: An RFID tag map is a data set storing information about tags arranged in a grid shape. The tag information is composed of specific IDs for distinguishing between the tags. An RFID reader is a sensor for sensing the tag. An actuator is drive unit for moving a robot. A particle filter employs a prediction model using the moving amount of the robot, and a sensor model of the recognized RFID sensor. [Reference numerals] (AA) RFID Tag map; (BB) Path data; (CC) Particle filter; (DD) Path following unit; (EE) RFID reader; (FF) Actuator

Description

POSITIONAL RECOGNITION APARATUS AND METHOD FOR AUTONOMOUS MOBILE ROBOT USING FOR RFID TAG AND POSITIONAL RECOGNITION SYSTEM}

The present invention relates to an autonomous driving position recognition apparatus, a method and a position recognition system of a mobile robot using an RFID tag.

In an intelligent robot, autonomous driving, which can move while avoiding obstacles to its destination, plays an important role. In order to perform autonomous driving smoothly, location recognition technology is essential. For this reason, various location recognition methods have been studied and developed as products.

Dead-reckoning position estimation method using Odometer information measured from wheel encoder is low cost and high sampling rate but error accumulates over time due to wheel slip and integration error. To reduce this cumulative error, a sensor that can measure the absolute position of the robot is required. Map matching based location recognition method using laser scanner, image based landmark and camera using method (StarGazer) and ultrasonic satellite using method (U-SAT) have been studied and developed.

However, StarGazer, a commercially available product, generates an error of a recognized landmark when the camera sensor is tilted, resulting in a large position error. In addition, if the camera receiving the landmark is obscured by an obstacle, the location is not recognized.

Since U-SAT also has four ultrasonic satellites mounted on the ceiling sequentially, the positional error increases due to the time difference between the ultrasonic waves reaching the moving robot. That is, as the speed of the robot increases, the error of the final recognized position increases. Also, similar to StarGazer, if there is an obstacle between the ultrasonic satellites and the receiver, the error of the recognized position increases.

An object of the present invention is to provide an autonomous driving position recognition apparatus, a method and a position recognition system of a mobile robot using an RFID tag that recognizes a current position using an RFID tag and follows a path.

According to an aspect of the present invention, it is possible to provide an autonomous driving position recognition apparatus of a mobile robot.

According to the exemplary embodiment of the present invention, the particle position filter reduces the current position and the estimated position error of the robot, thereby ensuring reliable position recognition performance.

1 is a block diagram schematically showing an autonomous driving position recognition apparatus of a mobile robot using an RFID tag according to an embodiment of the present invention.
2 is a graph illustrating a sensor model of an RFID tag.
3 is route data from a current position to a destination represented on a map.
Figure 4 is a graph showing the mechanism of the mobile robot
5 is a diagram illustrating a path estimation result using a control loop.
6 is a flowchart illustrating an autonomous driving method of a mobile robot according to an embodiment of the present invention.
7 is a graph of the sensor model.
8 is a diagram illustrating tag strips arranged in a line;
9 is an exemplary view of a tag strip shown at the bottom.
10 is a diagram illustrating an environment in which an RFID tag is installed.
11 to 14 are views illustrating a process of estimating a position by a robot.
15 is a photograph showing an RFID reader and an antenna mounted to a robot.
Figure 16 is a photograph showing an RFID tag installed on the floor.
17 is a photograph showing a StarGazer and a U-SAT mounted to a robot.
18 is a photograph illustrating an ultrasonic satellite transmitter installed on a ceiling.
19 is a photograph showing a landmark of StarGazer installed on the ceiling.
20 is a view showing an arrangement form of an RFID tag and a robot movement path.
21 is a graph showing the distance error between the estimated position and the actual position of the RFID position recognition.
Fig. 22 is a graph showing the error of the estimated direction angle of the RFID position recognition and the direction angle of the actual position.
23 is a graph showing the distance error between the RFID, StarGazer, U-SAT position and the actual position.
24 is a graph showing the error between the direction and the actual direction of the RFID, StarGazer, U-SAT.

The present invention may be variously modified and have various embodiments, and specific embodiments will be illustrated in the drawings and described in detail with reference to the accompanying drawings. It should be understood, however, that the invention is not intended to be limited to the particular embodiments, but includes all modifications, equivalents, and alternatives falling within the spirit and scope of the invention.

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS Hereinafter, the present invention will be described in detail with reference to the accompanying drawings. In addition, numerals (e.g., first, second, etc.) used in the description of the present invention are merely an identifier for distinguishing one component from another.

Also, in this specification, when an element is referred to as being "connected" or "connected" with another element, the element may be directly connected or directly connected to the other element, It should be understood that, unless an opposite description is present, it may be connected or connected via another element in the middle.

Hereinafter, an autonomous driving position recognition apparatus, a method and a position recognition system of a mobile robot using an RFID tag according to embodiments of the present invention will be described in detail with reference to the accompanying drawings.

1 is a block diagram schematically illustrating an autonomous driving position recognition apparatus of a mobile robot according to an exemplary embodiment of the present invention.

Referring to FIG. 1, an autonomous driving position recognition apparatus of a mobile robot according to an embodiment of the present invention may include an RFID tag map, route data, particle filters, a path follower, an RFID reader, and an actuator.

RFID tag maps, route data, particle filters, RFID readers and actuators are components for robot location recognition.

The RFID Tag Map is a data set that stores the information of the tags arranged in a grid in the real environment. The tag information consists of a unique ID for distinguishing tags, a position x, y value arranged in a real environment, and an antenna recognition radius distance value of an RFID reader. The RFID Reader is a sensor for recognizing a tag, and the actuator is a driving unit for moving the robot.

The particle filter uses a predictive model using a robot movement amount and a sensor model of a recognized RFID tag. Particle Fitler's prediction model for the amount of movement of the robot is shown in Equation 1.

[Equation 1]

When Equation 1 changes from time t to t + 1, the particles move in addition to the amount of movement error multiplied by the movement error to a random value obtained by the robot. At this time, the movement error is an error constant for the straight distance and the rotation angle caused by the frictional force of the robot wheel and the error of the mechanism. The larger the error constant, the larger the error during movement, so that the particles spread more.

Next, the sensor model of the particle filter RFID tag is shown in Equation 2.

&Quot; (2) "

In Equation 2, S is the recognition range of the RFID tag, is the position of the tag on the map, and represents any position on the map. Therefore, the probability for the position within the recognition range of the RFID tag has a maximum and the position twice the recognition range has a probability of zero. Equation 2 is shown in a graph as shown in FIG.

Particle filters are performed using the two models above. Particle filter execution is initialization, prediction, update, normalization, resampling, and position estimation.

Initialize creates particles at the location of the RFID tag that the robot first recognizes. The position of the generated particles is generated within the recognition range at the position of the recognized RFID tag and the angle has an arbitrary value. Second, Prediction measures the amount of movement of the robot and uses the predictive model to calculate the next position of the particles. Third, Update calculates the probability of each particle using the sensor model for the next recognized RFID tag. At this time, the probability of particles in the recognition range of the recognized RFID tag is high, and the probability of particles outside the range is small. Fourth, Normailze calculates that the probability is 1 when the particles are combined. Fifth, Resampling uses the probability of the final calculated particles to discard particles smaller than a certain probability value, leaving only the particles with the highest probability. Regenerate as many as the first few particles for the remaining particles. Sixth, estimate the position of the final robot by averaging the position and angle for all particles. Finally, we estimate the position of the robot by repeating the particle filter's Prediction to Position Estimates.

The route data, route follower and actuator are components that follow the route.

The route data is a data set of intermediate points on the route that the mobile robot passes from its current position to its final destination position. This route data set can be created automatically by the user by specifying any location or by using final destination location information from the current location.

The route data set includes positions x, y, theta (direction) on the route, including the current and destination positions, and the speed at which the robot moves.

The path follower takes the intermediate point to be reached from the current position from the path data, calculates the forward speed and rotation speed for moving the robot to the intermediate point, and finally delivers it to the actuator.

로 나태낼 수 있다. There are many ways to drive a mobile robot, but in this study, we used a two-wheel differential drive. The two-wheel differential drive is the coordinate system speed and the linear speed

Can be represented.

Equation 3 shows a kinematic equation of the differentially driven mobile robot.

&Quot; (3) "

The polar coordinates of the next destination at the current position of the robot are shown in Equation 4.

&Quot; (4) "

Equations 3 and 4 can be expressed as Equation 5.

[Equation 5]

The control loop for moving the robot from the path follower to the next destination can express the robot's forward and rotational speeds with error and gain constants from the current location to the next destination. Equation 6 shows a control loop equation.

&Quot; (6) "

The following control loop equation (7) can be obtained from equations (4) and (6).

[Equation 7]

는 직진에 대한 이득 상수, 목적지 방향에 대한 이득 상수, 자세에 대한 이득 상수이다. 이득 상수는 시뮬레이션을 통해 최적의 값을 찾는다.In equation (7)

Is the gain constant for going straight, the gain constant for the destination direction, and the gain constant for attitude. The gain constant finds the optimal value through simulation.

The path follower curves through the intermediate destination to the final destination using a control loop. 5 is a result of simulation using a control loop equation.

6 is a flowchart illustrating an autonomous driving method of a mobile robot according to an embodiment of the present invention.

Referring to FIG. 6, in the autonomous driving method of a mobile robot according to an exemplary embodiment of the present disclosure, an initialization step, an estimating position and direction of each particle, an update of a weight of each particle, a new particle proportional to the weight And generating the estimated position of the final robot.

Here, the step of estimating the position and direction of each particle, updating the weight of each particle, generating a new particle proportional to the weight, and calculating the estimated position of the final robot may be repeated.

When the RFID reader mounted on the robot recognizes the RFID tag in the initialization step, particles as many as the maximum number of particles (M) are generated. The location of each particle specifies a random value within the recognition radius around the location where the RFID tag is recognized, and the direction specifies a random value from 0 degree to 360 degrees.

로 표시하고 가중치를

로 표시할 때, M개의 파티클에 대한 집합은 수학식 8과 같이 표시된다.
To position the particles

And weight

When expressed as, the set of M particles is expressed as in Equation (8).

[Equation 8]

)과 이동량에 비례하는 표준편차(

)의 Gaussian 모델을 가지는데, 실제 로봇의 주행을 반복실험하여 계산한다.Each particle is moved according to the movement amount of the robot. At this time, the movement noise of the robot is added. Moving noise resulting from odometry errors is averaged (

) And the standard deviation (proportional to the amount of movement)

It has Gaussian model of), and it calculates by repeating experiment of actual robot's driving.

이고 왼쪽 바퀴의 회전량이

이며,

은 바퀴의 반지름,

은 양쪽 바퀴간의 거리일 때, 로봇의 이동거리

와 회전각도

는 수학식 9와 같다.
The amount of rotation of the right wheel

And the amount of rotation of the left wheel

,

Silver radius of the wheel,

Is the moving distance of the robot when the distance between the wheels

And rotation angle

Is the same as Equation 9.

[Equation 9]

와 회전각도에 대한 노이즈

은 평균 (

)과 표준편차 (

)를 가지는 정규분포로부터 무작위로 임의의 숫자를 선택한다.
Next, the step of estimating the position and direction of each particle adds the moving distance and rotation angle noise of the robot to each particle. Noise for travel

And noise for rotation angle

Is average (

) And standard deviation (

Select a random number from a normal distribution with).

&Quot; (10) "

의 위치와 방향은 수학식 11과 같이 업데이트 된다.
Each particle

The position and direction of are updated as in Equation (11).

&Quot; (11) "

와 인식된 태그의 위치

간의 거리

를 수학식 12를 통해 계산한다.
Next, updating the weight of each particle is the position of the particle with respect to each particle.

And location of recognized tags

Distance between

Is calculated through Equation 12.

[Equation 12]

를 계산한다. (

는 태그의 인식 반경)
Then, the probability from the relational expression

. (

Is the recognition radius of the tag)

&Quot; (13) "

Equation 13 shows the probability of recognizing the tag according to the distance away from the tag, as shown in FIG.

The weight of each particle is updated as shown in Equation 14 so that the sum of the weights of all particles is 1.

&Quot; (14) "

에 대한 예측과 업데이트 과정이 끝나 새로운 상태인 수학식 15와 같이 파티클 집단이 정의된다.
Then, a set of existing particles

After the prediction and update process for, the particle group is defined as shown in Equation (15).

&Quot; (15) "

을 그대로 물려받는다.Next, generating a new particle proportional to the weight generates new particles proportional to the weight of each particle. In other words, particles with lower weights are destroyed, and larger weights generate more new particles. New particles are the characteristics of existing particles

Take over as it is.

은 수학식 16과 같이 계산한다.

는 실수를 정수로 변환하는 연산자다.
The number of new particles the particle generates

Is calculated as in equation (16).

Is an operator that converts a real number to an integer.

&Quot; (16) "

를 계산한다.
Finally, the step of calculating the estimated position of the final robot is to add the product of the weights and positions of all the particles, the estimated position of the final robot as shown in Equation 17

.

&Quot; (17) "

는 -180도 에서 +180도의 값을 가지므로 +영역에 있는 각도와 -영역에 있는 각도를 따로 계산하는데, -영역의 값을 기준으로 +영역의 값과 -영역의 값의 차를 더하여 평균을 계산한다.
Attention should be paid to the operation of. For example, the average of -180 and 180 plus the mean is 0, but the mean should actually be 180.

Since the value is between -180 and +180 degrees, the angle in the + area and the angle in the-area are calculated separately, and the average is obtained by adding the difference between the value of the + area and the-area based on the value of the-area. Calculate

The autonomous driving system of the mobile robot according to an embodiment of the present invention performs the method shown in FIG. 6, recognizes an RFID tag installed on the floor, reads the movement amount from the encoder of the wheel, and matches the two position data with the particle filter. Perform robot position recognition. This system can vary the positioning method and density of RFID on the floor and adjust the position recognition accuracy according to each space.

To this end, the autonomous driving position recognition apparatus of the mobile robot may include an RFID tag map, route data, particle filter, route follower, RFID reader and encoder.

Particle filters are a sequential Monte Carlo algorithm. In other words, the probability distribution of the algorithm is expressed as particles. The location estimation of the particle filter using RFID estimates the location of the robot using the information of the robot's odometry and the location information of the tags recognized by the RFID reader. The particle filter algorithm enables the robot's direction recognition and the position estimation more precisely than the recognition radius of the RFID tag.

In an embodiment of the present invention, the RFID reader widens the diameter of the antenna to increase the recognition diameter and distance without increasing the output. The RFID reader can be designed with a recognition distance and diameter of up to 15cm, but it can be changed by the user in various ways.

In the case of an RFID antenna, the antenna lines may have the same length but may have a rectangular shape or a circular shape.

Here, when using two or more RFID readers to increase the recognition radius, each antenna is disposed at least 50 cm (at least 30 cm) to avoid interference.

The RFID tag uses a coin type that conforms to the ISO 15693 (ISO18000-3 Mode1) 13.56 MHz RFID standard.

The RFID positioning system can vary the positioning method and the density of the RFID on the floor and can precisely adjust the position recognition according to each space. As an example of batch density, in areas where the robot needs to have high positional accuracy due to work, etc., the positioning of the RFID tag can be densified to increase the positional accuracy and the corridor or square where the robot passes to move to the next destination. In areas such as this, there is no need for precise positioning, so there is no need to place RFID tags densely.

In order to facilitate the installation of RFID tags on the floor, this study proposes a strip of tags arranged in a line at regular intervals.

As shown in Figure 8, the tag strip is mounted with RFID tags at regular intervals. The strip is made of a thin, slightly width vinyl film that is rolled into a cylinder for easy storage and handling. In addition, it can be cut to the required length when installed on the floor.

When the interval of each tag in the tag strip is constant, the position of the remaining tags can be calculated by knowing the position of the first tag and the position of the last tag. This method makes it easy to calculate the position of the rest of the tags by starting and ending tag positions without entering the coordinates of every tag.

When the coordinates of the start S and end E of the tag are known and the number n of tags installed is calculated, the positions of the remaining tags are calculated as follows.

은 다음과 같이 표현할 수 있다. (

일때)
Equation of a straight line through the starting point S and the end point E

Can be expressed as (

when)

&Quot; (18) "

이며

일 때, 각 태그의 위치

는 수학식 19와 같이 계산가능하다.

And

When, position of each tag

Can be calculated as in Equation 19.

&Quot; (19) "

The tag installed on the band makes the construction easier on the floor, and does not affect the position recognition of the robot itself. The position recognition of the robot is possible from the position of each independent tag installed on the floor.

Data for representing one band is stored in a file in a format as shown in Equation (20).

&Quot; (20) "

는 시작 태그의 위치,

는 끝 태그의 위치,

은 태그의 개수,

는 태그의 개수만큼 존재하는 태그를 구별하는 유일한 식별값이다.

Is the location of the start tag,

Is the position of the end tag,

Is the number of tags,

Is a unique identification value that distinguishes tags that exist as many as the number of tags.

If more than one band is installed on the floor, the files that store them are stored in rows with data corresponding to each band. The location aware device reads this file to calculate the location of each tag and generates a tag map. For example, as illustrated in FIG. 9, a tag map is generated based on installation of a tag time on a floor.

The band can be installed along the path that the robot moves to increase the recognition rate of the tag.

The application program that implements the RFID position recognition algorithm is programmed to be used or simulated by connecting to a real robot. For simulation, a virtual motor that drives left and right wheels is implemented and the robot's odometry can be calculated.

FIG. 10 is a view illustrating an environment in which an RFID tag is installed. A total of 121 tags having a width of 5m x 5m and a recognition radius of 0.09m every 0.5m are installed at equal intervals. In FIG. 10, the horizontal and vertical solid lines are lines drawn at intervals of 1 m. The round gray circle indicates the position and the recognition radius of the RFID tag. This circle turns red when the tag is recognized. The gray robots show the actual robot's position, and the brown robots show the position estimated by RFID positioning algorithm. Blue dots mark the location of each particle.

In the simulation, the position and orientation of the actual robot was used by the dead-reckoning position and orientation.

The process of estimating the position of the robot is illustrated in FIGS. 11 to 14. The robot's position is unknown until the robot detects the first tag. When the robot detects the first tag, particles are randomly generated within the recognition radius of the tag as shown in FIG. 11. As shown in FIG. 12, the particles spread in a circle because the direction is unknown until the robot detects another tag. As shown in FIG. 13, when the robot detects the second tag, particles outside the recognition radius of the tag are destroyed so that only particles having the same direction remain. Referring to FIG. 14, when the robot moves, the particles spread out according to the probability distribution of the robots until the next tag is encountered, and when the tags are met, the particles repeat the convergence process again. At this time, the position estimation error of the robot increases in proportion to the area where the particles are spread.

During the simulation, the robot was moved manually by using a joystick. During the simulation, the robot made a proper position estimate and never lost the position of the robot.

The RFID position recognition device was applied to the robot according to an embodiment of the present invention and tested. In order to calculate the error between the estimated position and the actual position of the RFID positioning device, the laser tracker, a 3D positioning device, was used to measure the actual position of the robot.

15 is a photograph of an RFID reader and an antenna mounted on the bottom of the base frame of the robot. The height of the RFID antenna and the floor surface is 3cm.

In the experimental environment, as shown in FIG. 16, RFID tags are installed on the floor at intervals of 12 cm in a 2.4 m x 2.4 m space and are covered with wooden tiles.

FIG. 17 is a position system (StarGazer) using a landmark and a camera in the center for comparison experiments with existing commercial products, and a receiver of an ultrasonic satellite based position system (U-SAT), which is located at both sides. The small gray balls attached to the center and both edges are three-dimensional position recognition landmarks for precise positioning of the robot.

18 is a transmitter of an ultrasonic satellite based positioning system mounted on the ceiling. It is mounted on four corners of the space to measure the position and generates ultrasonic waves sequentially and transmits them to the receiver. 19 is a landmark of a StarGazer mounted on the ceiling. The landmark of StarGazer consists of a marker that can recognize the location and direction and a marker that can distinguish the ID of the landmark.

RFID tags were arranged in a grid of 20 horizontally and vertically spaced 12 cm apart on a 2.4m x 2.4m floor. The RFID reader and antenna are mounted at the center of the drive shaft at the bottom of the 2-Wheel Differential Wheel type mobile robot. The height of the RFID antenna and the ground is 3 cm.

In order to prove the performance of the developed RFID positioning device, we applied the mobile robot to experiment and compared the actual location of the robot with the estimated location based on RFID, and image-based location recognition system using landmarks and cameras. StarGazer) and ultrasonic satellite based position recognition system (U-SAT).

The robot was manipulated to move in a circle using a joystick. The maximum number of particles for performing the RFID position recognition algorithm was set to 1000 and the execution period was set to 130ms. At the same time as RFID position recognition, StarGazer position recognition and U-SAT position recognition were performed and compared with actual robot position measured in 3D position recognition system.

20 and 21 show graphs of distance and direction errors with respect to the estimated position and the actual position of the RFID position recognition device.

As shown in FIGS. 20 and 21, the distance error between the estimated position and the actual position of the RFID position recognition device is less than 0.04 m, and the direction error is less than 3 °.

To verify the performance of the RFID positioning device, we compared the performance with the existing commercial positioning sensors. 22 and 23 illustrate positional errors of an RFID device, a U-SAT using an ultrasound satellite, and a Stargazer using a landmark and a camera.

As shown in FIG. 22, the distance error from the actual position may be smaller than that of the conventional commercial position recognition device.

As shown in FIG. 23, the error with the actual direction may be smaller than that of the conventional commercial position recognition device.

Table 1 shows the average distance and the direction error of the RFID device, the U-SAT, the landmark and the camera, the Stargazer.

Average distance error (m) Average direction error (°) RFID 0.018 1.233 Stargazer 0.053 4.867 U-SAT 0.235 23.741

Therefore, the RFID positioning device is less than the image based positioning device (StarGazer) using landmark and camera, and much less than the ultrasonic satellite positioning device (U-SAT). have.

The average distance error between the estimated position of the robot and the actual position (measured by the laser tracker) in the RFID positioning device was 0.018m, and the average direction error was 1.233 °. Compared with the existing commercial products StarGazer and U-SAT, the error of RFID positioning device is the smallest, which proves the superiority of RFID positioning device.

RFID positioning devices, however, have a high initial installation cost. Since the RFID tag must be installed at the bottom of the space the robot wants to move, the cost depends on how tightly the tag is installed.

RFID positioning device has higher initial installation cost than existing commercial products, but it does not produce errors due to robot tilt due to ground curvature, errors due to robot's moving speed, and obstructions caused by obstacles. Positioning performance is guaranteed.

It will be apparent to those skilled in the art that various modifications and variations can be made in the present invention without departing from the spirit or scope of the invention as defined in the following claims And changes may be made without departing from the spirit and scope of the invention.

Claims (1)

Autonomous running position recognition device of a mobile robot according to the detailed description and the drawings.

Images