RU2740229C1 - Method of localizing and constructing navigation maps of mobile service robot - Google Patents

Abstract

FIELD: physics.

SUBSTANCE: invention relates to robotics. In the method of localizing and constructing navigation maps of a mobile service robot operated in a room, correcting the current coordinates of the robot and the trajectory traversed using stationary light-reflecting beacons with a RFID tag mounted on each of the beacons. Constructing the sub-map of obstacles in a specific position of the robot and updating layers of the constructed sub-map using the RGB-D camera. Obtaining the sub-map of obstacles to the coordinates system of the room according to the trajectory traversed. When the robot moves to the specified threshold value, a new sub-map of obstacles is constructed and the layers of the sub-map are updated. Combining the sub-map of obstacles in the obstacle map after construction of each new sub-map of obstacles. Global map is then constructed using the SLAM map data, the static map and taking into account the obstacle map data.

EFFECT: higher accuracy of robot coordinates determination.

1 cl, 8 dwg

Description

The invention relates to the field of robotics, namely to systems and methods for localizing and constructing obstacle maps and a global map of the premises for a mobile service robot performing technological functions at commercial facilities, in particular, operations related to inventory.

There is a known method of operation of a robot for cleaning floors (application WO No. 2019007038A1 from 23.01.2018), which includes the following steps: obtaining from a binocular camera two images of an object taken simultaneously from different angles; obtaining information about the obstacle based on the information of the two images and the selection of different modes of operation based on the information about the obstacles. The obstacle information contains the size of the obstacle, the shape of the obstacle, and the distance between the obstacle and the robot. The step of simultaneously acquiring two images of the same obstacle from different angles further comprises associating by the video image processing system all information about the obstacle with the corresponding operating mode.

The disadvantage of this method is the low accuracy of determining the contours of obstacles, due to the use of only an RGB camera for positioning.

There is a known method of compiling maps of the working area for an autonomous robot (application for invention US No. 2018 / 0004217A1 dated 11/18/2015), including: the robot performs identification of boundary lines between adjacent displayed sub-areas and non-displayed sub-areas of the work area, which should be displayed on the map, by comparing the distances traveled by a robot traveling for the first time across the working area, measured using coordinates and odometry; initiating display on the map of an unmapped sub-area adjacent to the boundary lines identified by the contact sensor while moving from a first point on the first boundary line to an unmapped sub-area; Generation of a map of the work area based on maps of sub-areas compiled by the robot.

The main disadvantage of this method is the lack of a method for correcting the calculated coordinates of the robot, which leads to an increase in the error in determining the coordinates of the location of the robot during movement, since the odometry data have a significant error.

Known is a method of constructing the robot map (Patent US № 8761925V2 on 23.10.2008) equipped with a three-dimensional measuring apparatus (e.g., a stereo camera or a video camera and the depth distance (ToF-camera)), comprising: sequentially receiving first and second data surfaces route by which moves robot; comparing the data of the first and second surfaces to calculate the difference between the data of the first and second surfaces; detecting a dynamic obstacle from the first and second surface data in accordance with the difference between the data on the first and second surfaces; generating third surface data by removing dynamic obstacle data from at least the first or second surface data; mapping third surface data without dynamic obstruction to first and second surface data to build a map, where the data is matched by iterating the nearest point ( ICP ) to capture the first and second surface data, and the robot acquires 3D image data as it moves.

The main disadvantage of this method is the lack of a method for correcting the calculated coordinate of the robot, which leads to an increase in the localization error during movement, since the odometry data have a significant error. Also, the disadvantage of this method is the use of only one type of sensor - a three-dimensional measuring device.

A known system and method of vehicle navigation using absolute and relative coordinates (application JP No. 2010519550A from 21.02.2008). The system includes a GPS absolute position sensor in addition to one or more additional sensors such as a camera, laser scanner or radar.

As you move, sensors detect the presence of objects and measure the relative position of the vehicle relative to these objects. This information is used in conjunction with absolute position information and additional map information to locate the vehicle and to support steering and collision avoidance functions. The system uses relative positioning according to one of the variants, without relying on storing information about the absolute position. The navigation method uses RFID tags to identify various physical objects on the map, such as road signs.

The described system and method uses GPS to correct the current coordinate, which does not allow their use for robots in closed rooms.

There is a known method for generating and updating room maps (application CN No. 107764270A dated 19.10.2017) by a robot with a laser radar (hereinafter - LIDAR) installed on its upper part, an encoder and a central processor chip installed in the robot; a signal antenna connecting the robot with the control computer, and LIDAR for scanning the surrounding space and transmitting the scanned information about discrete points to the central processor. Using LIDAR allows fast and accurate data collection and accurate creation and updating of a room map.

This method includes the following steps: converting data on 361 points of the robot's environment, obtained using LIDAR, in the range of 180 degrees with an interval of 0.5 degrees, represented by polar coordinates into the global coordinates of the room map in the Cartesian coordinate system; processing information about the data of discrete points, received by LIDAR, into a line segment used to represent information about the robot's environment (obstacles) and generate a local map; moving the robot in the room to get a lot of local maps; generating a global room map by combining multiple local maps; updating the global room map for precise positioning of the robot by calculating the relationship between the corresponding line segments for the corresponding line segments of the local and global maps; represent areas of maps with one or more line segments.

The main disadvantage of this method is the impossibility of detecting obstacles that are above and below the scanning level of the LIDAR.

The known method and device for the navigation of the robot (application CN No. 106681330A from 25.01.2017, prototype), using multisensor data fusion ( sensor fusion technology) to build a map.

The method comprises the following steps: creating a map of the surrounding space in accordance with the data obtained by the laser radar and odometry data; determination of the current position of the robot on the global map of the surrounding space in real time in accordance with the data obtained using laser radar, accelerometer, gyroscope and magnetometer, global map and odometry data; a map of the local environment is built in accordance with the data obtained in real time by a depth camera and a laser radar sensor; control of the robot to avoid obstacles is carried out using a local path planning algorithm, a strategy for choosing a preset direction; a map of the local environment (including obstacles), the current position and the planned route of the robot; real-time data is projected onto a 2D plane to create a 2D projection; if no two-dimensional projection data obtained by the laser radar exists for the corresponding position of the robot, the local environment map data of the corresponding position is filled with the corresponding two-dimensional projection data; If 2D projection data and laser radar data are available for the corresponding robot position, then the 2D projection data and laser radar data are weighted and averaged to obtain the local map data of the surrounding space for the corresponding robot position.

The main disadvantage of this method is the lack of a method for correcting the calculated coordinate of the robot, which leads to an increase in the localization error during movement, since the odometry data have a significant error.

The technical problem solved by the invention is to improve the accuracy of determining the coordinates of the location of the robot, the shape and location of obstacles in the path of the mobile service robot in real time.

The technical problem is solved in the method of localization and construction of navigation maps of a mobile service robot operating in the premises of a commercial facility, including: the use of stationary reflective beacons indoors; building an initial route using a floor plan (static map); determination of the current coordinate of the robot using a module for simultaneous localization and mapping (hereinafter - SLAM, English: simultaneous localization and mapping) ; the use of RFID tags to identify stationary reflective beacons in the space surrounding the robot; detection of object contours using a depth camera, LIDAR and ultrasonic rangefinders; SLAM SLAM construction module card using data LIDAR sensors odometry and inertial measurement unit (hereinafter - IMU); determination of the current coordinate of the robot using the SLAM module; construction of the trajectory traversed by the robot by the SLAM module; building a global map using SLAM and static map data; converting data from a depth camera into a projection onto a plane, the method including: adjusting the current coordinate of the robot using stationary reflective beacons with an RFID tag installed on each of the beacons; building a submap of obstacles in a specific position of the robot and updating the layers of the constructed submap of obstacles using data from the RGB-D camera, rear LIDAR and ultrasonic rangefinders, respectively, when building each of the layers in a random sequence; binding of the current submap of obstacles to the coordinate system of the room in accordance with the trajectory of the robot; when the robot moves by a predetermined threshold value, building a new submap of obstacles for the robot and updating the layers of the constructed submap of obstacles using data from the RGB-D camera, rear LIDAR and ultrasonic rangefinders when building each of the layers, respectively, in a random sequence; binding a new submap of obstacles to the coordinate system of the room in accordance with the trajectory of the robot; merging obstacle submaps into an obstacle map after building each new obstacle submap; building a global map using SLAM map data, static map and obstacle map.

The solution of the technical problem - increasing the accuracy of determining the current coordinate of the location of the robot, the shape and location of obstacles in the path of the mobile service robot in real time - is provided by the following set of distinctive features of the method for localizing and building a map of the mobile service robot, including: correcting the current coordinate of the robot using stationary reflective beacons with an RFID tag installed on each of the beacons; building a submap of obstacles in a specific position of the robot and updating the layers of the constructed submap of obstacles using data from the RGB-D camera, rear LIDAR and ultrasonic rangefinders, respectively, when building each of the layers in a random sequence; binding of the current obstacle submap to the coordinate system of the room in accordance with the trajectory of the robot; when the robot moves by a predetermined threshold value - building a new submap of the robot obstacles and updating the layers of the constructed submap of obstacles using data from the RGB-D camera, rear LIDAR and ultrasonic rangefinders, respectively, when building each of the layers in a random sequence; binding a new submap of obstacles to the coordinate system of the room in accordance with the trajectory of the robot; merging obstacle submaps into an obstacle map after building each new obstacle submap; building a global map using SLAM map data, static map and taking into account obstacle map data.

This set of distinctive features was not found during the patent information search, therefore, the technical solution meets the criterion of "novelty". It also does not follow explicitly from the prior art, therefore, it meets the criterion of "inventive step".

FIG. 1 shows a diagram of the premises of a commercial facility.

FIG. 2 shows a block diagram of a mobile service robot.

FIG. 3 shows a block diagram of the computer software of a mobile service robot.

FIG. 4 shows a block diagram of the complex of the software module for generating a global map of the room.

FIG. 5 shows a functional diagram of the formation of a global map of the premises of a commercial facility.

FIG. 6 shows a diagram of the algorithm for correcting the estimate of the current coordinate of the mobile service robot using stationary reflective beacons.

FIG. 7 shows a diagram of the algorithm for detecting stationary reflective beacons.

FIG. 8 shows a diagram of the obstacle map construction algorithm.

A method for localizing and constructing navigation maps of a mobile service robot 1 (Figs. 1-2) equipped with sensors: RFID 2, front LIDAR 3, rear LIDAR 4, RGB-D video camera 5 and ultrasonic rangefinders 6; a controller 7 for processing signals from sensors 2-6, wheels with engines 8 with an odometer sensor 9 and IMU 10, an engine controller 11 that processes data from the odometer sensor 9 and IMU 10; the computer 12 of the mobile service robot 1, which is by means of the unit 13 of the human-machine interface or by means of an external service device 14 connected to the computer 12 through the unit 15 of communications. In the room 17 of the commercial facility, divided into zones 18, with racks 19, there are stationary reflective beacons 16 with an RFID tag installed on each of the beacons (not shown).

Robot 1 with computer 12 has the following program modules (Fig. 3):

- a software module 20 for generating a static map,

- a complex of 21 software modules for forming a global map of the premises of a commercial facility,

- software module 22 for processing sensor signals,

- software module 23 navigation and recognition of racks,

- the program module 24 for the formation of the route of movement,

- a program module 25 for motion control.

The complex of 21 software modules for forming a global map of the premises of a commercial facility consists of interconnected modules:

- program module 21 a for correcting the coordinates of the robot's location,

- software module 21 b SLAM , which implements SLAM algorithms,

- software module 21 c creating a projection of data from an RGB-D camera onto a horizontal plane,

- software module 21 d for building a map of obstacles,

- the program module 21 e forming a global map.

The method for localizing and constructing navigation maps of a mobile service robot operating in a commercial facility is performed in the following sequence.

When the command "start" is received, the program module (hereinafter referred to as the PM) 21 receives a static card from the PM 20. Having received the initial data, the PM 21 generates a starting global map of the room 17 with the marked starting position of the robot 1, including the coordinate of the starting position. The generated global map and coordinates are transmitted to the PM 23, which generates waypoints from the received data. The generated waypoints are transmitted to the PM 24, which, on the basis of the coordinates of the received waypoints, generates a movement route transmitted for execution to the PM 25, which, taking into account the surrounding dynamic situation, received from sensors 2-6 through controller 7 (Fig. 2) and PM 22 generates control commands for motors 8 (Fig. 2) and transmits control commands to controller 11 (Fig. 2). The controller 11 gives the command to start the movement of the robot 1. The robot 1 starts moving from the starting position along the route formed by the PM 24.

Next, the parallel operation of the PM 21-25 is started (Fig. 3). PM 22 receives data from sensors 2-6 (Fig. 1, 2), scanning the surrounding space, and after primary data processing by controller 7 (Fig. 2) transmits summary information to PM 21 and PM 25 (Fig. 3). PM 21 (Fig. 3) forms, as the survey progresses, a global map of the room 17 and determines the coordinate of the location of the robot 1.

PM 21 (Fig. 3) works as follows.

PM 21 b SLAM (Fig. 4, 5) receives from PM 22 (Fig. 3) odometry and IMU data. PM 21 b processes the received data using the SLAM method

(http://ais.informatik.unifreiburg.de/teaching/ss12/robotics/slides/12-slam.pdf),

evaluates the trajectory traversed and the absolute coordinates of the robot 1 and transmits the result to the PM 21 a (Fig. 5). PM 21 a , using the data from the front LIDAR received from PM 22 (Fig. 3), detects the presence of stationary reflective beacons 16 (Fig. 1) with RFID tags (Fig. 1) in accordance with the algorithm (Fig. 7). If an RFID tag corresponding to a stationary reflective beacon 16 (Fig. 1) is detected in the environment of the robot 1 (Fig. 1), PM 21 a (Fig. 5), in accordance with the algorithm (Fig. 6), adjusts the location coordinate and the trajectory of the robot 1 (Fig. 1), taking into account the coordinates of the detected stationary reflective beacon 16 (Fig. 1).

The corrected result is returned to the PM 21 b (Fig. 5), which forms the current SLAM map and the trajectory traversed by the robot 1 (Fig. 1).

The current SLAM map is transferred to the software PM 21 e (Fig. 5) of the formation of the global map.

The trajectory traversed is transmitted to the PM 21 d (Fig. 5). PM 21 d works as follows (Fig. 8):

- PM 21 d (Fig. 5) receives from the PM 22 (Fig. 3) data from the rear LIDAR 4 (Fig. 1) and ultrasonic range finders 6 (Fig. 1); then the PM 21 d receives from the PM 22 (Fig. 3) through the PM 21 s from the RGB-D video camera 5 (Fig. 1) a projection onto a horizontal plane;

- a submap of obstacles is built at a specific location of the robot and the layers of the constructed submap of obstacles are updated using data from RGB-D video camera 5 (Fig. 1), rear LIDAR 4 (Fig. 1) and ultrasonic range finders 6 (Fig. 1), respectively, when building each from layers in any order;

- the current submap of obstacles is linked to the coordinate system of the room in accordance with the trajectory of the robot;

- when the robot moves to a predetermined threshold value, a new submap of the robot obstacles is built and the layers of the constructed submap of obstacles are updated using data from the RGB-D video camera 5 (Fig. 1), the rear LIDAR 4 (Fig. 1) and ultrasonic rangefinders 6 (Fig. 1 ) respectively, when building each of the layers in an arbitrary sequence;

- the new obstacle submap is linked to the coordinate system of the room in accordance with the trajectory of the robot.

According to the obtained data, the PM 21 d (Fig. 5), combining the sub-maps of obstacles, forms an obstacle map containing information about the contours of obstacles detected along the route of the robot 1 (Fig. 1).

The generated obstacle map is transmitted to the PM 21 e (Fig. 5). PM 21 e combines the received cards into one global map: from PM 21 b - SLAM card; from PM 21 d - obstacle map; from PM 20 (Fig. 3) - a static map.

The work of all PMs ends upon the return of the robot 1 (Fig. 1) to the starting position.

Building submaps and obstacle maps allows you to improve the accuracy of determining the shape and location of obstacles in the path of the robot in real time. Correction of the current coordinate of the robot using stationary reflective beacons improves the accuracy of determining the current coordinate of the location of the robot when performing warehouse operations.

Claims (1)

A method for localizing and constructing navigation maps of a mobile service robot operating in a commercial facility, including: using stationary beacons indoors; using RFID tags to identify stationary objects on the global map; detection of object contours using a depth camera, LIDAR and ultrasonic rangefinders; building a SLAM map using LIDAR and an inertial measuring device; determination of the current coordinate of the robot using the SLAM module; construction of the trajectory traversed by the robot using the SLAM module; building a global map using SLAM and static map data; converting data from a depth camera into a projection onto a plane, characterized in that it includes: adjusting the current coordinate of the robot and the trajectory traversed by the robot using stationary reflective beacons with an RFID tag installed on each of the beacons; building a submap of obstacles in a specific position of the robot and updating the layers of the constructed submap of obstacles using data from the RGB-D camera, rear LIDAR and ultrasonic rangefinders, respectively, when building each of the layers in a random sequence; binding of the current obstacle submap to the coordinate system of the room in accordance with the trajectory of the robot; when the robot moves by a predetermined threshold value - building a new submap of the robot obstacles and updating the layers of the constructed submap of obstacles using data from the RGB-D camera, rear LIDAR and ultrasonic rangefinders, respectively, when building each of the layers in a random sequence; binding a new submap of obstacles to the coordinate system of the room in accordance with the trajectory of the robot; merging obstacle submaps into an obstacle map after building each new obstacle submap; building a global map using SLAM map data, static map and taking into account obstacle map data.

Images