Map stitching method in changing indoor and outdoor environments

Abstract

The present invention discloses a map splicing method for indoor and outdoor variable environments, comprising a house, a rectangular yard connected to the house door, an electromagnetic sensor, a laser radar, and a GPS positioning system provided on a robot; a processor of the robot is respectively connected to the electromagnetic sensor, the laser radar, and the GPS positioning system; the present invention has the characteristics of realizing seamless indoor and outdoor navigation by splicing indoor and outdoor raster maps as a priori static maps for the mobile robot in indoor and outdoor environments; and by adjusting C, the success rate of map splicing based on the characteristics of the raster image can be effectively optimized.

Description

Technical Field

[0001] The present invention relates to the technical field of intelligent robots, and in particular to a map splicing method in indoor and outdoor variable environments for a robot to work indoors and outdoors. Background Art

[0002] As the application demand for robots becomes more and more extensive, the actual environment they face is becoming more and more complex. When intelligent robots work in changing indoor and outdoor environments, they will face many problems such as inaccurate positioning and inaccurate navigation of indoor and outdoor seamless navigation technology.

[0003] For example, compared with indoor sweeping robots, the working environment of lawn mowing robots is an open outdoor environment without boundary fences. The indoor navigation and obstacle avoidance system used by indoor sweeping robots cannot be directly applied to lawn mowing robots.

[0004] Therefore, when faced with practical indoor and outdoor environmental problems, multiple independent robots are usually selected to complete indoor and outdoor tasks respectively.

[0005] Although multiple independent robots working together can complete tasks to a certain extent, the following problems still exist:

[0006] First, using multiple robots to divide the work and complete the task incurs a high cost.

[0007] Second, multiple robots are difficult to operate, which increases the difficulty for users. Summary of the Invention

[0008] The purpose of the invention is to overcome the shortcomings of the existing technology of using multiple independent robots to divide the work to complete indoor and outdoor work, which is costly and difficult to use. A map splicing method in indoor and outdoor variable environments is provided to facilitate a single robot to work indoors and outdoors.

[0009] In order to achieve the above object, the present invention adopts the following technical solutions:

[0010] A map stitching method for indoor and outdoor variable environments includes a house, a rectangular yard connected to the house door, an electromagnetic sensor, a laser radar, and a GPS positioning system installed on a robot; a processor of the robot is respectively connected to the electromagnetic sensor, the laser radar, and the GPS positioning system; and the method includes the following steps:

[0011] Step 1: Create a map of the indoor environment map1:

[0012] The processor is installed with a Linux system. The processor obtains the indoor environmental data detected by the LiDAR. The Linux system's ROS platform calls the SLAM mapping algorithm - the Cartagrapher algorithm, to locate and create the indoor two-dimensional grid map map1 in real time.

[0013] Step 2: Create a map of the outdoor environment (map 2): Starting from the left and right sides of the door, rectangular rings of magnetic strips are laid along the edge of the yard. The processor controls the robot to move along the magnetic strips for a circle based on the magnetic field information detected by the electromagnetic sensor and the robot's position information detected by the GPS positioning system. The robot identifies the working boundary created by the magnetic strips and uses the GPS positioning system to locate and track the magnetic strip boundaries in real time. The ROS platform uses the laser SLAM mapping algorithm (Cartagrapher) to create a two-dimensional grid map of the outdoor area with fixed boundaries (map 2).

[0014] The present invention utilizes magnetic strips to set the working boundary of an outdoor environment, which can effectively prevent the robot from running out of the working range and improve the safety of the robot.

[0015] In step 3, the processor uses the raster image feature matching method to stitch map1 and map2. After the maps are stitched together, the indoor and outdoor global static stitched map map_merge is published.

[0016] Preferably, step 3 includes the following specific steps:

[0017] The processor calls the map information of map1: node1_robot1 and saves it to the map node of robot_1; the map information of map2 is called: node2_robot2 and saved to the map node of robot_2; the ROS platform calls the map_merge map splicing node to automatically add map1 and map2. The ROS platform uses the Open-CV feature matching method to identify and match the similarity of the grids at the overlap of map1 and map2, perform position estimation and transformation, and realize map splicing.

[0018] Preferably, the open-cv feature matching method includes the following specific steps:

[0019] The ROS platform calls map1 a two-dimensional grid map G, and map2 a two-dimensional grid map W;

[0020] G is represented by a set of pixels:

[0021] W is represented by a set of pixels:

[0022] Among them, g i1 and wi2 are all vectors, N g and N w Represent the number of pixels in G and W respectively;

[0023] Use G ε Represents the edge pixel set of the overlapping area of ​​G and W, let G ε is a subset of G, and ε represents the overlap percentage;

[0024] The ROS platform uses the following formula to calculate the point pair relationship f at time k k (i)

[0025]

[0026] Where i = 1, 2, ... N g ;

[0027] Assume that the rigid body transformation T = {R, t} is calculated as follows: where R is the rotation matrix and t is the translation vector;

[0028] R k-1 is the rotation matrix at time k-1, t k-1 is the translation vector at time k-1;

[0029] According to g of G at time k-1 i1 and W's w i2 , use the rigid body transformation matrix T to calculate the map feature point pair relationship f at time k k (i);

[0030] The ROS platform calculates the overlap percentage ε k and update subset

[0031]

[0032] ε∈[ε min ,1], |G ε |=ε|G|;

[0033] Where: p i represents the i-th group of pixels that are successfully matched between G and W, |·| represents the number of elements in the calculation set, and ε min Indicates the set minimum overlap percentage, C is the set dynamic confidence parameter, P o is the set of pixels from the first group to the Nth group where G and W are successfully matched, where N is the number of pixels where G and W are successfully matched;

[0034] The ROS platform utilizes k (i) Calculate (i, f k(i)) The overlap percentage ε in the state k and matching subsets

[0035] After the rigid body transformation, the point set T(map1) can be well matched with map2, and the grid stitching problem is converted into a minimization problem.

[0036] As a preference, since the edge feature points of the raster image are blurred and difficult to stitch, the ROS platform sets the initial value of C to 1; when the raster image is complex and the features of the indoor and outdoor overlaps are not obvious, and the stitching is unsuccessful, C is gradually reduced with a step size of -0.1 until the image stitching is successful. The ROS platform publishes the indoor and outdoor global static stitching map map_merge.

[0037] The present invention enables the intelligent robot to recognize the global working environment indoors and outdoors and realize seamless navigation indoors and outdoors, which can effectively improve its intelligence level, efficiently solve practical problems and reduce the difficulty of user operation.

[0038] Therefore, the present invention has the following beneficial effects:

[0039] In indoor and outdoor environments, the spliced ​​indoor and outdoor raster maps are used as the prior static map for the mobile robot to achieve seamless indoor and outdoor navigation; by adjusting C, the success rate of map splicing based on the characteristics of raster images can be effectively optimized, making it easier for a robot to work indoors and outdoors. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] Figure 1 is a flow chart of the present invention;

[0041] Figure 2 It is a schematic diagram of map1 of the present invention;

[0042] Figure 3 is a schematic diagram of map2 of the present invention;

[0043] Figure 4 This is a schematic diagram of map_merge of the present invention;

[0044] Figure 5 This is a schematic diagram of an indoor and outdoor map stitching process according to an embodiment of the present invention;

[0045] Figure 6 This is a navigation test chart for indoor and outdoor environments according to an embodiment of the present invention. DETAILED DESCRIPTION

[0046] The present invention will be further described below with reference to the accompanying drawings and specific embodiments.

[0047] like Figure 1-Figure 4The illustrated embodiment is a map stitching method for a variable indoor and outdoor environment, comprising a house 1, a rectangular courtyard 2 connected to a door 11 of the house, an electromagnetic sensor, a laser radar, and a GPS positioning system installed on a robot; the robot's processor is respectively connected to the electromagnetic sensor, the laser radar, and the GPS positioning system; and the method comprises the following steps:

[0048] Step 1: Create a map of the indoor environment map1:

[0049] The processor is installed with Linux system, which obtains the environmental data of the room detected by LiDAR. The ROS platform of Linux system calls SLAM mapping algorithm - Cartagrapher algorithm, and establishes real-time positioning. Figure 2 The indoor two-dimensional grid map map1 shown;

[0050] Step 2, build a map of the outdoor environment map2: starting from the left and right sides of the door, lay a rectangular ring of magnetic strips along the edge of the yard 3. The processor controls the robot to move along the magnetic strips according to the magnetic field information of the magnetic strips detected by the electromagnetic sensor and the robot position information detected by the GPS positioning system, so that the robot can identify the working boundary constructed by the magnetic strips, and locate and track the magnetic strip boundaries in real time through the GPS positioning system. The ROS platform calls the laser SLAM mapping algorithm - Cartagrapher algorithm to build an outdoor fixed boundary such as Figure 3 The two-dimensional grid map map2 shown;

[0051] Step 3: The processor uses the raster image feature matching method to splice map1 and map2. After the maps are spliced, they are published as follows: Figure 4 The indoor and outdoor global static mosaic map map_merge is shown:

[0052] The processor calls the map information of map1: node1_robot1 and saves it to the map node of robot_1; the map information of map2 is called: node2_robot2 and saved to the map node of robot_2; the ROS platform calls the map_merge map splicing node to automatically add map1 and map2. The ROS platform uses the Open-CV feature matching method to identify and match the similarity of the grids at the overlap of map1 and map2, perform position estimation and transformation, and realize map splicing.

[0053] Among them, the open-cv feature matching method includes the following specific steps:

[0054] When the robot builds a grid map, the environment is divided into two-dimensional grids of the same resolution, and the grids in the environment correspond to pixels in the map. The ROS platform calls map1 a two-dimensional grid map G, and map2 a two-dimensional grid map W.

[0055] G is represented by a set of pixels:

[0056] W is represented by a set of pixels:

[0057] Among them, g i1 and w i2 are all vectors, N g and N w Represent the number of pixels in G and W respectively;

[0058] Use G ε Represents the edge pixel set of the overlapping area of ​​G and W, let G ε is a subset of G, and ε represents the overlap percentage;

[0059] Calculate the rigid body transformation T = {R, t};

[0060] Where R is the rotation matrix and t is the translation vector, so that the transformed point set T(map1) can match map2 well, and the grid splicing problem is converted into a minimization problem;

[0061] The ROS platform uses the following formula to calculate the point pair relationship f at time k k (i)

[0062]

[0063] Where i = 1, 2, ... N g ;

[0064] Assume that the rigid body transformation T = {R, t}, R k-1 is the rotation matrix at time k-1, t k-1 is the translation vector at time k-1;

[0065] According to g of G at time k-1 i1 and W's w i2 , use the rigid body transformation matrix T to calculate the map feature point pair relationship f at time k k (i);

[0066] The ROS platform calculates the overlap percentage ε k and update subset

[0067]

[0068] ε∈[εmin ,1], |G ε |=ε|G|;

[0069] Where: p i represents the i-th group of pixels that are successfully matched between G and W, |·| represents the number of elements in the calculation set, and ε min Indicates the set minimum overlap percentage, C is the set dynamic confidence parameter, P o is the set of pixels from the first group to the Nth group where G and W are successfully matched, where N is the number of pixels where G and W are successfully matched;

[0070] The ROS platform utilizes k (i) Calculate (i, f k (i)) The overlap percentage ε in the state k and matching subsets

[0071] Among them, because the edge feature points of the raster image are fuzzy and difficult to stitch, the ROS platform sets the initial value of C to 1; when the raster image is complex and the features of the indoor and outdoor overlaps are not obvious, and the stitching is unsuccessful, C is gradually reduced with a step size of -0.1 until the image stitching is successful. The ROS platform releases the global static stitching map map_merge of indoor and outdoor.

[0072] Figure 5 (a) in the figure is the unjoined raster maps map1 and map2. Figure 5 (b) in the figure is the map_merger.

[0073] From the stitching result in (b), we can see that the stitched map not only matches the overlapping areas of the two raster maps but also preserves all the pixel data of the two raster maps map1 and map2.

[0074] like Figure 6 As shown in the figure, from the original coordinate position of the starting point to the door in the indoor grid map map1, we can know the position relationship between the starting point and the door is M1[(x1,y1,z1,yaw1)-(x2,y2,z2,yaw2)]. At the same time, according to the original outdoor grid map map2, we can know the position relationship between the door and the target point is M2[(x2,y2,z2,yaw2)-(x3,y3,z3,yaw3)].

[0075] The specific test method is based on the above conditions. The starting point, key path point 1 and target point are set as the navigation detection path on the spliced ​​map map_merge, so that the robot starts to navigate autonomously from the starting point to the target point, and uses ros to publish and monitor the robot's position coordinates (x, y, z, yaw) in real time. It can be seen that the actual position relationship between the starting point and the target point is N[(x1, y1, z1, yaw1)-(x3, y3, z3, yaw3)]. By analyzing the robot's trajectory, it can be judged that the actual driving path deviates very little from the ideal path.

[0076] By splicing indoor and outdoor local grid maps, a priori static maps for seamless indoor and outdoor positioning and navigation are matched with the robot's real-time sensor signals for positioning and navigation.

[0077] The actual experimental test results show that when C is around 0.7, the indoor and outdoor spliced ​​map can be used as a priori static map to achieve seamless indoor and outdoor navigation with good results.

[0078] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A map splicing method in indoor and outdoor changing environments, characterized by: The invention comprises a house (1), a rectangular yard (2) connected to a door (11) of the house, an electromagnetic sensor, a laser radar and a GPS positioning system installed on a robot; a processor of the robot is respectively connected to the electromagnetic sensor, the laser radar and the GPS positioning system; and the invention comprises the following steps: Step 1: Create a map of the indoor environment map1: The processor is installed with a Linux system. The processor obtains the indoor environmental data detected by the LiDAR. The Linux system's ROS platform calls the SLAM mapping algorithm - the Cartographer algorithm, to locate and create the indoor two-dimensional grid map map1 in real time. Step 2, establish a map map2 of the outdoor environment: starting from the left and right sides of the door, a rectangular ring of magnetic strips (3) is laid along the edge of the yard. The processor controls the robot to move along the magnetic strips according to the magnetic field information of the magnetic strips detected by the electromagnetic sensor and the robot position information detected by the GPS positioning system, so that the robot can identify the working boundary constructed by the magnetic strips, and locate and track the magnetic strip boundaries in real time through the GPS positioning system. The ROS platform calls the laser SLAM mapping algorithm - Cartographer algorithm to establish a two-dimensional grid map map2 with a fixed boundary outdoors; In step 3, the processor uses a raster image feature matching method to stitch map1 and map2 together. After the maps are stitched together, the processor publishes a global static stitched map of indoor and outdoor areas, map_merge. Step 3 includes the following specific steps: The processor calls the map information of map1: node1_robot1 and saves it to the map node of robot_1; the map information of map2: node2_robot2 and saves it to the map node of robot_2; the ROS platform calls the map_merge map splicing node to automatically add map1 and map2. The ROS platform uses the Open-CV feature matching method to identify and match the similarity of the grids at the overlap of map1 and map2, perform position estimation and transformation, and realize map stitching; Open-CV feature matching has a dynamic confidence parameter C. Because the edge feature points of raster images are fuzzy and difficult to stitch, the ROS platform sets C to an initial value of 1; when the raster image is complex and the features at the overlap of indoor and outdoor areas are not obvious, and the stitching is unsuccessful, C is gradually reduced by a step size of -0.1 until the image stitching is successful. The ROS platform then publishes the global static stitching map map_merge of indoor and outdoor areas.

2. The map splicing method under indoor and outdoor changing environments according to claim 1 is characterized in that: The Open-CV feature matching method includes the following specific steps: The ROS platform calls map1 a two-dimensional grid map G, and map2 a two-dimensional grid map W; G is represented by a set of pixels: W is represented by a set of pixels: Among them, g i1 and w i2 are all vectors, N g and N w Represent the number of pixels in G and W respectively; Use G ε Represents the edge pixel set of the overlapping area of ​​G and W, let G ε is a subset of G, and ε represents the overlap percentage; The ROS platform uses the following formula to calculate the point pair relationship f at time k k (i) Where i = 1, 2, ... N g ; Assume that the rigid body transformation T = {R, t}, R k-1 is the rotation matrix at time k-1, t k-1 is the translation vector at time k-1; According to g of G at time k-1 i1 and W's w i2 , use the rigid body transformation matrix T to calculate the map feature point pair relationship f at time k k (i); The ROS platform calculates the overlap percentage ε k and update subset Among them, p i represents the i-th group of pixels that are successfully matched between G and W, |·| represents the number of elements in the calculation set, and ε min Indicates the set minimum overlap percentage, C is the set dynamic confidence parameter, P o is the set of pixels from the first group to the Nth group where G and W are successfully matched, where N is the number of pixels where G and W are successfully matched; The ROS platform utilizes k (i) Calculate (i, f k (i)) The overlap percentage ε in the state k and matching subsets

Images