A method and system for generating semantic maps of construction robots

Abstract

The present invention discloses a method and system for generating a semantic map of a construction robot. The method processes a plurality of grid maps obtained at different heights to extract basic indoor building structure feature information; constructs a robot passage cost information map based on the basic indoor building structure feature information and the robot body size; converts the indoor height information obtained by two-dimensional laser as the original information set into a coordinate system and adds it to the navigation map, and synthesizes the information with the passage cost information map to obtain a semantic map of the construction robot. The semantic map provided by the present invention extracts the door and narrow aisle area, and increases the travel cost, so that the path is considered in terms of safety when generating the path, and the inaccessible area is screened out, thereby improving the computing power and making the path take into account both safety and efficiency; the map is also divided into different areas, and when the points are connected under the condition that the working point order is not sensitive, the work is carried out according to the nearest order between the areas, so as to further improve the work efficiency.

Description

Technical Field

[0001] The present invention relates to the field of map navigation technology, and in particular to a method and system for generating a semantic map for a construction robot. Background Art

[0002] Traditional mobile robot maps include three-dimensional maps and two-dimensional maps. Three-dimensional maps are mainly constructed through three-dimensional laser point clouds. The advantage is that they cover all the spatial information required for robot navigation, but the disadvantage is that they require a lot of computing power to construct. The corresponding two-dimensional maps have the advantage of low cost and high efficiency in map construction, but the disadvantage is that they cannot obtain three-dimensional information, which makes it easy for robots to touch fixed obstacles above or below their own laser scanning planes during single-point navigation or other automatic path planning. This greatly increases the manual processing workload of later path planning, and at the same time requires high computing power, which reduces efficiency. Summary of the invention

[0003] Therefore, the technical problem to be solved by the present invention is to overcome the defects of the maps in the prior art in that the maps have high requirements on computing power and poor navigation efficiency when used for robot navigation.

[0004] In order to achieve the above object, the present invention provides the following technical solutions:

[0005] In a first aspect, an embodiment of the present invention provides a method for generating a semantic map of a construction robot, comprising the following steps:

[0006] Get multiple raster maps at different altitudes;

[0007] Process multiple raster maps at different heights to extract basic indoor building structure feature information;

[0008] According to the basic indoor building structure feature information and the construction robot body size information, a construction robot passage cost information map is constructed;

[0009] The indoor height information obtained by two-dimensional laser is used as the original information set, and after coordinate system conversion, it is attached to the navigation map and synthesized with the construction robot's travel cost information map to obtain the construction robot's semantic map.

[0010] In one embodiment, the process of training the model to obtain a trained prediction model includes:

[0011] The obtaining of multiple grid maps at different heights includes:

[0012] Get a grid map with a height higher than the upper edge of the door and lower than the lower edge of the beam as the ground Figure 1 ;

[0013] Get a grid map with a height higher than the bottom edge of the floating window and lower than the top edge of the floating window as the map. Figure 2 ;

[0014] Get a grid map with a height lower than the bottom edge of the floating window as the map Figure 3 ;

[0015] Get a ceiling height information map as the height Figure 1 ;

[0016] Get a floating window height information map as the height Figure 2 .

[0017] In one embodiment, the process of processing a plurality of grid maps at different heights to extract basic indoor building structure feature information includes:

[0018] To the ground Figure 1 ,land Figure 2 ,land Figure 3 Perform noise reduction, edge detection and smoothing of detected contours to obtain the original contour set, including: Figure 1 ,land Figure 2 ,land Figure 3 The corresponding indoor ceiling area contour set, indoor medium height area contour set, and indoor low height area contour set;

[0019] Based on the original contour set or combined with height Figure 1 or combined with height Figure 2 , extract door information, bay window information, window information, narrow passage information, beam information and ceiling information, and associate the relative relationships between different building structures.

[0020] In one embodiment, door information is extracted based on the indoor ceiling area contour set, the indoor medium height area contour set, and the indoor low height area contour set. The extraction process includes:

[0021] Use the indoor medium height area contour set and the indoor low height area contour set to perform door feature detection, and save the identified door area information as the indoor medium height area door area set and the indoor low height area door area set respectively;

[0022] The indoor ceiling area contour set is compared with the indoor medium height area contour set and the indoor low height area contour set respectively, and the suspicious door areas corresponding to the indoor medium height area and the indoor low height area are screened out, and the suspicious door areas are expanded. If the four corner points of the door are in the indoor ceiling area contour set and are distributed in at least two different contours, they are saved in the corresponding door area set, and the door areas identified by comparison are used as the indoor medium height area door area supplementary set and the indoor low height area door area supplementary set respectively;

[0023] Compare the indoor medium height area door area supplement set and the indoor low height area door area supplement set with the indoor medium height area outline set and the indoor low height area outline set respectively, complete the door information not recognized by the indoor medium height area outline set and the indoor low height area outline set, and compare the indoor medium height area door area set and the indoor low height area door area set. If the door area under the medium height area cannot find the corresponding door area in the door area of ​​the low height area, remove the door area under the medium height area from the indoor medium height area door area set and add it to the window outline set;

[0024] Use the identified door information to split the indoor medium-height area contour set and the indoor low-height area contour set, and add the split contours to the original set. If a contour has multiple door areas, split them in the order of the doors.

[0025] In one embodiment, based on the indoor medium height area contour set, the indoor low height area contour set and the height Figure 2 Extract pop-up window information. The extraction process includes:

[0026] Compare the indoor medium-height area contour set and the indoor low-height area contour set, extract the corresponding floating window information, and save all the floating window information into the floating window contour set;

[0027] According to the bay window outline collection and height Figure 2 The height information of the floating window in the image is extracted, and the height information corresponding to each floating window outline is extracted, and the highest point in the outline area is taken as the height information of the floating window sill and saved in the floating window information set;

[0028] The bay window information is used to split the contour set of the medium-height area in the room. If the bay window information is detected under a certain contour, the bay window feature area contained in this contour is deleted, and then a new contour is generated from the remaining part of the contour to replace the original contour.

[0029] In one embodiment, the window information is extracted based on the indoor medium-height area contour set and the indoor low-height area contour set, and the extraction process includes:

[0030] Detect window features based on the indoor medium-height area contour set and the indoor low-height area contour set to obtain a window contour set;

[0031] The window contour set is used to split the indoor medium-height area contour set and the indoor low-height area contour set. If window area information is detected under a certain contour, the window feature area contained in this contour is deleted, and then a new contour is generated from the remaining part of the contour to replace the original contour.

[0032] In one embodiment, the narrow channel information is extracted based on the indoor medium-height area contour set and the indoor low-height area contour set, and the extraction process includes:

[0033] Traverse the indoor medium-height area contour set and the indoor low-height area contour set, save the channels in the contour that belong to the set threshold range separately, add a new contour to the original set, and mark the corresponding newly added contour information set in the original contour set as a narrow channel, and save the remaining part to replace the original contour.

[0034] In one embodiment, based on the indoor medium height area contour set, the indoor low height area contour set and the height Figure 1 Information extraction Ceiling and beam information, the extraction process includes:

[0035] According to the indoor medium height area contour set, indoor low height area contour set and height Figure 1 The ceiling height information is extracted, the height information corresponding to each contour is extracted, the extracted height information is traversed and counted, the connected area with an area greater than a preset area threshold and a height greater than a preset height threshold is taken as the ceiling, the rectangular contour feature with a height less than the preset height threshold is taken as the beam, and the extracted ceiling height set and beam area contour set are associated with their corresponding contour areas and saved respectively.

[0036] In one embodiment, the process of associating the relative relationships between different building structures includes: obtaining the association relationship between doors and areas, and the association process is:

[0037] The indoor medium height area door area set and the indoor low height area door area set are respectively connected to the indoor medium height area contour set and the indoor low height area contour set for area detection, and the detection information is respectively added to the indoor medium height area door area information set and the indoor low height area door area information set.

[0038] In one embodiment, the process of associating the relative relationships between different building structures includes: determining the relationship of the bay windows, and the associating process is:

[0039] The floating window contour set is connected with the indoor medium-height area contour set and the indoor low-height area contour set for detection, and the detected area information is added to the floating window information set.

[0040] In one embodiment, the process of constructing a construction robot passage cost information map based on basic indoor building structure feature information and construction robot body size information includes:

[0041] The ordinary area enclosed by the contours in the original contour set is drawn on the map with preset pixel values, the pixel values ​​of the narrow channel information in the original contour information set are assigned to the first preset multiples of the ordinary area values ​​and drawn on the map, and the pixel values ​​of the door area are drawn on the map with the second preset multiples of the ordinary area values ​​to obtain a construction robot travel cost information map, wherein the first preset multiple is determined according to the robot body size, and the second preset multiple is not less than the first preset multiple.

[0042] In a second aspect, an embodiment of the present invention provides a construction robot semantic map generation system, including: a map data source acquisition module, for acquiring a plurality of raster maps at different heights;

[0043] The building structure feature information acquisition module is used to process multiple grid maps at different heights and extract basic indoor building structure feature information;

[0044] A robot passage cost information map construction module is used to construct a construction robot passage cost information map based on basic indoor building structure feature information and construction robot body size information;

[0045] The construction robot semantic map generation module is used to obtain the indoor height information obtained by the two-dimensional laser as the original information set, and attach it to the navigation map after coordinate system conversion, and synthesize the information with the construction robot's travel cost information map to obtain the construction robot semantic map.

[0046] In a third aspect, an embodiment of the present invention provides a computer-readable storage medium, wherein the computer-readable storage medium stores computer instructions, and the computer instructions are used to enable the computer to execute the construction robot semantic map generation method of the first aspect of the embodiment of the present invention.

[0047] In a fourth aspect, an embodiment of the present invention provides a computer device, comprising: a memory and a processor, wherein the memory and the processor are communicatively connected to each other, the memory stores computer instructions, and the processor executes the construction robot semantic map generation method of the first aspect of the embodiment of the present invention by executing the computer instructions.

[0048] The technical solution of the present invention has the following advantages:

[0049] 1. The present invention provides a method and system for generating semantic maps of construction robots. First, a plurality of grid maps at different heights are obtained, and then the grid maps at different heights are processed to extract basic indoor building structure feature information. Basic information input is provided for automatic path generation, and these semantics are interrelated, providing a good interface and providing favorable guarantees for secondary development, which has good effects on both two-dimensional traversal path generation and special work point generation.

[0050] 2. The present invention provides a method and system for generating a semantic map of a construction robot. According to basic indoor building structure feature information and construction robot body size information, a construction robot travel cost information map is constructed. By combining a two-dimensional map with a construction robot travel cost information map with three-dimensional information to obtain a semantic map of the construction robot, the robot can obtain three-dimensional information of the space without three-dimensional laser navigation. During real-time path planning, a map can be used to efficiently and accurately obtain the passable area, the characteristics of the point area, the height of the point, and the spatial information near the point, thereby narrowing the search range of single-point navigation and dynamic planning, providing more types of effective information, and effectively improving the efficiency and safety of single-point navigation and dynamic planning.

[0051] 3. The present invention provides a method and system for generating a semantic map of a construction robot. The generated semantic map of a construction robot will extract door areas and narrow corridor areas, increase the travel cost of these areas, and consider the path comprehensively when generating the path. In addition, the inaccessible areas have been screened out, reducing the planning scope of the path-finding algorithm, so that the path takes into account both safety and efficiency. In addition, the semantic map will also divide the map into different areas. When connecting points when the order of the working points is not sensitive, it works according to the nearest order between the areas instead of simply working according to the order of the points, which can further improve the work efficiency of the machine. BRIEF DESCRIPTION OF THE DRAWINGS

[0052] In order to more clearly illustrate the specific implementation methods of the present invention or the technical solutions in the prior art, the drawings required for use in the specific implementation methods or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are some implementation methods of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0053] Figure 1 A workflow diagram of a specific example of a method for generating a semantic map for a construction robot provided in an embodiment of the present invention;

[0054] Figure 2It is an indoor building structure feature information diagram provided in an embodiment of the present invention;

[0055] Figure 3 A module composition diagram of a construction robot semantic map generation system provided by an embodiment of the present invention;

[0056] Figure 4 A composition diagram of a specific example of a computer device provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0057] The technical solution of the present invention will be described clearly and completely below in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0058] In addition, the technical features involved in the different embodiments of the present invention described below can be combined with each other as long as they do not conflict with each other.

[0059] Example 1

[0060] The spatial three-dimensional information of the working environment of the construction robot is relatively fixed. According to this feature of the construction robot, the embodiment of the present invention provides a method for generating a semantic map of the construction robot, which constructs a two-dimensional navigation map with three-dimensional spatial information, while meeting the automatic navigation requirements of the construction robot, reducing costs and improving navigation efficiency. Figure 1 As shown, the method comprises the following steps:

[0061] Step S1: Obtain multiple grid maps at different altitudes.

[0062] In an embodiment of the present invention, multiple grid maps at different altitudes are obtained, including the following five maps:

[0063] (1) Obtain a grid map with a height higher than the upper edge of the door and lower than the lower edge of the beam as the ground Figure 1 ;

[0064] (2) Obtain a grid map with a height higher than the bottom edge of the floating window and lower than the top edge of the floating window as the map Figure 2 ;

[0065] (3) Obtain a grid map with a height lower than the bottom edge of the floating window as the ground Figure 3 ;

[0066] (4) Get a ceiling height information map as the height Figure 1 ;

[0067] (5) Get a floating window height information map as the height Figure 2 .

[0068] Among them, for the ground Figure 1 ,land Figure 2 ,land Figure 3 The input image can be in the form of various data sources, including: architectural CAD drawings, BIM drawings, or maps constructed based on measured 3D laser data.

[0069] If the ground Figure 1 ,land Figure 2 ,land Figure 3 The input image is an architectural CAD drawing. All graphic information belonging to the wall layer and the bay window layer in the solid segment part is drawn onto an image with the wall color represented by the navigation map (for example: 8-bit image pixel value 0), and the room information surrounded by the wall is drawn with the color of the passable area represented by the navigation map (for example: 8-bit image pixel value 255), and the rest is drawn with the color of the unknown area represented by the navigation map (for example: 8-bit image pixel value 127). In this way, the architectural CAD drawing can be converted into a raster map. The raster map generated based on the architectural CAD drawing reduces the cost, greatly reduces the computing power requirements, and improves the navigation efficiency.

[0070] If the ground Figure 1 ,land Figure 2 ,land Figure 3 The input image is a BIM map or a map built based on actual measured 3D laser data. Because it is a raster map itself, when using a 2D raster map for processing, the amount of data processed is greatly reduced compared to processing a section of a 3D image. Moreover, when looking for architectural features, only the 2D feature relationship needs to be considered, and there is no need for a large amount of 3D feature solving, which reduces the complexity of the program and the required computing power.

[0071] Step S2: Process multiple grid maps at different heights to extract basic indoor building structure feature information.

[0072] In the embodiment of the present invention, Figure 1 ,land Figure 2 ,land Figure 3 The noise reduction process, edge detection (such as canny edge detection, which is used as an example only and is not limited to this) and the detected contour smoothing process are respectively performed to obtain the original contour set, including: Figure 1 ,land Figure 2 ,land Figure 3 The corresponding indoor ceiling area contour set, indoor medium height area contour set, and indoor low height area contour set are respectively based on the original contour set or combined with the height Figure 1 or combined with height Figure 2, extract door information, bay window information, window information, narrow passage information, beam information and ceiling information, and associate the relative relationships between different building structures.

[0073] Since the 2D laser navigation system cannot obtain 3D information by 2D laser alone, it cannot obtain the building structure feature information required for the production process path through its own laser, so the machine cannot automatically generate work sites, and a large amount of manual participation is required. In order to solve this problem, the embodiment of the present invention extracts the building structure features that the construction robot is concerned about from several input grid maps.

[0074] In the present embodiment, door information is extracted based on the indoor ceiling area contour set, the indoor medium height area contour set and the indoor low height area contour set. The extraction process includes:

[0075] Step 211: Obtain the initial door area, use the indoor medium height area contour set and the indoor low height area contour set to perform door feature detection (detect four corner points of the door that have a special relationship, such as the four corner points of the door), and save the identified door area information as an indoor medium height area door area set and an indoor low height area door area set, respectively.

[0076] Step 212: Obtain a supplementary door area, compare the indoor ceiling area contour set with the indoor medium height area contour set and the indoor low height area contour set respectively, screen out suspicious door areas corresponding to the indoor medium height area and the indoor low height area, and dilate the suspicious door area. If the four corner points of the door are within the indoor ceiling area contour set and are distributed in at least two different contours, save them in the corresponding door area set, and use the door areas identified by comparison as a supplementary set of indoor medium height area door areas and a supplementary set of indoor low height area door areas.

[0077] Step 213: Reintegrate the door areas, compare the indoor medium height area door area supplement set and the indoor low height area door area supplement set with the indoor medium height area outline set and the indoor low height area outline set respectively, complete the door information not recognized by the indoor medium height area outline set and the indoor low height area outline set, and compare the indoor medium height area door area set and the indoor low height area door area set. If the door area under the medium height area cannot find the corresponding door area in the door area of ​​the low height area, then remove the door area under the medium height area from the indoor medium height area door area set and add it to the window outline set.

[0078] Step 214: Preliminary segmentation of the area information, using the identified door information to segment the indoor medium-height area contour set and the indoor low-height area contour set, and adding the segmented contours to the original set. If a contour has multiple door areas, they are segmented in the order of the doors.

[0079] The embodiment of the present invention is based on the indoor medium height area contour set, the indoor low height area contour set and the height Figure 2 Extract pop-up window information. The extraction process includes:

[0080] Step 221: extract the bay window contour set, compare the indoor medium height area contour set and the indoor low height area contour set, extract the corresponding bay window information, and save all the bay window information to the bay window contour set; Figure 2 The height information of the floating window in the image is extracted, and the height information corresponding to each floating window outline is extracted, and the highest point in the outline area is taken as the floating window sill height information and saved in the floating window information set.

[0081] Step 222: The second splitting of the area information uses the bay window information to split the contour set of the medium-height area in the room. If the bay window information is detected under a certain contour, the bay window feature area contained in this contour is deleted, and then a new contour is generated from the remaining part of the contour to replace the original contour.

[0082] Window information is extracted based on the indoor medium-height area contour set and the indoor low-height area contour set. The extraction process includes:

[0083] Step 231: Extract the window contour set, perform window feature detection based on the indoor medium-height area contour set and the indoor low-height area contour set, and obtain the window contour set; in this embodiment, the window feature is also determined by the relationship between the four corner points of the window, that is, the four corner points are in the order of positive corner (protruding corner on the wall)-negative corner (recessed corner on the wall)-negative corner-positive corner.

[0084] Step 232: The third splitting of the area information uses the window contour set to split the indoor medium-height area contour set and the indoor low-height area contour set. If window area information is detected under a certain contour, the window feature area contained in this contour is deleted, and then a new contour is generated from the remaining part of the contour to replace the original contour.

[0085] It should be noted that the embodiment of the present invention performs multiple splitting processes on the region to remove redundant building feature information from the regional outline information, only retaining the area within the region where the robot body can pass, and saving the removed building feature information into the corresponding building information set, which can achieve three purposes: first, for removing door features, the entire outline can be divided into rooms. If not removed, the entire floor may be in the same outline; second, it makes it easier and more convenient for the robot body to plan the path later, without having to consider whether there are bay windows or the like in the regional outline that hinder the body's progress; third, when generating process paths, the data source classification is clearer, and no certain feature appears in multiple sets.

[0086] The embodiment of the present invention extracts narrow channel information based on an indoor medium-height area contour set and an indoor low-height area contour set, and the extraction process includes:

[0087] Step 241: traverse the indoor medium height area contour set and the indoor low height area contour set, save the channels in the contour that belong to the set threshold range separately, add a new contour to the original set (for example, those found from the indoor medium height area contour set are added to the indoor medium height area contour set, and those found from the indoor low height area contour set are added to the indoor low height area contour set), and mark the corresponding newly added contour information set in the original contour set as a narrow channel, and save the remaining part to replace the original contour (for example: the contour of the area originally containing the narrow channel area is counted as contour A, and the area of ​​contour A is larger than the area of ​​the identified narrow channel, that is, in addition to the narrow channel, there is a part of other areas in contour A, and the identified narrow channel is counted as contour B, and the remaining part of contour A after removing contour B is counted as contour C, and finally the information of contour B is added to the original contour set and the information of contour C replaces the information of contour A). In this embodiment, the narrow channel is determined and the threshold range is set according to the size of the robot body, and the width is generally set to 1.1m (for example only, not limited to this).

[0088] The embodiment of the present invention is based on the indoor medium height area contour set, the indoor low height area contour set and the height Figure 1 Information extraction Ceiling and beam information, the extraction process includes:

[0089] Step 251: Based on the indoor medium height area contour set, the indoor low height area contour set and the height Figure 1The ceiling height information is extracted, and the height information corresponding to each contour is extracted. The extracted height information is traversed and counted, and the connected area with an area greater than the preset area threshold and a height greater than the preset height threshold is taken as the ceiling, and the rectangular contour feature with a height less than the preset height threshold is taken as the beam (in this embodiment, the highest height of the bottom surface of the beam is taken as the preset height threshold, and the preset area threshold is reasonably set according to the actual building area, and is not limited here). The extracted ceiling height set and beam area contour set are respectively associated with their corresponding contour areas and saved. For example: Figure 2 In the figure, three beam information are detected in the height information corresponding to the contour of area No. 1 in the indoor medium-height area contour set, and the architectural feature information of area No. 1 in the indoor medium-height area contour set stores the three beam information and the ceiling information of this area.

[0090] The process of associating the relative relationships between different building structures in the embodiment of the present invention includes: obtaining the association relationship between doors and areas and determining the belonging relationship of bay windows, wherein the process of obtaining the association relationship between doors and areas is:

[0091] Step 261: Perform connection area detection on the indoor medium height area door area set and the indoor low height area door area set, respectively, and the indoor medium height area contour set and the indoor low height area contour set, and add the detection information to the indoor medium height area door area information set and the indoor low height area door area information set. Figure 2 In the process, detect the door area No. 1 in the door area set of medium-height areas in the room, expand the door area No. 1, detect the two areas where the four corner points are located, and save the two areas to the connection area 1 and the connection area 2 in the door connection information. When the vehicle body enters the two areas, the angle of the vehicle body is saved as the angle of entering area 1 and the angle of entering area 2 respectively. At the same time, the length, width, center point and angle of the door are saved. After obtaining the association information between the door and the area, it can be known whether any two areas are passable, which provides convenience for single-point navigation and cross-area process path generation.

[0092] The process of determining the relationship of the floating window in the embodiment of the present invention is as follows:

[0093] Step 262: Connect the floating window outline set with the indoor medium height area outline set and the indoor low height area outline set, and add the detected area information to the floating window information set. Figure 2In the process, the outline of the floating window outline set No. 1 is detected, the area of ​​door No. 1 is expanded, the area where the two corner points without windows are located in the four corner points is detected, and the area number is saved in the floating window information set. The association information between the floating window and the area can avoid the collision between the vehicle body and the floating window. For example: when the laser radar scanning plane is higher than the bottom of the floating window, the bottom of the floating window is displayed on the map as a passable area. If there is no detection value input of the corresponding sensor at the bottom layer during dynamic obstacle avoidance, it will cause the vehicle body to collide with the floating window. However, if the floating window information set is queried during planning to see if there is a floating window connected to the current area, if there is, the floating window range is obtained for avoidance, which improves safety.

[0094] After processing multiple grid maps at different heights, the embodiment of the present invention can extract architectural feature information such as doors, windows, bay windows, beams, narrow passages, ceilings, etc., providing basic information input for automatic path generation. In addition, these semantics are interrelated, providing a good interface, and providing favorable guarantees for secondary development, which has good effects on both two-dimensional traversal path generation and special work point generation.

[0095] Step S3: Construct a construction robot travel cost information map based on basic indoor building structure feature information and construction robot body size information.

[0096] Since most indoor construction robots are large in size, low path safety margins can easily lead to problems when passing through doors or narrow corridors. In traditional single-point navigation and dynamic planning, the cost of the traversable area is the same, and the planned path may repeatedly pass through narrow areas, increasing the probability of robot failure. Therefore, it is necessary to increase the cost of passing through doors and corridors to reduce the cost of passing through these relatively dangerous areas. A weighted path planning map is needed as a map of the robot's travel cost information.

[0097] The embodiment of the present invention draws a solid outline of the original contour set, that is, the ordinary area surrounded by the contour in the original contour set is drawn on the map with a preset pixel value, the pixel value of the narrow channel information in the original contour information set is assigned to the first preset multiple of the ordinary area value and drawn on the map, and the pixel value of the door area is drawn on the map with the second preset multiple of the ordinary area value, so as to obtain a construction robot passage cost information map, wherein the first preset multiple is determined according to the robot body size, and the second preset multiple is not less than the first preset multiple. The larger the body, the larger the value of the first preset multiple. In a specific embodiment, the pixel value of the door area is set to between 1.3 and 2.5 times the pixel value of the ordinary area and drawn on the map, and then the door area is drawn on the image with 3 times the pixel value of the ordinary area, so that the cost required to pass through the pixel point can be known by the numerical value of the pixel point during planning.

[0098] Step S4: The indoor height information obtained by the two-dimensional laser is used as the original information set, and after the coordinate system conversion, it is attached to the navigation map, and the information is synthesized with the construction robot's travel cost information map to obtain the construction robot's semantic map.

[0099] When a traditional robot is navigating in two dimensions, it only uses a map with the same height as its own laser radar as the basis for its own positioning and path planning. At this time, when doing single-point navigation or dynamic planning, the problem of vehicle body collision may occur. For example, when the laser radar is too high above the bottom of the floating window, the bottom of the floating window is displayed on the map as a passable area. If there is no detection value input of the corresponding sensor in the bottom layer during dynamic obstacle avoidance, it will cause the vehicle body to collide with the floating window. The navigation map in the embodiment of the present invention can be a map of robot travel cost information, or it can be an ordinary two-dimensional navigation map. By combining the two-dimensional map with the construction robot travel cost information map with three-dimensional information to obtain a construction robot semantic map, the robot can obtain three-dimensional information of the space without three-dimensional laser navigation. When using a map for real-time path planning, the passable area, the characteristics of the point area, the height of the point and the spatial information near the point can be obtained efficiently and accurately, which narrows the search range of single-point navigation and dynamic planning, provides more types of effective information, and effectively improves the efficiency and safety of single-point navigation and dynamic planning.

[0100] In practical applications, most indoor building robots are large in size. When passing through doors or narrow corridors, the path safety margin is low, which can easily lead to navigation failure problems. In traditional single-point navigation and dynamic planning, the cost of the passable area is the same, and the planned path may repeatedly pass through the narrow area, increasing the probability of robot failure. The construction robot semantic map generated by the construction robot semantic map generation method provided in the embodiment of the present invention (semantics refers to building mechanism feature information such as doors, windows, and passages) will extract the door area and the narrow passage area, increase the travel cost of these areas, so that the path is considered in a comprehensive way when generating the path. In addition, the inaccessible area has been screened out, reducing the planning scope of the path-finding algorithm, so that the path takes into account both safety and efficiency. In addition, the semantic map will also divide the map into different areas. When connecting points when the working point order is not sensitive, the robot can be further improved in terms of the nearest order between areas instead of simply working in the order of points.

[0101] Example 2

[0102] The embodiment of the present invention provides a construction robot semantic map generation system, such as Figure 3 As shown, including:

[0103] The map data source acquisition module 1 is used to obtain multiple grid maps at different altitudes; this module executes the method described in step S1 of embodiment 1, which will not be repeated here.

[0104] The building structure characteristic information acquisition module 2 is used to process multiple grid maps at different heights to extract basic indoor building structure characteristic information; this module executes the method described in step S2 of embodiment 1, which will not be repeated here.

[0105] The robot travel cost information map construction module 3 is used to construct a construction robot travel cost information map based on basic indoor building structure feature information and construction robot body size information; this module executes the method described in step S3 of Example 1 and will not be repeated here.

[0106] The construction robot semantic map generation module 4 is used to use the indoor height information obtained by the two-dimensional laser as the original information set, and after the coordinate system conversion, it is attached to the navigation map, and the information is synthesized with the construction robot passage cost information map to obtain the construction robot semantic map. This module executes the method described in step S4 of embodiment 1, which will not be repeated here.

[0107] The semantic map generation system for construction robots provided by the embodiment of the present invention generates a semantic map for construction robots, which extracts door areas and narrow corridor areas, increases the travel costs of these areas, and considers the path comprehensively for safety when generating the path. In addition, the inaccessible areas have been screened out, reducing the planning scope of the path-finding algorithm, so that the path takes into account both safety and efficiency. In addition, the semantic map will also divide the map into different areas. When connecting points in the case where the order of the working points is not sensitive, it works according to the closest order between the areas instead of simply working according to the order of the points, which can further improve the work efficiency of the machine.

[0108] Example 3

[0109] An embodiment of the present invention provides a computer device, such as Figure 4 As shown, the device may include a processor 51 and a memory 52, wherein the processor 51 and the memory 52 may be connected via a bus or other means. Figure 4 Take the example of connecting via a bus.

[0110] The memory 52 is a non-transitory computer-readable storage medium that can be used to store non-transitory software programs, non-transitory computer executable programs and modules, such as the corresponding program instructions / modules in the embodiments of the present invention. The processor 51 executes various functional applications and data processing of the processor by running the non-transitory software programs, instructions and modules stored in the memory 52, that is, the construction robot semantic map generation method in the above method embodiment 1 is implemented.

[0111] The memory 52 may include a program storage area and a data storage area, wherein the program storage area may store an operating system and applications required for at least one function; the data storage area may store data created by the processor 51, etc. In addition, the memory 52 may include a high-speed random access memory, and may also include a non-volatile memory, such as at least one disk storage device, a flash memory device, or other non-volatile solid-state storage devices. In some embodiments, the memory 52 may optionally include a memory remotely arranged relative to the processor 51, and these remote memories may be connected to the processor 51 via a network. Examples of the above-mentioned network include, but are not limited to, the Internet, an intranet, an intranet, a mobile communication network, and combinations thereof.

[0112] One or more modules are stored in the memory 52, and when executed by the processor 51, the construction robot semantic map generation method in Example 1 is executed.

[0113] The specific details of the above-mentioned computer device can be understood by referring to the corresponding related descriptions and effects in Example 1, and will not be repeated here.

[0114] Those skilled in the art can understand that the implementation of all or part of the processes in the above-mentioned embodiments can be completed by instructing the relevant hardware through a computer program. The program can be stored in a computer-readable storage medium, and when the program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, the storage medium can be a disk, an optical disk, a read-only memory (ROM), a random access memory (RAM), a flash memory, a hard disk drive (HDD) or a solid-state drive (SSD), etc.; the storage medium can also include a combination of the above-mentioned types of memory.

[0115] Obviously, the above embodiments are merely examples for the purpose of clear explanation, and are not intended to limit the implementation methods. For those skilled in the art, other different forms of changes or modifications can be made based on the above description. It is not necessary and impossible to list all the implementation methods here. The obvious changes or modifications derived from these are still within the scope of protection of the invention.

Claims

1. A method for generating a semantic map of a construction robot, characterized in that: The following steps are involved: Get multiple raster maps at different altitudes; Process multiple grid maps at different heights to extract basic indoor building structure feature information; According to the basic indoor building structure feature information and the construction robot body size information, a construction robot passage cost information map is constructed; The indoor height information obtained by two-dimensional laser is used as the original information set, and after coordinate system conversion, it is attached to the navigation map and synthesized with the construction robot's travel cost information map to obtain the construction robot's semantic map.

2. The method for generating a semantic map of a construction robot according to claim 1, characterized in that: The obtaining of multiple grid maps at different heights includes: Obtain a grid map with a height higher than the upper edge of the door and lower than the lower edge of the beam as map 1; Obtain a grid map with a height higher than the bottom edge of the floating window and lower than the top edge of the floating window as map 2; Obtain a grid map whose height is lower than the bottom edge of the floating window as map three; Obtain a ceiling height information map as height map 1; Get a floating window height information map as height map 2.

3. The method for generating a semantic map of a construction robot according to claim 2, characterized in that: The process of processing multiple grid maps at different heights to extract basic indoor building structure feature information includes: Map 1, Map 2, and Map 3 are subjected to noise reduction, edge detection, and detected contour smoothing, respectively, to obtain an original contour set, including: an indoor ceiling area contour set, an indoor medium-height area contour set, and an indoor low-height area contour set corresponding to Map 1, Map 2, and Map 3, respectively; Based on the original contour set or in combination with the height map 1 or in combination with the height map 2, door information, bay window information, window information, narrow passage information, beam information and ceiling information are extracted, and the relative relationships between different building structures are associated.

4. The method for generating a semantic map of a construction robot according to claim 3, characterized in that: Door information is extracted based on the indoor ceiling area contour set, the indoor medium-height area contour set, and the indoor low-height area contour set. The extraction process includes: Use the indoor medium height area contour set and the indoor low height area contour set to perform door feature detection, and save the identified door area information as the indoor medium height area door area set and the indoor low height area door area set respectively; The indoor ceiling area contour set is compared with the indoor medium height area contour set and the indoor low height area contour set respectively, and the suspicious door areas corresponding to the indoor medium height area and the indoor low height area are screened out, and the suspicious door areas are expanded. If the four corner points of the door are in the indoor ceiling area contour set and are distributed in at least two different contours, they are saved in the corresponding door area set, and the door areas identified by comparison are used as the indoor medium height area door area supplementary set and the indoor low height area door area supplementary set respectively; Compare the indoor medium height area door area supplement set and the indoor low height area door area supplement set with the indoor medium height area outline set and the indoor low height area outline set respectively, complete the door information not recognized by the indoor medium height area outline set and the indoor low height area outline set, and compare the indoor medium height area door area set and the indoor low height area door area set. If the door area under the medium height area cannot find the corresponding door area in the door area of ​​the low height area, remove the door area under the medium height area from the indoor medium height area door area set and add it to the window outline set; Use the identified door information to split the indoor medium-height area contour set and the indoor low-height area contour set, and add the split contours to the original set. If a contour has multiple door areas, split them in the order of the doors.

5. The method for generating a semantic map of a construction robot according to claim 3, characterized in that: The bay window information is extracted based on the indoor medium-height area contour set, the indoor low-height area contour set and the height map 2. The extraction process includes: Compare the indoor medium-height area contour set and the indoor low-height area contour set, extract the corresponding floating window information, and save all the floating window information into the floating window contour set; According to the floating window outline set and the floating window height information in the height map 2, the height information corresponding to each floating window outline is extracted, and the highest point in the outline area is taken as the floating window sill height information and saved in the floating window information set; The bay window information is used to split the contour set of the medium-height area in the room. If the bay window information is detected under a certain contour, the bay window feature area contained in this contour is deleted, and then a new contour is generated from the remaining part of the contour to replace the original contour.

6. The method for generating a semantic map of a construction robot according to claim 3, characterized in that: Window information is extracted based on the indoor medium-height area contour set and the indoor low-height area contour set. The extraction process includes: Detect window features based on the indoor medium-height area contour set and the indoor low-height area contour set to obtain a window contour set; The window contour set is used to split the indoor medium-height area contour set and the indoor low-height area contour set. If window area information is detected under a certain contour, the window feature area contained in this contour is deleted, and then a new contour is generated from the remaining part of the contour to replace the original contour.

7. The method for generating a semantic map of a construction robot according to claim 3, characterized in that: Based on the indoor medium-height area contour set and the indoor low-height area contour set, the narrow channel information is extracted. The extraction process includes: Traverse the indoor medium-height area contour set and the indoor low-height area contour set, save the channels in the contour that belong to the set threshold range separately, add a new contour to the original set, and mark the corresponding newly added contour information set in the original contour set as a narrow channel, and save the remaining part to replace the original contour.

8. The method for generating a semantic map of a construction robot according to claim 3, characterized in that: Based on the indoor medium-height area contour set, the indoor low-height area contour set and the height map information, the ceiling and beam information is extracted. The extraction process includes: According to the indoor medium-height area contour set, the indoor low-height area contour set and the ceiling height information of height map 1, the height information corresponding to each contour is extracted, and the extracted height information is traversed and counted. The connected area with an area greater than the preset area threshold and a height greater than the preset height threshold is taken as the ceiling, and the rectangular contour feature with a height less than the preset height threshold is taken as the beam. The extracted ceiling height set and beam area contour set are respectively associated with their corresponding contour areas and saved.

9. The method for generating a semantic map of a construction robot according to claim 4, characterized in that: The process of associating the relative relationships between different building structures includes: obtaining the association relationship between doors and areas, and the association process is: The indoor medium height area door area set and the indoor low height area door area set are respectively connected to the indoor medium height area contour set and the indoor low height area contour set for area detection, and the detection information is respectively added to the indoor medium height area door area information set and the indoor low height area door area information set.

10. The method for generating a semantic map of a construction robot according to claim 5, characterized in that: The process of associating the relative relationships between different building structures includes: determining the belonging relationship of the bay windows, and the associating process is: The floating window contour set is connected with the indoor medium-height area contour set and the indoor low-height area contour set for detection, and the detected area information is added to the floating window information set.

11. The method for generating a semantic map of a construction robot according to claim 7, characterized in that: The process of constructing a construction robot passage cost information map based on basic indoor building structure feature information and construction robot body size information includes: The ordinary area enclosed by the contours in the original contour set is drawn on the map with preset pixel values, the pixel values ​​of the narrow channel information in the original contour information set are assigned to the first preset multiples of the ordinary area values ​​and drawn on the map, and the pixel values ​​of the door area are drawn on the map with the second preset multiples of the ordinary area values ​​to obtain a construction robot travel cost information map, wherein the first preset multiple is determined according to the robot body size, and the second preset multiple is not less than the first preset multiple.

12. A construction robot semantic map generation system, characterized in that: include: A map data source acquisition module is used to obtain multiple raster maps at different altitudes; The building structure feature information acquisition module is used to process multiple grid maps at different heights and extract basic indoor building structure feature information; A robot passage cost information map construction module is used to construct a construction robot passage cost information map based on basic indoor building structure feature information and construction robot body size information; The construction robot semantic map generation module is used to obtain the indoor height information obtained by the two-dimensional laser as the original information set, and attach it to the navigation map after coordinate system conversion, and synthesize the information with the construction robot's travel cost information map to obtain the construction robot semantic map.

13. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores computer instructions, and the computer instructions are used to enable the computer to execute the construction robot semantic map generation method as described in any one of claims 1-11.

14. A computer device, characterized in that: include: A memory and a processor, wherein the memory and the processor are communicatively connected to each other, the memory stores computer instructions, and the processor executes the construction robot semantic map generation method as described in any one of claims 1-11 by executing the computer instructions.

Images