Robot control method and system based on self-intelligent navigation

Abstract

The invention discloses a robot control method and system based on self-intelligent navigation, and belongs to the technical field of robot control, and the method comprises the following steps: S1, continuously collecting environment images; s2, constructing an environment view; s3, dividing a plurality of sub-environment views; s4, defining attributes of a plurality of sub-environment views; s5, inputting the target position environment image into the robot; s6, positioning a target position; s7, acquiring an environment image in an initial navigation state; s8, positioning the position of the robot; s9, planning a navigation path of the robot based on the position of the robot, the target position and the attributes of a plurality of sub-environment views in the environment view; s10, the robot performs navigation control based on the planned navigation path, and the robot moves; according to the method, the straight line between the robot position and the target position is optimized based on the sub-environment view attributes, environment data does not need to be repeatedly collected for planning, the data processing amount is small, the planning speed is high, and the navigation control speed is high.

Description

Technical Field

[0001] The present invention belongs to the technical field of robot control, and in particular relates to a robot control method and system based on self-intelligent navigation. Background Art

[0002] A robot refers to a machine system that can perform tasks automatically and is widely used in industries such as industry, medicine, military, services, and scientific research.

[0003] Autonomous control of robots, especially autonomous navigation control, is the key to improving the robot's own intelligence level and application efficiency.

[0004] Chinese patent application No. 202410472650.X discloses a zero-sample self-intelligent navigation robot control method, system, medium and product, including: semantically segmenting the instructions to be executed to obtain a vocabulary group of instructions to be executed, wherein the instructions to be executed include a target object; inputting a first image and the vocabulary group of instructions to be executed into a SAM model to generate a semantic map with a multi-dimensional feature channel, wherein the first image is collected during the movement of the robot, and the SAM model includes an encoder and a decoder; based on the semantic map with a multi-dimensional feature channel, an exploration area is obtained based on the robot position; path planning is performed on the exploration area to obtain a planned route; based on the planned route, the robot is controlled to move so that the robot completes the instructions to be executed.

[0005] The above technology has the following problems: Although the above self-intelligent navigation robot control method can achieve navigation with zero samples, data processing planning is required for each exploration area, the data processing volume is large, and the navigation speed is slow.

[0006] In view of this, a control method and system of a robot based on self-intelligent navigation are designed to solve the above problems. Summary of the invention

[0007] To solve the problems raised in the above background technology. The present invention provides a self-intelligent navigation robot control method and system, which has the characteristics of optimizing the straight line between the robot position and the target position based on the sub-environment view attribute, no need to repeatedly collect environmental data for planning, small data processing volume, fast planning speed, that is, fast navigation control speed.

[0008] Another object of the present invention is to provide a robot control system based on its own intelligent navigation.

[0009] To achieve the above object, the present invention provides the following technical solution: a control method of a robot based on self-intelligent navigation, comprising the following steps:

[0010] S1: The robot camera continuously collects environmental images;

[0011] S2: Constructing an environment view based on continuously acquired environment images;

[0012] S3: Divide the environment view into multiple sub-environment views based on the edge line of the environment view;

[0013] S4: Based on the robot collision factors, multiple sub-environment view attributes are defined, including movable areas and immovable areas, where the immovable areas are obstacle areas;

[0014] S5: inputting the target location environment image into the robot;

[0015] S6: The robot matches the target location environment image based on the environment view and locates the target location;

[0016] S7: The robot camera collects the environment image in the initial state of navigation;

[0017] S8: The robot matches the environment image based on the environment view and locates its own position;

[0018] S9: planning the navigation path of the robot based on the robot's own position, the target position, and the attributes of multiple sub-environment views in the environment view;

[0019] S10: The robot performs navigation control based on the planned navigation path, and the robot moves.

[0020] Furthermore, the specific steps of step S2 include:

[0021] S201: preprocessing the collected multiple environment images, including denoising and contrast enhancement;

[0022] S202: extracting and matching feature points of multiple environment images by SIFT algorithm;

[0023] S203: filtering out multiple environmental image feature points that are incorrectly matched by using the RANSC algorithm;

[0024] S204: Arrange and obtain feature matching point sets of multiple environment images, and calculate the global homography matrix between the multiple environment images, which is expressed as:

[0025]

[0026] Where: K i，j Represented as the camera's internal parameter matrix, R i,j is represented as the rotation parameter, I is represented as the unit matrix, (n j , d j ) is expressed as the plane parameter, T j,i Expressed as a translation parameter;

[0027] S205: multiple environment images are transformed based on the global homography matrix;

[0028] S206: construct an image segmentation energy function, and find the optimal stitching line of the converted multiple environment images based on the energy function. When the energy function value is compared with a preset threshold and meets the set condition, it is the optimal stitching line. The energy function expression is:

[0029] E=E1+E2

[0030] E1=|S T -S O |*l TO

[0031] E2=|S T -S O |*g TO

[0032] Where: S T Represents the entire image area, S O Represented as the image overlap area, l TO Expressed as color difference, g TO It is expressed as gradient difference;

[0033] S207: Stitching the multiple environment images based on the optimal stitching lines between them to form an environment view.

[0034] Furthermore, the specific steps of step S3 include:

[0035] S301: Detect edge lines of the environment view based on the Prewitt operator edge detection algorithm;

[0036] S302: Divide different sub-environment views based on the inside and outside of the closed edge line until all the environment views are divided.

[0037] Furthermore, in step S6 and step S8, the specific steps of locating the target position based on the environmental image matching the target position and locating the robot's own position based on the environmental view matching the environmental image include: preprocessing the environmental image and the environmental view according to steps S201-205, matching feature points, screening out mismatched feature points, calculating the global homography matrix and converting based on the global homography matrix, aligning the converted environmental image with the environmental view, and determining the positions of the target and the robot itself in the environmental view.

[0038] Furthermore, the specific steps of step S9 include:

[0039] S901: In the environment view, a straight line is drawn with the robot's own position as the starting point and the target position as the end point, as the robot's initial navigation path;

[0040] S902: Determine whether there is a sub-environment view of the immovable area on the initial navigation path. If not, the initial navigation path is the final navigation path. If so, the edge line of the sub-environment view of the immovable area close to the sub-environment view of the movable area is taken as the starting point, and the line that expands outward beyond the volume of the robot is taken as the final navigation path.

[0041] The system based on the self-intelligent navigation robot control method comprises:

[0042] An environmental image acquisition module, which acquires environmental images;

[0043] An environment view construction module, which constructs an environment view based on the collected environment image;

[0044] A sub-environment view division module, which divides the environment view into multiple sub-environment views based on the edge line of the environment view;

[0045] A sub-environment view attribute definition module defines multiple sub-environment view attributes based on robot collision factors;

[0046] A target location positioning module matches the target location environment image based on the environment view and locates the target location;

[0047] The robot position positioning module locates the robot's own position based on the matching of the environment image with the environment view;

[0048] A navigation path planning module plans the robot's navigation path based on the robot's own position, the target position, and the attributes of multiple sub-environment views in the environment view;

[0049] The navigation control module controls the movement of the robot based on the planned navigation path.

[0050] Compared with the prior art, the present invention has the following beneficial effects:

[0051] The present invention constructs an environment view based on the continuous acquisition of environment images by a robot camera, divides the sub-environment view based on the edge line of the environment view, and defines the sub-environment view attributes based on collision factors. During navigation, the robot position is defined based on the environment image acquired by the robot camera, a straight line connecting the robot position and the target position is used as an initial planned navigation path, and the navigation path is optimized based on the sub-environment view attributes to form a final navigation path. Navigation control is then performed based on the final navigation path. Compared with the prior art, the straight line between the robot position and the target position is optimized based on the sub-environment view attributes, and there is no need to repeatedly acquire environment data for planning. The data processing volume is small and the planning speed is fast, that is, the navigation control speed is fast. BRIEF DESCRIPTION OF THE DRAWINGS

[0052] Figure 1 is a flow chart of the method of the present invention;

[0053] Figure 2 It is a system framework diagram of the present invention;

[0054] Figure 3 A navigation route map for robot control of the present invention;

[0055] In the figure: 1. Environmental image acquisition module; 2. Environmental view construction module; 3. Sub-environment view division module; 4. Sub-environment view attribute definition module; 5. Target position positioning module; 6. Robot position positioning module; 7. Navigation path planning module; 8. Navigation control module. DETAILED DESCRIPTION

[0056] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. 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.

[0057] The present invention provides the following technical solutions: Figure 1 , a control method of a robot based on its own intelligent navigation, comprising the following steps:

[0058] S1: The robot camera continuously collects environmental images;

[0059] S2: Constructing an environment view based on continuously acquired environment images;

[0060] S3: Divide the environment view into multiple sub-environment views based on the edge line of the environment view;

[0061] S4: Based on the robot collision factors, multiple sub-environment view attributes are defined, including movable areas and immovable areas, where the immovable areas are obstacle areas;

[0062] S5: inputting the target location environment image into the robot;

[0063] S6: The robot matches the target location environment image based on the environment view and locates the target location;

[0064] S7: The robot camera collects the environment image in the initial state of navigation;

[0065] S8: The robot matches the environment image based on the environment view and locates its own position;

[0066] S9: planning the navigation path of the robot based on the robot's own position, the target position, and the attributes of multiple sub-environment views in the environment view;

[0067] S10: The robot performs navigation control based on the planned navigation path, and the robot moves.

[0068] Specifically, the specific steps of step S2 include:

[0069] S201: preprocessing the collected multiple environment images, including denoising and contrast enhancement;

[0070] In the process of digitization and transmission, digital images in reality are often affected by the interference of imaging devices and external environmental noise, thus generating noise, which may cause image distortion and quality degradation. Image denoising refers to the process of reducing noise in digital images, including mean filtering, median filtering and Gaussian filtering.

[0071] Image contrast enhancement refers to increasing the difference between the brightest and darkest parts of an image through processing technology, so that the visual effect of the image is more vivid and clear. Contrast enhancement can improve the clarity of the image, enhance the layering and depth of the image, and make it more vivid and visually impactful, including linear contrast broadening and histogram equalization;

[0072] S202: extracting and matching feature points of multiple environment images by SIFT algorithm;

[0073] SIFT algorithm, i.e. scale-invariant feature transform, processes the original image in scale space through Gaussian difference function to detect extreme points with different scales, and then uses DoG operator to detect these extreme points, i.e. compares the difference between two adjacent layers of Gaussian images in Gaussian pyramids of different scales and spatial positions, so as to obtain scale-invariant key points, and accurately locates the candidate key points with sub-pixel precision, determines the position of the key points with sub-pixel precision, calculates the gradient direction histogram in the area around the key points, selects the main direction as the direction of the key points, and assigns the direction to the key points, extracts local features in the area around the key points, and generates descriptors of the key points, and calculates the similarity between features using Euclidean distance by comparing the feature descriptors in different images, and performs feature matching;

[0074] S203: filtering out multiple environmental image feature points that are incorrectly matched by using the RANSC algorithm;

[0075] The RANSC algorithm, i.e., random sampling consistency, randomly selects four non-collinear matching point pairs from all matching point pairs as the inlier candidate set, calculates the homography matrix through these four matching point pairs, and uses them as the current model to calculate the projection errors of all other matching point pairs to the model. If the projection error of a matching point pair is less than the preset threshold, it is judged as an inlier, otherwise it is judged as an outlier, i.e., a mismatched point. The number of inliers in the current model is counted and compared with the previous optimal model. If the number of inliers in the current model is greater, the optimal model is updated as the current model. The above steps are repeated until the set number of iterations is reached or a sufficiently good model is found. The optimal homography matrix is ​​determined based on the number of inliers and the cost function recorded during the iteration process. All matching point pairs are screened based on the optimal model, and only the inliers with projection errors less than the preset threshold are retained as the final matching point pairs.

[0076] S204: Arrange and obtain feature matching point sets of multiple environment images, and calculate the global homography matrix between the multiple environment images, which is expressed as:

[0077]

[0078] Where: K i，j Represented as the camera's internal parameter matrix, R i,j is represented as the rotation parameter, I is represented as the unit matrix, (n j , d j ) is expressed as the plane parameter, T j,i Expressed as a translation parameter;

[0079] S205: multiple environment images are transformed based on the global homography matrix;

[0080] S206: construct an image segmentation energy function, and find the optimal stitching line of the converted multiple environment images based on the energy function. When the energy function value is compared with a preset threshold and meets the set condition, it is the optimal stitching line. The energy function expression is:

[0081] E=E1+E2

[0082] E1=|S T -S O |*l TO

[0083] E2=|S T -S O |*g TO

[0084] Where: S T Represents the entire image area, S O Represented as the image overlap area, l TO Expressed as color difference, g TOIt is expressed as gradient difference;

[0085] S207: Stitching the multiple environment images based on the optimal stitching lines between them to form an environment view.

[0086] Specifically, the specific steps of step S3 include:

[0087] S301: Detect edge lines of the environment view based on the Prewitt operator edge detection algorithm;

[0088] The Prewitt operator edge detection algorithm is an edge detection method based on the first-order differential. It uses the grayscale difference of the upper, lower, left, and right neighboring points of a pixel to reach an extreme value at the edge to detect the edge.

[0089] Define the horizontal edge detection operator as [[-1, 0, 1], [-1, 0, 1], [-1, 0, 1]], and the vertical edge detection operator as [[-1, -1, -1], [0, 0, 0, [1, 1, 1]], use the horizontal edge detection operator and the vertical edge detection operator as the convolution kernel, move pixel by pixel on the image, and calculate the convolution result at each pixel position. For each pixel position, calculate its horizontal and vertical gradient amplitudes, and then add these two gradient amplitudes to obtain the total gradient amplitude of the pixel, and compare it with the preset threshold. If the gradient amplitude is greater than or equal to the threshold, the pixel is marked as an edge point, otherwise, it is marked as a non-edge point;

[0090] S302: Dividing different sub-environment views based on the inside and outside of the closed edge line until all the environment views are divided;

[0091] As attached Figure 3 As shown, the black edge line is a closed edge line, and the black part and the white part are divided into different sub-environment views.

[0092] As attached Figure 3 As shown, since the black area is a wall, the robot has been hit here during the movement, so the attribute of the black part is an immovable area, and the white area has not been hit during the movement of the robot, so the attribute of the white part is a movable area.

[0093] Specifically, in step S6 and step S8, the specific steps of locating the target position based on the environmental image matching the target position and locating the robot's own position based on the environmental view matching the environmental image include: preprocessing the environmental image and the environmental view according to steps S201-205, matching feature points, screening out mismatched feature points, calculating the global homography matrix and converting based on the global homography matrix, aligning the converted environmental image with the environmental view, and determining the positions of the target and the robot itself in the environmental view.

[0094] Specifically, the specific steps of step S9 include:

[0095] S901: In the environment view, a straight line is drawn with the robot's own position as the starting point and the target position as the end point, as the robot's initial navigation path;

[0096] As attached Figure 3 As shown, the initial navigation path is the solid line part;

[0097] S902: Determine whether there is a sub-environment view of the immovable area on the initial navigation path. If not, the initial navigation path is the final navigation path. If so, the edge line of the sub-environment view of the immovable area close to the sub-environment view of the movable area is taken as the starting point, and the line that exceeds the volume of the robot is expanded outward as the final navigation path;

[0098] As attached Figure 3 As shown, the final navigation path is the combination of the solid line and the dotted line, wherein the outward expansion area based on the edge line is preset.

[0099] See attached Figure 2 , a system based on a self-intelligent navigation robot control method, comprising:

[0100] Environmental image acquisition module 1, collecting environmental images;

[0101] An environment view construction module 2 constructs an environment view based on the collected environment image;

[0102] A sub-environment view division module 3, which divides the environment view into a plurality of sub-environment views based on the edge line of the environment view;

[0103] A sub-environment view attribute definition module 4 is used to define multiple sub-environment view attributes based on robot collision factors;

[0104] A target position positioning module 5 is configured to match the target position environment image based on the environment view and locate the target position;

[0105] The robot position positioning module 6 locates the robot's own position based on the matching of the environment image with the environment view;

[0106] A navigation path planning module 7 plans the navigation path of the robot based on the robot's own position, the target position and the attributes of multiple sub-environment views in the environment view;

[0107] The navigation control module 8 controls the movement of the robot based on the planned navigation path.

[0108] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A control method for a robot based on its own intelligent navigation, characterized in that: The following steps are involved: S1: The robot camera continuously collects environmental images; S2: Constructing an environment view based on continuously acquired environment images; S3: Divide the environment view into multiple sub-environment views based on the edge line of the environment view; S4: Based on the robot collision factors, multiple sub-environment view attributes are defined, including movable areas and immovable areas, where the immovable areas are obstacle areas; S5: inputting the target location environment image into the robot; S6: The robot matches the target location environment image based on the environment view and locates the target location; S7: The robot camera collects the environment image in the initial state of navigation; S8: The robot matches the environment image based on the environment view and locates its own position; S9: planning the navigation path of the robot based on the robot's own position, the target position, and the attributes of multiple sub-environment views in the environment view; S10: The robot performs navigation control based on the planned navigation path, and the robot moves.

2. A control method for a robot based on self-intelligent navigation according to claim 1, characterized in that: The specific steps of step S2 include: S201: preprocessing the collected multiple environment images, including denoising and contrast enhancement; S202: extracting and matching feature points of multiple environment images by SIFT algorithm; S203: filtering out multiple environmental image feature points that are incorrectly matched by using the RANSC algorithm; S204: Arrange and obtain feature matching point sets of multiple environment images, and calculate the global homography matrix between the multiple environment images, which is expressed as: Where: K i，j Represented as the camera's internal parameter matrix, R i,j is represented as the rotation parameter, I is represented as the unit matrix, (n j , d j ) is expressed as the plane parameter, T j,i Expressed as a translation parameter; S205: multiple environment images are transformed based on the global homography matrix; S206: construct an image segmentation energy function, and find the optimal stitching line of the converted multiple environment images based on the energy function. When the energy function value is compared with a preset threshold and meets the set condition, it is the optimal stitching line. The energy function expression is: E=E1+E2 E1=|S T -S O |*l TO E2=|S T -S O |*g TO Where: S T Represents the entire image area, S O Represented as the image overlap area, l TO Expressed as color difference, g TO It is expressed as gradient difference; S207: Stitching the multiple environment images based on the optimal stitching lines between them to form an environment view.

3. A control method for a robot based on self-intelligent navigation according to claim 1, characterized in that: The specific steps of step S3 include: S301: Detect edge lines of the environment view based on the Prewitt operator edge detection algorithm; S302: Divide different sub-environment views based on the inside and outside of the closed edge line until all the environment views are divided.

4. A control method for a robot based on self-intelligent navigation according to claim 2, characterized in that: In step S6 and step S8, the specific steps of locating the target position based on the environmental image matching the target position of the environmental view and locating the robot's own position based on the environmental view matching the environmental image include: preprocessing the environmental image and the environmental view according to steps S201-205, matching feature points, screening out mismatched feature points, calculating the global homography matrix and converting based on the global homography matrix, aligning the converted environmental image with the environmental view, and determining the positions of the target and the robot itself in the environmental view.

5. The control method of a robot based on self-intelligent navigation according to claim 1, characterized in that: The specific steps of step S9 include: S901: In the environment view, a straight line is drawn with the robot's own position as the starting point and the target position as the end point, as the robot's initial navigation path; S902: Determine whether there is a sub-environment view of the immovable area on the initial navigation path. If not, the initial navigation path is the final navigation path. If so, the edge line of the sub-environment view of the immovable area close to the sub-environment view of the movable area is taken as the starting point, and the line that expands outward beyond the volume of the robot is taken as the final navigation path.

6. A system based on a self-intelligent navigation robot control method according to claims 1-5, characterized in that: include: An environmental image acquisition module (1) is used to acquire environmental images; An environment view construction module (2) constructs an environment view based on the collected environment image; A sub-environment view division module (3), which divides the environment view into a plurality of sub-environment views based on the edge line of the environment view; A sub-environment view attribute definition module (4), which defines a plurality of sub-environment view attributes based on robot collision factors; A target location positioning module (5) matches the target location environment image based on the environment view to locate the target location; A robot position positioning module (6) locates the robot's own position based on the matching of the environment view with the environment image; A navigation path planning module (7) plans the navigation path of the robot based on the robot's own position, the target position and the attributes of multiple sub-environment views in the environment view; A navigation control module (8) controls the movement of the robot based on the planned navigation path.

Images