Automobile windshield transferring robot and method based on artificial intelligence

Abstract

The invention relates to the technical field of program control robots, in particular to an automobile windshield transferring robot and method based on artificial intelligence, and a system comprises a surface characteristic analysis module, a dynamic environment sensing module, a grabbing area screening module, a path and action generation module and an action execution and correction module. According to the method, the boundary coordinates, the curvature distribution value and the light intensity deviation value of the glass surface are accurately extracted, so that the characteristic identification problem of a transparent object under complex optical interference is solved, the accuracy of surface data extraction is improved, obstacle motion data are collected in real time in a dynamic environment, the interference area and intensity are rapidly identified, and the accuracy of surface data extraction is improved. Grabbing points select comprehensive contact angles and stress uniformity parameters, stability and reliability of grabbing actions are enhanced, path planning improves task execution efficiency and motion coordination in a dynamic environment through low-interference path optimization and time adjustment, action execution is based on pose correction and force feedback adjustment, and the stability and reliability of the grabbing actions are improved. And the operation precision and the adaptive capacity are obviously improved.

Description

Technical Field

[0001] The invention relates to the technical field of program-controlled robots, and in particular to an automobile windshield transfer robot and method based on artificial intelligence. Background Art

[0002] The field of program-controlled robotics technology includes robotic systems used to achieve automated operations through program control. The core content of this technology is to achieve high-precision operations of robots in industrial manufacturing, assembly, handling and other scenarios through programmed instructions, mainly involving aspects such as robot structure design, motion control, path planning and task execution. The overall scope of program-controlled robotics technology covers multi-axis motion control technology, the design of end effectors and their functional expansion, as well as the development and optimization of control programs, enabling robots to complete specific tasks according to preset programs or real-time input data. In the field of automobile manufacturing, program-controlled robots are widely used in operations such as welding, painting, handling and assembly, and are characterized by high repeatability and high precision.

[0003] Among them, the AI-based automobile windshield transfer robot refers to the use of AI technology to optimize the robot's task execution process in the handling and assembly of automobile windshields. The technical matters covered by this patent subject include robot motion control and path optimization, precise grasping and placement of specific positions of windshields, real-time visual recognition systems for glass positioning and feature recognition, and real-time adjustment algorithms based on neural networks. Specifically, the sensor data is analyzed through artificial intelligence algorithms to determine the spatial position of the glass, and the robot motion trajectory is generated in combination with the path planning algorithm. In addition, a special fixture design is used to ensure the stability and safety of the windshield during the transfer process, and the combination of robots and artificial intelligence technology is comprehensively realized to complete this specific operation task.

[0004] The existing technology for identifying the characteristics of transparent objects is greatly affected by optical interference, and it is difficult to accurately extract surface features under the interference of reflection and refraction, resulting in deviations in characteristic data. In dynamic environment perception, it relies on a single parameter to analyze the movement trend of obstacles, which makes it difficult to accurately predict dynamic changes in complex environments, increasing the risk of path planning failure or obstacle avoidance delays. The selection of grasping points lacks a comprehensive evaluation of contact angles and force uniformity, which can easily lead to problems such as uneven distribution of grasping points or unstable grasping movements. The path planning process is not able to respond to real-time changes in the environment, and the path rhythm cannot be flexibly adjusted, resulting in reduced efficiency of the robot in a dynamic environment. The processing capabilities of path deviation correction and real-time force feedback during action execution are limited, which can easily lead to problems such as insufficient action accuracy or target grasping failure, restricting the operation effect in complex environments. Summary of the invention

[0005] The purpose of the present invention is to solve the shortcomings of the prior art and to propose an automobile windshield transfer robot and method based on artificial intelligence.

[0006] In order to achieve the above object, the present invention adopts the following technical solution: The automobile windshield transfer robot based on artificial intelligence comprises: The surface characteristic analysis module extracts the coordinates of the surface boundary contour points, the curvature distribution value and the light intensity offset value of the transparent area based on the windshield sampling data, superimposes and calculates the light intensity offset value and the curvature distribution value, analyzes the curvature change trend, identifies the curvature peak and the coordinates of the stable area, and generates the glass surface characteristic distribution data; The dynamic environment perception module analyzes the speed value and direction change rate based on the dynamically collected obstacle position coordinates, speed value and movement direction change rate, calculates the obstacle movement area, cross-analyzes the spatial position of the area and the glass surface feature distribution data, and generates the environmental interference distribution value; The grabbing area screening module extracts the boundary curvature peak point and the stable area coordinates based on the glass surface feature distribution data and the environmental interference distribution value, screens the clamping coordinate points, and generates the optimal grabbing point coordinate data; The path and action generation module extracts the spatial position sequence and time sequence of the path nodes based on the coordinate data of the preferred grasping points, accumulates the distances of the node paths of the spatial position sequence, compares them point by point with the environmental interference distribution value, adjusts the time differences between the path nodes, and generates optimized grasping path sequence data; The action execution and correction module extracts the path node posture parameters and the contact force feedback value of the fixture based on the optimized grasping path sequence data, calculates and generates the posture correction value, analyzes the error between the fixture contact force feedback value and the grasping point force trend value, identifies the contact force adjustment amount, and outputs the execution action calibration parameters.

[0007] As a further solution of the present invention, the glass surface feature distribution data includes boundary contour features, curvature change characteristics, light intensity offset characteristics, and optical interference characteristics; the environmental interference distribution values ​​include interference area coordinates, interference intensity data, motion range distribution, and overlapping range distribution; the preferred grasping point coordinate data include grasping point coordinates, force uniformity characteristics, contact angle distribution, and interference rejection results; the optimized grasping path sequence data include path node position sequence, time distribution sequence, low interference path sequence, and time interval adjustment value; the execution action calibration parameters include posture correction parameters, force adjustment parameters, path correction parameters, and grasping point force adjustment parameters.

[0008] As a further solution of the present invention, the surface characteristic analysis module includes a data extraction submodule, an optical interference calculation submodule, and a characteristic distribution generation submodule; The data extraction submodule extracts the coordinates of the surface boundary contour points, the light intensity offset value and the curvature distribution value of the transparent area based on the windshield sampling data, classifies and screens according to the characteristics of the transparent area, identifies the distribution law of the boundary data and the transparent area, and generates a surface feature data set; The optical interference calculation submodule is based on the surface feature data set, superimposes the transparent area light intensity offset value and the curvature distribution value, calculates the regional optical interference intensity, classifies the calculation results into segments to form a characteristic distribution range, analyzes the relationship between the optical interference and the surface geometric characteristics, marks the difference distribution of the regional interference intensity, and generates optical interference characteristic data; The characteristic distribution generation submodule analyzes the changing trend of the curvature distribution in segments based on the optical interference characteristic data, extracts characteristic key points by summarizing the curvature changing trend, maps the key points point by point to the glass surface coordinate system, marks the coordinate positions of the curvature peak and the stable area, and generates glass surface characteristic distribution data.

[0009] As a further solution of the present invention, the dynamic environment perception module includes a motion parameter extraction submodule, a dynamic prediction range generation submodule, and a spatial interference analysis submodule; The motion parameter extraction submodule extracts the real-time position and motion state data of the obstacle based on the dynamically collected obstacle position coordinates, speed value and motion direction change rate, calculates the correlation between the speed value and the direction change rate, extracts the position coordinates and motion state data, and generates the obstacle motion parameter data; The dynamic prediction range generation submodule determines the movement trend of the obstacle based on the obstacle motion parameter data, combined with the speed value and the direction change rate, calculates the impact value of the trend change on the spatial range, gradually predicts the obstacle movement area, spatially annotates the predicted area to generate the trend change range, and generates the obstacle dynamic prediction area data; The spatial interference analysis submodule compares the spatial distribution of the predicted area and the surface features point by point based on the obstacle dynamic prediction area data and the glass surface feature distribution data, calculates the interference intensity value in the intersection range, marks the spatial distribution and intensity data of the interference area, and generates the environmental interference distribution value.

[0010] As a further solution of the present invention, the grabbing area screening module includes a feature point extraction submodule, an interference elimination submodule, and a preferred point generation submodule; The feature point extraction submodule extracts the boundary curvature peak point and the coordinates of the stable area based on the glass surface feature distribution data, analyzes the curvature change value and coordinate distribution characteristics of the boundary area point by point, selects the points with significant curvature change as the curvature peak point, and identifies the area range where the curvature change tends to be stable, and generates boundary key feature point data; The interference elimination submodule compares the spatial coordinates of each feature point with the interference area range point by point based on the boundary key feature point data and the environmental interference distribution value, calculates the interference intensity value of the area where the feature point is located, eliminates the feature points above the interference threshold, and re-marks the coordinates of the remaining feature points and the interference area range to which they belong, to generate low-interference feature point data; The preferred point generation submodule calculates the clamping contact angle and force uniformity of the feature points based on the low-interference feature point data, sorts them according to the numerical values ​​of the contact angle and force uniformity parameters, selects the feature points that meet the clamping conditions, and generates preferred grasping point coordinate data.

[0011] As a further solution of the present invention, the interference intensity value calculation formula is specifically: ; in, Representative feature points The interference intensity value, Representative and characteristic points The number of associated interference regions, Representative feature points In the interference area The interference intensity value within Represents the interference area The area of Representative feature points Interference area The distance between the center points.

[0012] As a further solution of the present invention, the path and action generation module includes a path node extraction submodule, a path optimization submodule, and a time adjustment submodule; The path node extraction submodule extracts the spatial position sequence and time sequence of the path nodes based on the coordinate data of the preferred grab points, analyzes the distribution relationship between the nodes, calculates the connection attributes between adjacent nodes, generates the path framework point by point according to the spatial connectivity of the path nodes, accumulates the distances between the nodes to form the initial information of the path nodes, and generates the initial node sequence of the path; The path optimization submodule compares the spatial coordinates of the path nodes with the intersection of the interference area point by point based on the path initial node sequence and the environmental interference distribution value, calculates the interference intensity value of the node in the interference area, removes the nodes whose interference intensity exceeds the threshold, reconnects the nodes, establishes the interference optimization path, marks the adjusted path nodes, and generates low interference path data; Based on the low-interference path data, the time adjustment submodule extracts the time series of path nodes, analyzes the time intervals between adjacent nodes in the time series, adjusts the intervals and optimizes the movement rhythm between nodes, corrects the time distribution, verifies the consistency of the time series and the spatial path, and generates optimized grasping path sequence data.

[0013] As a further solution of the present invention, the action execution and correction module includes a posture parameter correction submodule, a contact force adjustment submodule, and an action calibration generation submodule; The posture parameter correction submodule extracts the posture parameters of the path nodes based on the optimized grasping path sequence data, compares the posture of the path nodes with the expected posture of the grasping points point by point, calculates the offset of the path nodes in multiple directions, analyzes the directionality and cumulative offset of the offset, and generates posture correction data; The contact force adjustment submodule compares the feedback value with the force trend value of the grasping point point by point based on the posture correction data and the contact force feedback value of the fixture, calculates the contact force error value, identifies its force direction, analyzes the spatial relationship between the change trend of the feedback force and the force of the grasping point, extracts the adjustment range of the force deviation, and marks the correction amplitude to generate the contact force adjustment data; Based on the contact force adjustment data, the motion calibration generation submodule superimposes the contact force adjustment value and the posture correction value point by point, calculates the adjusted motion state parameters, analyzes the state differences between the adjustment data and the path nodes item by item, and converts the differences into motion amplitude correction values, calibrates the motion outputs of the path nodes, and generates execution motion calibration parameters.

[0014] As a further solution of the present invention, the interference cumulative value calculation formula is specifically: ; in, Represents the cumulative interference value of the path segment, Represents the number of interference regions contained in the path segment, Represents the node in the path segment in the interference area The interference intensity value, Represents the interference area The area of Represents the distance from the node in the path segment to the center point of the interference area.

[0015] The method for transferring automobile windshield by robot based on artificial intelligence comprises the following steps: S1: Based on the windshield sampling data, the coordinates of the surface boundary contour points, the curvature distribution value and the light intensity offset value of the transparent area are extracted, the curvature change trend is analyzed, the curvature peak and the coordinates of the stable area are identified, and the glass surface feature distribution data is generated; S2: Based on the dynamically collected obstacle position coordinates, speed value and movement direction change rate, the obstacle movement area is calculated, the spatial position of the cross-analysis area and the glass surface feature distribution data is analyzed, and the environmental interference distribution value is generated; S3: based on the glass surface feature distribution data and the environmental interference distribution value, extracting the boundary curvature peak point and the stable area coordinates, and generating the optimal grabbing point coordinate data; S4: Based on the coordinate data of the preferred grasping point, extract the spatial position sequence and time sequence of the path nodes, accumulate the distances of the node paths of the spatial position sequence, and compare them point by point with the environmental interference distribution value to generate optimized grasping path sequence data; S5: Based on the optimized grasping path sequence data, the path node posture parameters and the contact force feedback value of the fixture are extracted, the error between the fixture contact force feedback value and the grasping point force trend value is analyzed, the contact force adjustment amount is identified, and the execution action calibration parameters are output.

[0016] Compared with the prior art, the advantages and positive effects of the present invention are: In the present invention, by accurately extracting the boundary coordinates, curvature distribution value and light intensity offset value of the glass surface, the problem of characteristic identification of transparent objects under complex optical interference is solved, the accuracy of surface data extraction is improved, and obstacle motion data is collected in real time in a dynamic environment to achieve rapid identification of interference areas and intensities. The grasping point selection comprehensively considers contact angles and force uniformity parameters to enhance the stability and reliability of the grasping action. Path planning improves task execution efficiency and motion coordination in a dynamic environment through low-interference path optimization and time adjustment. Action execution is based on posture correction and force feedback adjustment, which significantly improves operation accuracy and adaptability. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0018] Figure 1 is a system flow chart of the present invention; Figure 2 It is a submodule flow chart of the present invention; Figure 3 It is a flow chart of the surface characteristic analysis module of the present invention; Figure 4 This is a flow chart of the dynamic environment perception module of the present invention; Figure 5 This is a flow chart of the grab area screening module of the present invention; Figure 6 A flow chart of the path and action generation module of the present invention; Figure 7 It is a flow chart of the action execution and correction module of the present invention; Figure 8 The figure is a flow chart of the steps of the method of the present invention. DETAILED DESCRIPTION

[0019] The technical solution of the present invention is described below in conjunction with the accompanying drawings.

[0020] In the embodiments of the present invention, words such as "exemplarily" and "for example" are used to indicate examples, illustrations or explanations. Any embodiment or design described as "example" in the present invention should not be interpreted as being more preferred or more advantageous than other embodiments or designs. Specifically, the use of the word "example" is intended to present the concept in a specific way. In addition, in the embodiments of the present invention, the meaning expressed by "and / or" can be both, or it can be either of the two.

[0021] In the embodiments of the present invention, "image" and "picture" can sometimes be used interchangeably. It should be noted that when the difference between them is not emphasized, the meanings they intend to express are the same. "of", "corresponding, relevant" and "corresponding" can sometimes be used interchangeably. It should be noted that when the difference between them is not emphasized, the meanings they intend to express are the same.

[0022] In the embodiments of the present invention, sometimes a subscript such as W1 may be written as a non-subscript such as W1. When the difference is not emphasized, the meanings to be expressed are the same.

[0023] In order to make the technical problems, technical solutions and advantages to be solved by the present invention more clear, a detailed description will be given below with reference to the accompanying drawings and specific embodiments.

[0024] See also Figure 1 and Figure 2 , the AI-based car windshield transfer robot includes: The surface characteristic analysis module extracts the coordinates of the surface boundary contour points, the curvature distribution value and the light intensity offset value of the transparent area based on the windshield sampling data, superimposes and calculates the light intensity offset value and the curvature distribution value, analyzes the curvature change trend, identifies the curvature peak and the coordinates of the stable area, and generates the glass surface characteristic distribution data; The dynamic environment perception module analyzes the speed value and direction change rate based on the dynamically collected obstacle position coordinates, speed value and movement direction change rate, calculates the obstacle movement area, cross-analyzes the spatial position of the area and the glass surface feature distribution data, and generates the environmental interference distribution value; The grabbing area screening module extracts the boundary curvature peak point and the stable area coordinates based on the glass surface feature distribution data and the environmental interference distribution value, screens the clamping coordinate points, and generates the optimal grabbing point coordinate data; The path and action generation module extracts the spatial position sequence and time sequence of the path nodes based on the coordinate data of the preferred grasping points, accumulates the distances of the node paths in the spatial position sequence, compares them point by point with the environmental interference distribution value, adjusts the time differences between the path nodes, and generates optimized grasping path sequence data; The action execution and correction module is based on the optimized grasping path sequence data, extracts the path node posture parameters and the contact force feedback value of the fixture, calculates and generates the posture correction value, analyzes the error between the fixture contact force feedback value and the grasping point force trend value, identifies the contact force adjustment amount, and outputs the execution action calibration parameters.

[0025] The glass surface feature distribution data includes boundary contour features, curvature change characteristics, light intensity offset characteristics, and optical interference characteristics. The environmental interference distribution values ​​include interference area coordinates, interference intensity data, motion range distribution, and overlapping range distribution. The preferred grasping point coordinate data include grasping point coordinates, force uniformity characteristics, contact angle distribution, and interference rejection results. The optimized grasping path sequence data includes path node position sequence, time distribution sequence, low interference path sequence, and time interval adjustment value. The execution action calibration parameters include posture correction parameters, force adjustment parameters, path correction parameters, and grasping point force adjustment parameters.

[0026] See also Figure 2 and Figure 3 ,The surface characteristic analysis module includes a data extraction submodule, an optical interference calculation submodule, and a characteristic distribution generation submodule; The data extraction submodule extracts the coordinates of the surface boundary contour points, the light intensity offset value and the curvature distribution value of the transparent area based on the windshield sampling data, classifies and screens according to the characteristics of the transparent area, identifies the distribution law of the boundary data and the transparent area, and generates a surface feature data set; Firstly, the coordinates of the surface boundary contour points are sampled point by point through a high-precision boundary recognition device, and the sampling error value is eliminated using a data denoising algorithm. At the same time, the variation range of the light intensity offset value in the transparent area is recorded, and the curvature distribution value of the transparent area is calculated in real time in combination with the curvature analysis device. The classification and screening rules are constructed using the curvature change curve. The classification process is based on the comparison of the transparent characteristics of adjacent areas to determine whether the area belongs to the transparent boundary range, and then the distribution law of the boundary data and the transparent area is identified, and the surface feature data set is generated based on the recognition results of the transparent area.

[0027] The optical interference calculation submodule is based on the surface feature data set, superimposes the light intensity offset value and curvature distribution value of the transparent area, calculates the regional optical interference intensity, classifies the calculation results into segments to form a characteristic distribution range, analyzes the relationship between the optical interference and the surface geometric characteristics, marks the differential distribution of the regional interference intensity, and generates optical interference characteristic data; Superimpose the transparent area light intensity offset value and curvature distribution value, according to the formula ; In the formula, represents the optical interference intensity, represents the regional curvature distribution value, Represents the light intensity distribution value, Represents the partial derivative of the light intensity distribution value with respect to the spatial position, and is the integration interval.

[0028] In the calculation of the optical interference intensity of the region, the curvature distribution value Measured by high-precision curvature analysis equipment, the light intensity distribution value The brightness values ​​and partial derivatives of different spatial positions are obtained through sampling devices. It is calculated by dividing the difference between the sampling points by the spatial position difference. Assuming that in an area, the curvature value is measured is 0.05, light intensity distribution Satisfies linear distribution in the interval [0,10] , the calculation process is as follows: , which means that the partial derivative of light intensity with respect to position is a constant.

[0029] Substitute into the formula to calculate: ; The results show that the calculated optical interference intensity is 0.1, which represents the interference effect value under the combined action of the transparency and geometric characteristics of the region, and its value reflects the differences in optical characteristics of the region.

[0030] The characteristic distribution generation submodule analyzes the changing trend of curvature distribution in sections based on the optical interference characteristic data, extracts characteristic key points by summarizing the curvature changing trend, maps the key points point by point to the glass surface coordinate system, marks the coordinate positions of the curvature peak and the stable area, and generates the glass surface characteristic distribution data; Through the segmented curvature analysis of the transparent area, the curvature change rate calculation model is used to extract the key points of curvature change. First, the curvature data in the transparent area is collected and the change points are recorded. According to the curvature change trend, the segmented analysis is performed, and the points with a change rate greater than a certain threshold are extracted as feature key points. The extracted feature key points are transformed and mapped to the glass surface coordinate system. The distribution of curvature peak points and stable points is analyzed, and the feature distribution results of the glass surface are generated by combining the mapped feature distribution data.

[0031] See also Figure 2 and Figure 4 ,The dynamic environment perception module includes a motion parameter extraction submodule, a dynamic prediction range generation submodule, and a spatial interference analysis submodule; The motion parameter extraction submodule extracts the real-time position and motion state data of the obstacle based on the dynamically collected obstacle position coordinates, speed value and motion direction change rate, calculates the correlation between the speed value and the direction change rate, extracts the position coordinates and motion state data, and generates the obstacle motion parameter data; The speed value is calculated through continuous position sampling and the rate of change of the motion direction is recorded. Combined with the position coordinate data, the motion trajectory is obtained by calculating the Euclidean distance between continuous position points. The real-time position data and motion state data are extracted through trajectory segmentation analysis. The correlation between the speed value and the rate of change of direction is compared to calculate the correlation between the two. The correlation coefficient method is used to calculate the correlation degree. The motion parameters of the obstacle are further extracted and recorded in the data set, and finally the obstacle motion parameter data is generated.

[0032] The dynamic prediction range generation submodule determines the movement trend of the obstacle based on the obstacle motion parameter data, combined with the speed value and the direction change rate, calculates the impact value of the trend change on the spatial range, gradually predicts the obstacle movement area, spatially annotates the predicted area to generate the trend change range, and generates the obstacle dynamic prediction area data; Combine the speed value and the direction change rate to determine the movement trend of the obstacle, according to the formula ; Calculate the impact of trend changes on spatial extent.

[0033] In the formula, Represents the impact of trend changes on the spatial range, represents the speed value of the obstacle, represents the time derivative of the rate of change of direction, and is the time integration interval.

[0034] Calculation of the impact of motion trend changes on spatial range, speed value The direction change rate is collected by real-time obstacle monitoring equipment. The derivative is obtained by calculating the change in the angle of the obstacle's movement direction within the time interval. Obtained from the time difference of the rate of change of direction. Assuming the velocity value 3m / s, direction change rate Satisfy linear changes in the interval [0,2] seconds , the calculation process is as follows: The time derivative of the rate of change of direction: .

[0035] Substitute into the formula to calculate: ; The results show that the impact value of trend change on spatial range is 3, which represents the degree of interference of the change of obstacle movement trend on the future prediction area, and its value can be used for the calculation of spatial range prediction.

[0036] The spatial interference analysis submodule compares the spatial distribution of the predicted area and the surface features point by point based on the obstacle dynamic prediction area data and the glass surface feature distribution data, calculates the interference intensity value in the intersection range, marks the spatial distribution and intensity data of the interference area, and generates the environmental interference distribution value; By combining the dynamically generated trend change range with the glass surface feature distribution data, point-by-point spatial coordinate matching technology is used for comparison, and the spatial intersection range is calculated for each point in the obstacle prediction area and the glass surface feature data point. The spatial intersection range is calculated by judging the overlapping area through the distance threshold, and the interference intensity value within the intersection range is calculated. The total interference intensity is calculated by gradually accumulating the intensity values ​​in the interference area, and the interference area coordinates in the spatial distribution map are marked. The environmental interference distribution value is generated by combining the interference intensity value and the distribution coordinates.

[0037] See also Figure 2 and Figure 5 ,The grabbing area screening module includes a feature point extraction submodule, an interference elimination submodule, and a preferred point generation submodule; The feature point extraction submodule extracts the boundary curvature peak point and the coordinates of the stable area based on the glass surface feature distribution data, analyzes the curvature change value and coordinate distribution characteristics of the boundary area point by point, selects the points with significant curvature change as the curvature peak point, and identifies the area range where the curvature change tends to be stable, and generates the boundary key feature point data; The curvature distribution value of the glass surface is collected by a high-precision sensor, and the curvature change value is calculated using the differential calculation method. The points with significant curvature changes are selected as the curvature peak points. The sliding window technology is used to analyze the continuity characteristics of the curvature change, and the area where the curvature change tends to be stable is extracted. The coordinates of the feature points are calibrated through the correspondence between the boundary curvature value and the position coordinates, and the above extracted contents are formed into a boundary key feature point data set.

[0038] The interference removal submodule compares the spatial coordinates of each feature point with the interference area based on the boundary key feature point data and the environmental interference distribution value, calculates the interference intensity value of the area where the feature point is located, removes the feature points that are higher than the interference threshold, and re-marks the coordinates of the remaining feature points and the interference area to which they belong, generating low-interference feature point data; The calculation formula of interference intensity value is as follows: ; in, Representative feature points The interference intensity value, Representative and characteristic points The number of associated interference regions, Representative feature points In the interference area The interference intensity value within Represents the interference area The area of Representative feature points Interference area The distance between the center points.

[0039] Feature point interference correction coefficient The calculation of involves the following steps and parameters: 1. Representing feature points In the interference area The interference intensity value within is calculated by measuring the optical or mechanical influence value of the interference area. The measurement method uses an interference intensity sensor to integrate the intersection range of the feature point and the area to obtain the value; 2. Indicates the interference area The area is calculated using the coordinate data of the boundary points of the interference area through the geometric area formula, and the area unit is square meters; 3. Representing feature points To the interference area The distance of the center point is calculated using the Euclidean distance formula between two points; 4. Representation and feature points The number of relevant interference regions is determined by the intersection of the feature points and the interference regions; 5. The summation operation is used to accumulate the effects of all interference areas on the feature points, the absolute value is taken to ensure that all effects are positive, and the square root process is used to smooth the results.

[0040] Calculation example: Feature Points Located at coordinates (2,3), there are three interference regions, with center coordinates (4,4), (6,7), and (3,5), and their areas 5 square meters, 8 square meters and 6 square meters. Interference intensity value The values ​​are 0.8, 1.2 and 0.5. The calculations are: Calculate the distance between each interference area and the feature point: ; ; ; Calculate the impact value of each interference area on the feature point: ; ; ; Sum and calculate the interference correction coefficient: ; The results show that the interference intensity value of the feature point It is 1.268, which represents the average interference correction value of the feature point in all interference areas. The value is used to further eliminate high interference points and optimize the distribution of interference areas.

[0041] The optimal point generation submodule calculates the clamping contact angle and force uniformity of the feature points based on the low-interference feature point data, sorts them according to the numerical values ​​of the contact angle and force uniformity parameters, selects the feature points that meet the clamping conditions, and generates the optimal grasping point coordinate data; The clamping contact angle and force distribution of the feature points are measured by high-precision force sensors. The clamping contact angle is calculated using the angle distribution mean method. The force uniformity is calculated using the standard deviation of the force distribution of the feature points. A sorted list of feature points is generated based on the calculation results, and feature points with clamping angles within the target range and good force uniformity are screened. The coordinates of the screened feature points are output as the coordinate data of the preferred grasping points.

[0042] See also Figure 2 and Figure 6 ,The path and action generation module includes a path node extraction submodule, a path optimization submodule, and a time adjustment submodule; The path node extraction submodule extracts the spatial position sequence and time sequence of the path nodes based on the coordinate data of the preferred grab points, analyzes the distribution relationship between nodes, calculates the connection properties between adjacent nodes, generates the path framework point by point according to the spatial connectivity of the path nodes, accumulates the distances between nodes to form the initial information of the path nodes, and generates the initial node sequence of the path; The spatial position sequence and time series of path nodes are extracted point by point, and the Euclidean distance between adjacent nodes is calculated using spatial analysis tools. The total length of the path is calculated by accumulating the distances, and the spatial connectivity of adjacent nodes is further determined. The path framework is generated point by point using geometric topological structure analysis. At the same time, the time series of path nodes are synchronously analyzed, and the node time intervals and sequence distribution are recorded. The initial information of path nodes is generated through a comprehensive analysis of the position and time between nodes, and finally the initial node sequence of the path is output.

[0043] The path optimization submodule compares the spatial coordinates of the path nodes with the intersection of the interference area point by point based on the path initial node sequence and the environmental interference distribution value, calculates the interference intensity value of the node in the interference area, removes the nodes whose interference intensity exceeds the threshold, reconnects the nodes, establishes the interference optimization path, marks the adjusted path nodes, and generates low-interference path data; The calculation formula of interference cumulative value is as follows: ; in, Represents the cumulative interference value of the path segment, Represents the number of interference regions contained in the path segment, Represents the node in the path segment in the interference area The interference intensity value, Represents the interference area The area of Represents the distance from the node in the path segment to the center point of the interference area.

[0044] Cumulative interference value of path segments It represents the cumulative interference effect of all interference areas in the path segment on the path node. The calculation involves the following parameters: Indicates that the nodes in the path segment are in the interference area The interference intensity value is measured by an interference intensity sensor and is expressed in Newton or the corresponding interference intensity unit.

[0045] Indicates the interference area The area is calculated by measuring the boundary coordinate points of the interference area and using the geometric area formula, in square meters.

[0046] Indicates the node to interference area in the path segment The distance of the center point is calculated using the Euclidean distance formula and is in meters.

[0047] Represents the total number of interference regions within a path segment, determined by the spatial intersection data of the path segment and the interference regions.

[0048] The summation operation accumulates the interference contribution value of each interference area, takes the absolute value to ensure the numerical positivity, and the inverse square is used to weight the effect of distance on interference.

[0049] Calculation example: The path segment contains three interference areas, and the corresponding parameters are as follows: Interference area 1: , , the distance from the path node to the center point .

[0050] Interference area 2: , , the distance from the path node to the center point .

[0051] Interference area 3: , , the distance from the path node to the center point .

[0052] Calculate the contribution of each interference region: Contribution of interference region 1: ; Contribution of interference region 2: ; Contribution of interference region 3: ; Cumulative contributions from all interfering regions: ; This result shows that the interference accumulation value of the path segment The value is 2.65, which indicates the total interference strength contribution value between the nodes in the path segment and the interference area. This value is used to determine whether the path node exceeds the interference threshold and perform path optimization.

[0053] The time adjustment submodule extracts the time series of path nodes based on low-interference path data, analyzes the time intervals between adjacent nodes in the time series, adjusts the intervals and optimizes the movement rhythm between nodes, corrects the time distribution, verifies the consistency between the time series and the spatial path, and generates optimized grasping path sequence data; The time interval analysis model is used to optimize the node time interval. The time interval is adjusted by analyzing the time distribution of adjacent nodes and optimizing the movement rhythm. The time interval distribution is corrected to be consistent with the spatial path. The adjustment effect of the time series is verified point by point, and the optimized crawling path sequence data is regenerated in combination with the spatial path distribution, and the final correspondence between time and space nodes is marked.

[0054] See also Figure 2 and Figure 7 ,The action execution and correction module includes a posture parameter correction submodule, a contact force adjustment submodule, and an action calibration generation submodule; The posture parameter correction submodule extracts the posture parameters of the path nodes based on the optimized grasping path sequence data, compares the posture of the path nodes with the expected posture of the grasping points point by point, calculates the offset of the path nodes in multiple directions, analyzes the directionality and cumulative offset of the offset, and generates posture correction data; High-precision posture sensing equipment is used to collect the posture information of path nodes point by point, and the angle and position offset of the nodes in multiple directions are obtained. By comparing the posture parameters of the path nodes with the expected posture of the grasping points, the size and direction of the offset are calculated, and the offset trend and cumulative offset in multiple directions are analyzed. Combined with the spatial relationship between the nodes, the final posture correction data is generated.

[0055] The contact force adjustment submodule compares the feedback value with the force trend value of the grasping point point by point based on the posture correction data and the contact force feedback value of the fixture, calculates the contact force error value, identifies its force direction, analyzes the spatial relationship between the change trend of the feedback force and the force of the grasping point, extracts the adjustment range of the force deviation, and marks the correction amplitude to generate the contact force adjustment data; Compare the feedback value and the force trend value of the grasping point point by point, according to the formula ; Calculate the contact force error value.

[0056] In the formula, represents the adjustment value of the contact force, represents the contact force distribution of feedback, Indicates the force trend value of the grasping point, is the integral range of the force-bearing area.

[0057] The contact force adjustment value is obtained by calculating the difference between the feedback contact force and the expected contact force point by point and accumulating them. The feedback contact force is recorded by the real-time contact force sensor, and the expected contact force is given by the force distribution model of the gripping point. Assume that the feedback contact force function is , the expected contact force function is , the integral range is [0,4]×[0,4]. The calculation process is as follows: Difference function: ; Adjustment value calculation: ; Decomposition of the integral: ; Item-by-item integration: ; The results show that the contact force adjustment value is 13.6, which represents the overall error adjustment range between the feedback force and the desired force, and is used for the subsequent optimization of the contact force parameters.

[0058] Based on the contact force adjustment data, the motion calibration generation submodule superimposes the contact force adjustment value and the posture correction value point by point, calculates the adjusted motion state parameters, analyzes the state difference between the adjustment data and the path node item by item, and converts the difference into the motion amplitude correction value, calibrates the motion output of the path node, and generates the execution motion calibration parameters; The corrected action state parameters are calculated by point-by-point superposition, and the difference between the adjustment data and the current state of the path node is further analyzed. The correction value of the action amplitude is calculated through difference analysis. The action amplitude correction value is obtained by the difference between the adjustment data and the historical state recorded in real time by the sensor. The corrected amplitude value re-marks the action output parameters in units of path nodes, and finally generates the calibrated execution action calibration parameter data.

[0059] See also Figure 8 , an artificial intelligence-based automobile windshield transfer robot method, comprising the following steps: S1: Based on the windshield sampling data, the coordinates of the surface boundary contour points, the curvature distribution value and the light intensity offset value of the transparent area are extracted, the curvature change trend is analyzed, the curvature peak and the coordinates of the stable area are identified, and the glass surface feature distribution data is generated; S2: Based on the dynamically collected obstacle position coordinates, speed value and movement direction change rate, the obstacle movement area is calculated, the spatial position of the cross-analysis area and the glass surface feature distribution data is analyzed, and the environmental interference distribution value is generated; S3: Based on the glass surface feature distribution data and the environmental interference distribution value, the boundary curvature peak point and the stable area coordinates are extracted to generate the optimal grabbing point coordinate data; S4: Based on the coordinate data of the preferred grasping points, the spatial position sequence and time series of the path nodes are extracted, the distances of the node paths of the spatial position sequence are accumulated, and compared point by point with the environmental interference distribution value to generate the optimized grasping path sequence data; S5: Based on the optimized grasping path sequence data, the path node posture parameters and the contact force feedback value of the fixture are extracted, the error between the fixture contact force feedback value and the grasping point force trend value is analyzed, the contact force adjustment amount is identified, and the execution action calibration parameters are output.

[0060] The above is only a specific embodiment of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art can easily think of changes or substitutions within the technical scope disclosed by the present invention, which should be included in the protection scope of the present invention. Therefore, the protection scope of the present invention should be based on the protection scope of the claims.

Claims

1. An artificial intelligence-based automobile windshield transfer robot, characterized by: The system comprises: The surface characteristic analysis module extracts the coordinates of the surface boundary contour points, the curvature distribution value and the light intensity offset value of the transparent area based on the windshield sampling data, superimposes and calculates the light intensity offset value and the curvature distribution value, analyzes the curvature change trend, identifies the curvature peak and the coordinates of the stable area, and generates the glass surface characteristic distribution data; The dynamic environment perception module analyzes the speed value and direction change rate based on the dynamically collected obstacle position coordinates, speed value and movement direction change rate, calculates the obstacle movement area, cross-analyzes the spatial position of the area and the glass surface feature distribution data, and generates the environmental interference distribution value; The grabbing area screening module extracts the boundary curvature peak point and the stable area coordinates based on the glass surface feature distribution data and the environmental interference distribution value, screens the clamping coordinate points, and generates the optimal grabbing point coordinate data; The path and action generation module extracts the spatial position sequence and time sequence of the path nodes based on the coordinate data of the preferred grasping points, accumulates the distances of the node paths of the spatial position sequence, compares them point by point with the environmental interference distribution value, adjusts the time differences between the path nodes, and generates optimized grasping path sequence data; The action execution and correction module extracts the path node posture parameters and the contact force feedback value of the fixture based on the optimized grasping path sequence data, calculates and generates the posture correction value, analyzes the error between the fixture contact force feedback value and the grasping point force trend value, identifies the contact force adjustment amount, and outputs the execution action calibration parameters.

2. The artificial intelligence-based automobile windshield transfer robot according to claim 1, characterized in that: The glass surface feature distribution data includes boundary contour features, curvature change characteristics, light intensity offset characteristics, and optical interference characteristics. The environmental interference distribution values ​​include interference area coordinates, interference intensity data, motion range distribution, and overlapping range distribution. The preferred grasping point coordinate data include grasping point coordinates, force uniformity characteristics, contact angle distribution, and interference rejection results. The optimized grasping path sequence data includes path node position sequence, time distribution sequence, low interference path sequence, and time interval adjustment value. The execution action calibration parameters include posture correction parameters, force adjustment parameters, path correction parameters, and grasping point force adjustment parameters.

3. The automobile windshield transfer robot based on artificial intelligence according to claim 1 is characterized in that: The surface characteristic analysis module includes a data extraction submodule, an optical interference calculation submodule, and a characteristic distribution generation submodule; The data extraction submodule extracts the coordinates of the surface boundary contour points, the light intensity offset value and the curvature distribution value of the transparent area based on the windshield sampling data, classifies and screens according to the characteristics of the transparent area, identifies the distribution law of the boundary data and the transparent area, and generates a surface feature data set; The optical interference calculation submodule is based on the surface feature data set, superimposes the transparent area light intensity offset value and the curvature distribution value, calculates the regional optical interference intensity, classifies the calculation results into segments to form a characteristic distribution range, analyzes the relationship between the optical interference and the surface geometric characteristics, marks the difference distribution of the regional interference intensity, and generates optical interference characteristic data; The characteristic distribution generation submodule analyzes the changing trend of the curvature distribution in segments based on the optical interference characteristic data, extracts characteristic key points by summarizing the curvature changing trend, maps the key points point by point to the glass surface coordinate system, marks the coordinate positions of the curvature peak and the stable area, and generates glass surface characteristic distribution data.

4. The automobile windshield transfer robot based on artificial intelligence according to claim 1 is characterized in that: The dynamic environment perception module includes a motion parameter extraction submodule, a dynamic prediction range generation submodule, and a spatial interference analysis submodule; The motion parameter extraction submodule extracts the real-time position and motion state data of the obstacle based on the dynamically collected obstacle position coordinates, speed value and motion direction change rate, calculates the correlation between the speed value and the direction change rate, extracts the position coordinates and motion state data, and generates the obstacle motion parameter data; The dynamic prediction range generation submodule determines the movement trend of the obstacle based on the obstacle motion parameter data, combined with the speed value and the direction change rate, calculates the impact value of the trend change on the spatial range, gradually predicts the obstacle movement area, spatially annotates the predicted area to generate the trend change range, and generates the obstacle dynamic prediction area data; The spatial interference analysis submodule compares the spatial distribution of the predicted area and the surface features point by point based on the obstacle dynamic prediction area data and the glass surface feature distribution data, calculates the interference intensity value in the intersection range, marks the spatial distribution and intensity data of the interference area, and generates the environmental interference distribution value.

5. The automobile windshield transfer robot based on artificial intelligence according to claim 1 is characterized in that: The grabbing area screening module includes a feature point extraction submodule, an interference elimination submodule, and a preferred point generation submodule; The feature point extraction submodule extracts the boundary curvature peak point and the coordinates of the stable area based on the glass surface feature distribution data, analyzes the curvature change value and coordinate distribution characteristics of the boundary area point by point, selects the points with significant curvature change as the curvature peak point, and identifies the area range where the curvature change tends to be stable, and generates boundary key feature point data; The interference elimination submodule compares the spatial coordinates of each feature point with the interference area range point by point based on the boundary key feature point data and the environmental interference distribution value, calculates the interference intensity value of the area where the feature point is located, eliminates the feature points above the interference threshold, and re-marks the coordinates of the remaining feature points and the interference area range to which they belong, to generate low-interference feature point data; The preferred point generation submodule calculates the clamping contact angle and force uniformity of the feature points based on the low-interference feature point data, sorts them according to the numerical values ​​of the contact angle and force uniformity parameters, selects the feature points that meet the clamping conditions, and generates preferred grasping point coordinate data.

6. The automobile windshield transfer robot based on artificial intelligence according to claim 5 is characterized in that: The interference intensity value calculation formula is specifically: ； in, Representative feature points The interference intensity value, Representative and characteristic points The number of associated interference regions, Representative feature points In the interference area The interference intensity value within Represents the interference area The area of Representative feature points Interference area The distance between the center points.

7. The artificial intelligence-based automobile windshield transfer robot according to claim 1, characterized in that: The path and action generation module includes a path node extraction submodule, a path optimization submodule, and a time adjustment submodule; The path node extraction submodule extracts the spatial position sequence and time sequence of the path nodes based on the coordinate data of the preferred grab points, analyzes the distribution relationship between the nodes, calculates the connection attributes between adjacent nodes, generates the path framework point by point according to the spatial connectivity of the path nodes, accumulates the distances between the nodes to form the initial information of the path nodes, and generates the initial node sequence of the path; The path optimization submodule compares the spatial coordinates of the path nodes with the intersection of the interference area point by point based on the path initial node sequence and the environmental interference distribution value, calculates the interference intensity value of the node in the interference area, removes the nodes whose interference intensity exceeds the threshold, reconnects the nodes, establishes the interference optimization path, marks the adjusted path nodes, and generates low interference path data; Based on the low-interference path data, the time adjustment submodule extracts the time series of path nodes, analyzes the time intervals between adjacent nodes in the time series, adjusts the intervals and optimizes the movement rhythm between nodes, corrects the time distribution, verifies the consistency of the time series and the spatial path, and generates optimized grasping path sequence data.

8. The artificial intelligence-based automobile windshield transfer robot according to claim 7 is characterized in that: The interference cumulative value calculation formula is specifically: ； in, Represents the cumulative interference value of the path segment, Represents the number of interference regions contained in the path segment, Represents the node in the path segment in the interference area The interference intensity value, Represents the interference area The area of Represents the distance from the node in the path segment to the center point of the interference area.

9. The automobile windshield transfer robot based on artificial intelligence according to claim 1 is characterized in that: The action execution and correction module includes a posture parameter correction submodule, a contact force adjustment submodule, and an action calibration generation submodule; The posture parameter correction submodule extracts the posture parameters of the path nodes based on the optimized grasping path sequence data, compares the posture of the path nodes with the expected posture of the grasping points point by point, calculates the offset of the path nodes in multiple directions, analyzes the directionality and cumulative offset of the offset, and generates posture correction data; The contact force adjustment submodule compares the feedback value with the force trend value of the grasping point point by point based on the posture correction data and the contact force feedback value of the fixture, calculates the contact force error value, identifies its force direction, analyzes the spatial relationship between the change trend of the feedback force and the force of the grasping point, extracts the adjustment range of the force deviation, and marks the correction amplitude to generate the contact force adjustment data; Based on the contact force adjustment data, the motion calibration generation submodule superimposes the contact force adjustment value and the posture correction value point by point, calculates the adjusted motion state parameters, analyzes the state differences between the adjustment data and the path nodes item by item, and converts the differences into motion amplitude correction values, calibrates the motion outputs of the path nodes, and generates execution motion calibration parameters.

10. A method for transferring automobile windshields by a robot based on artificial intelligence, characterized in that: The automobile windshield transfer robot based on artificial intelligence according to any one of claims 1 to 9 comprises the following steps: S1: Based on the windshield sampling data, the coordinates of the surface boundary contour points, the curvature distribution value and the light intensity offset value of the transparent area are extracted, the curvature change trend is analyzed, the curvature peak and the coordinates of the stable area are identified, and the glass surface feature distribution data is generated; S2: Based on the dynamically collected obstacle position coordinates, speed value and movement direction change rate, the obstacle movement area is calculated, the spatial position of the cross-analysis area and the glass surface feature distribution data is analyzed, and the environmental interference distribution value is generated; S3: based on the glass surface feature distribution data and the environmental interference distribution value, extracting the boundary curvature peak point and the stable area coordinates, and generating the optimal grabbing point coordinate data; S4: Based on the coordinate data of the preferred grasping point, extract the spatial position sequence and time sequence of the path nodes, accumulate the distances of the node paths of the spatial position sequence, and compare them point by point with the environmental interference distribution value to generate optimized grasping path sequence data; S5: Based on the optimized grasping path sequence data, the path node posture parameters and the contact force feedback value of the fixture are extracted, the error between the fixture contact force feedback value and the grasping point force trend value is analyzed, the contact force adjustment amount is identified, and the execution action calibration parameters are output.

Images