A Motion Planning and Control Method and System for a Robot to Sort Commodities

Abstract

The present invention relates to a motion planning and control method and system for a robot to sort goods. The method includes: capturing an image of goods to be sorted; mapping the image of goods to be sorted to obtain initial features of each grid point in a three-dimensional grid; calculating a camera depth map based on the initial features of each grid point in the three-dimensional grid according to a feature correction method and a volume rendering method; correcting the camera depth map to obtain a corrected depth map; calculating the pose information of the goods to be sorted; and planning the motion trajectory of a robotic arm and the grasping posture of a dexterous hand based on an improved whale optimization algorithm according to the corrected depth map and the pose information. Based on the feature correction method and the volume rendering method, the present invention can accurately calculate the camera depth map and perform correction to control the dexterous hand to quickly and accurately grasp goods; at the same time, the method of the present invention is simple and greatly improves the grasping speed of the robot; in addition, the present invention adopts an improved whale optimization algorithm, which can ensure the accurate grasping of the target object.

Description

Technical Field

[0001] The present invention relates to the field of robot control, and particularly to a motion planning and control method and system for a robot to sort goods. Background Art

[0002] With the rapid development of technology, robot technology has penetrated into various industries. A robot sorting system usually consists of technologies such as a robot body, sensors, a control system, and computer vision. During the sorting process, the robot obtains goods information through sensors, uses computer vision technology to identify the types and positions of goods, and then drives the robot body through the control system for precise grasping and sorting. The whole process realizes high automation and intelligence, greatly improving the sorting efficiency and accuracy.

[0003] In the prior art, during the process of a robot grasping goods, Yolov8 and DenseFusion are generally used for motion planning and control. The main function of Yolov8 is to accurately identify goods, helping the robot accurately identify various different types of goods. The main function of DenseFusion is to perform 6D pose estimation of an object, helping the robot accurately locate and grasp the object.

[0004] However, the existing combination method of Yolov8 and DenseFusion for the motion planning and control of a robot matches the target information with an object of a known three-dimensional model after detecting the target pose, and obtains the grasping pose through the matching. However, the target objects in the actual scene are complex and changeable, and the camera carried by the robot usually can only obtain target information of a single view, which results in a low success rate of the matching and the trajectory planning algorithm is prone to falling into local optimization. Therefore, accurate and rapid grasping of the target object cannot be achieved. Summary of the Invention

[0005] The present invention provides a motion planning and control method and system for a robot to sort goods to solve at least one of the above technical problems.

[0006] The technical solution of the present invention to solve the above technical problems is as follows: A motion planning and control method for a robot to sort goods includes:

[0007] S1, using a monocular camera to photograph the goods to be sorted in the sorting area to obtain an image of the goods to be sorted;

[0008] S2, mapping the image of the goods to be sorted based on a three-dimensional grid to obtain initial features of each grid point in the three-dimensional grid;

[0009] S3. Based on the feature correction method and the volume rendering method, calculate the camera depth map according to the initial features of each grid point in the three-dimensional grid.

[0010] S4. Correct the camera depth map according to the current position information of the dexterous hand at the end of the robotic arm and the position information of the monocular camera to obtain a corrected depth map.

[0011] S5. Adopt an adaptive technique to improve the whale optimization algorithm by introducing an improved potential field factor and virtual obstacles to obtain an improved whale optimization algorithm.

[0012] S6. Use the pose estimation method to calculate the pose information of the goods to be sorted from the image to be sorted; based on the improved whale optimization algorithm, plan the motion trajectory of the robotic arm and the grasping posture of the dexterous hand according to the corrected depth map and the pose information, so as to control the dexterous hand at the end of the robotic arm to sort the goods to be sorted in the sorting area.

[0013] Based on the above technical solutions, the present invention can be further improved as follows.

[0014] Further, the S2 is specifically as follows:

[0015] S21. Extract features from the image to be sorted to obtain a two-dimensional feature map.

[0016] S22. Use the internal and external parameters of the monocular camera to project each grid point in the three-dimensional grid onto the two-dimensional feature map to obtain the initial features of each grid point in the three-dimensional grid.

[0017] Further, in the S21, specifically use the CNN convolutional neural network to extract features from the image to be sorted.

[0018] Further, the S3 is specifically as follows:

[0019] S31. Input the initial features of each grid point in the three-dimensional grid into the occupancy probability prediction model to obtain the initial occupancy probability of each grid point in the three-dimensional grid.

[0020] S32. Calculate the correction parameters of the initial features of each grid point in the three-dimensional grid according to the initial occupancy probability of each grid point in the three-dimensional grid.

[0021] S33. Correct the initial features of each grid point in the three-dimensional grid based on the correction parameters of the initial features of each grid point in the three-dimensional grid to obtain the corrected features of each grid point in the three-dimensional grid.

[0022] S34. Determine whether to continue the correction process according to the corrected features of each grid point in the three-dimensional grid; if so, execute S35; if not, execute S36.

[0023] S35. Take the calibration features of each grid point in the three-dimensional grid as the initial features of each grid point in the three-dimensional grid and return. Iteratively execute S31 until the calibration features of each grid point in the three-dimensional grid meet the preset conditions;

[0024] S36. Input the calibration features of each grid point in the three-dimensional grid into the occupancy probability prediction model to obtain the occupancy probability of each grid point in the three-dimensional grid;

[0025] Based on the volume rendering method, render the image to be sorted according to the occupancy probability of each grid point in the three-dimensional grid to obtain a camera depth map.

[0026] Further, in S31, the occupancy probability prediction model is specifically the Occupancy model.

[0027] Further, in S32, the calculation formula of the calibration parameter is:

[0028] ;

[0029] where, represents the initial occupancy probability of the th grid point in the three-dimensional grid, represents a preset refinement parameter, represents the calibration parameter.

[0030] Further, S33 is specifically:

[0031] Construct a feature interpolation model and adjust the model parameters of the feature interpolation model according to the calibration parameters of the initial features of each grid point in the three-dimensional grid;

[0032] Substitute the calibration features of each grid point in the three-dimensional grid into the feature interpolation model with adjusted model parameters to output a feature function;

[0033] Substitute the coordinates of each grid point in the three-dimensional grid into the feature function to obtain the calibration features of each grid point in the three-dimensional grid.

[0034] Further, S34 is specifically:

[0035] Calculate the average value of the calibration features of all grid points in the three-dimensional grid to obtain an average calibration feature;

[0036] Perform a difference operation on the calibration features of each grid point in the three-dimensional grid and the average calibration feature respectively to obtain the calibration feature differences of each grid point in the three-dimensional grid;

[0037] Determine whether the calibration feature differences of each grid point in the three-dimensional grid are within a preset feature difference threshold range, and count the point values of the grid points whose calibration feature differences are not within the preset feature difference threshold range;

[0038] Judge whether to continue the calibration process according to the point value; if so, execute S35; if not, execute S36;

[0039] Among them, if the point value is greater than or equal to the preset point value, execute S35; if the point value is less than the preset point value, execute S36.

[0040] Further, the specific content of S4 is as follows:

[0041] Construct a camera coordinate system according to the position information, internal and external parameters of the monocular camera, and construct a dexterous hand coordinate system according to the current position information and pose of the dexterous hand at the end of the robotic arm;

[0042] Calculate the calibration matrix between the monocular camera and the dexterous hand according to the conversion relationship between the camera coordinate system and the dexterous hand coordinate system;

[0043] Calibrate the camera depth map according to the calibration matrix to obtain a calibrated depth map.

[0044] Based on the above-mentioned motion planning and control method for a robot to sort goods, the present invention also provides a motion planning and control system for a robot to sort goods.

[0045] A motion planning and control system for a robot to sort goods, including a monocular camera, a robotic arm, and a toolbox arranged in sequence from top to bottom; the monocular camera is suspended above the toolbox through a bent pipe; the robotic arm is installed on the toolbox, and a suction and grasping integrated dexterous hand is installed on the robotic arm; a display is provided on the toolbox, four rolling wheels are provided at the bottom of the toolbox, and an industrial control computer is provided inside the toolbox;

[0046] The monocular camera is used to photograph the goods to be sorted in the sorting area to obtain an image of the goods to be sorted;

[0047] The industrial control computer is used to map the to-be-sorted image based on a three-dimensional grid to obtain the initial features of each grid point in the three-dimensional grid; based on a feature correction method and a volume rendering method, calculate a camera depth map according to the initial features of each grid point in the three-dimensional grid; correct the camera depth map according to the current position information of the dexterous hand at the end of the robotic arm and the position information of the monocular camera to obtain a corrected depth map; adopt an adaptive technique to improve the whale optimization algorithm by introducing an improved potential field factor and virtual obstacles to obtain an improved whale optimization algorithm; use a pose estimation method to calculate the pose information of the to-be-sorted commodity from the to-be-sorted image; based on the improved whale optimization algorithm, plan the motion trajectory of the robotic arm and the grasping pose of the dexterous hand according to the corrected depth map and the pose information to control the dexterous hand at the end of the robotic arm to sort the to-be-sorted commodity in the sorting area.

[0048] The beneficial effects of the present invention are as follows: For a method and system for motion planning and control of a robot for sorting commodities according to the present invention, by mapping a two-dimensional image into a three-dimensional space to obtain three-dimensional features, and then based on a feature correction method and a volume rendering method, a camera depth map can be accurately calculated. Since the dexterous hand and the camera are in different coordinate systems, correcting the camera depth map can obtain a corrected depth map relative to the dexterous hand, so that the dexterous hand can be controlled to quickly and accurately grasp the commodity; at the same time, the method of the present invention is simple, so it can save computer computing memory, improve the computing speed, and thus greatly improve the grasping speed of the robot; in addition, the present invention adopts an improved whale optimization algorithm, which can overcome the problem that the trajectory planning algorithm is prone to fall into local optimization, thereby improving the planning accuracy and ensuring the accurate grasping of the target object. Description of the Drawings

[0049] Figure 1 It is a flowchart of a method for motion planning and control of a robot for sorting commodities according to the present invention;

[0050] Figure 2 It is a flowchart of an improved whale optimization algorithm in a method for motion planning and control of a robot for sorting commodities according to the invention;

[0051] Figure 3 It is a structural diagram of a system for motion planning and control of a robot for sorting commodities according to the present invention.

[0052] In the drawings, the list of components represented by each reference numeral is as follows:

[0053] 1. Monocular camera; 2. Robotic arm; 3. Dexterous hand; 4. Toolbox; 5. Elbow pipe; 6. Display; 7. Pulley. Detailed Embodiments

[0054] The principles and features of the present invention will be described below in conjunction with the accompanying drawings. The examples given are only for explaining the present invention and are not intended to limit the scope of the present invention.

[0055] As Figure 1 shown, a motion planning and control method for a robot to sort goods includes:

[0056] S1. Use a monocular camera to capture the goods to be sorted in the sorting area to obtain an image of the goods to be sorted;

[0057] S2. Map the image of the goods to be sorted based on a three-dimensional grid to obtain the initial features of each grid point in the three-dimensional grid;

[0058] S3. Based on a feature correction method and a volume rendering method, calculate a camera depth map according to the initial features of each grid point in the three-dimensional grid;

[0059] S4. Correct the camera depth map according to the current position information of the dexterous hand at the end of the robotic arm and the position information of the monocular camera to obtain a corrected depth map;

[0060] S5. Adopt an adaptive technique to improve the whale optimization algorithm by introducing an improved potential field factor and virtual obstacles to obtain an improved whale optimization algorithm;

[0061] S6. Use a pose estimation method to calculate the pose information of the goods to be sorted from the image of the goods to be sorted; based on the improved whale optimization algorithm, plan the motion trajectory of the robotic arm and the grasping posture of the dexterous hand according to the corrected depth map and the pose information, so as to control the dexterous hand at the end of the robotic arm to sort the goods to be sorted in the sorting area.

[0062] In the present invention, a monocular camera is used for shooting, which can avoid image distortion caused by shooting with a multi-camera, so that the distortion correction process can be saved and the processing speed can be improved.

[0063] In some embodiments, S2 is specifically as follows:

[0064] S21. Extract features from the image of the goods to be sorted to obtain a two-dimensional feature map;

[0065] S22. Use the internal and external parameters of the monocular camera to project each grid point in the three-dimensional grid onto the two-dimensional feature map to obtain the initial features of each grid point in the three-dimensional grid.

[0066] Among them, in S21, a CNN convolutional neural network is specifically used to extract features from the image of the goods to be sorted.

[0067] In the present invention, by projecting the three-dimensional points in the three-dimensional grid onto the two-dimensional feature map, three-dimensional features can be quickly obtained.

[0068] In some embodiments: Specifically, S3 is as follows:

[0069] S31, input the initial features of each grid point in the three-dimensional grid into the occupancy probability prediction model to obtain the initial occupancy probability of each grid point in the three-dimensional grid;

[0070] S32, calculate the correction parameters of the initial features of each grid point in the three-dimensional grid according to the initial occupancy probability of each grid point in the three-dimensional grid;

[0071] S33, correct the initial features of each grid point in the three-dimensional grid based on the correction parameters of the initial features of each grid point in the three-dimensional grid to obtain the corrected features of each grid point in the three-dimensional grid;

[0072] S34, determine whether to continue the correction process according to the corrected features of each grid point in the three-dimensional grid; if so, execute S35; if not, execute S36;

[0073] S35, use the corrected features of each grid point in the three-dimensional grid as the initial features of each grid point in the three-dimensional grid and return to S31 for iterative execution until the corrected features of each grid point in the three-dimensional grid meet the preset conditions;

[0074] S36, input the corrected features of each grid point in the three-dimensional grid into the occupancy probability prediction model to obtain the occupancy probability of each grid point in the three-dimensional grid;

[0075] Based on the volume rendering method, render the image to be sorted according to the occupancy probability of each grid point in the three-dimensional grid to obtain the camera depth map.

[0076] Specifically, in S31, the occupancy probability prediction model is specifically the Occupancy model.

[0077] Specifically, in S32, the calculation formula of the correction parameter is:

[0078] ;

[0079] Wherein, represents the initial occupancy probability of the th grid point in the three-dimensional grid, represents the preset refinement parameter, represents the correction parameter.

[0080] Specifically, S33 is specifically:

[0081] Construct a feature interpolation model and adjust the model parameters of the feature interpolation model according to the correction parameters of the initial features of each grid point in the three-dimensional grid;

[0082] Substitute the corrected features of each grid point in the three-dimensional grid into the feature interpolation model with adjusted model parameters to output a feature function;

[0083] Substitute the coordinates of each grid point in the three-dimensional grid into the feature function to obtain the corrected features of each grid point in the three-dimensional grid.

[0084] Specifically, the S34 is specifically as follows:

[0085] Calculate the average value of the corrected features of all grid points in the three-dimensional grid to obtain an average corrected feature;

[0086] Subtract the average corrected feature from the corrected features of each grid point in the three-dimensional grid respectively to obtain the corrected feature differences of each grid point in the three-dimensional grid;

[0087] Judge whether the corrected feature differences of each grid point in the three-dimensional grid are within the preset feature difference threshold range, and count the number of grid points whose corrected feature differences are not within the preset feature difference threshold range;

[0088] Judge whether to continue the correction process according to the number of points; if so, execute S35; if not, execute S36;

[0089] Among them, if the number of points is greater than or equal to the preset number of points, execute S35; if the number of points is less than the preset number of points, execute S36.

[0090] Based on the feature correction method and the volume rendering method, the present invention can accurately calculate the camera depth map, laying a foundation for improving the grasping accuracy subsequently.

[0091] In some embodiments, the S4 is specifically as follows:

[0092] Construct a camera coordinate system according to the position information and internal and external parameters of the monocular camera, and construct a dexterous hand coordinate system according to the current position information and pose of the dexterous hand at the end of the robotic arm;

[0093] Calculate the correction matrix between the monocular camera and the dexterous hand according to the conversion relationship between the camera coordinate system and the dexterous hand coordinate system;

[0094] Correct the camera depth map according to the correction matrix to obtain a corrected depth map.

[0095] In the present invention, since the dexterous hand and the camera are in different coordinate systems, correcting the depth map of the camera can obtain a corrected depth map relative to the dexterous hand, so that the dexterous hand can be controlled to quickly and accurately grasp the commodity.

[0096] In step S5 of the present invention, an adaptive technique is used to automatically adjust the algorithm convergence factor to improve the convergence speed of the algorithm; an improved potential field factor is used to achieve the dynamic obstacle avoidance ability of the algorithm through the action of gravity and repulsion; virtual obstacles are used to avoid the local optimization trap of the algorithm, so as to achieve the purpose of capturing unreachable targets and local optimal solutions.

[0097] The flow chart of the improved whale optimization algorithm is as Figure 2 shown.

[0098] Among them, the formula for population initialization is: , represents the position of individual , and respectively represent the upper and lower bounds of the search space, represents a random number between 0 and 1. In the stage of surrounding the prey, , , ; is the current iteration number, A and C are coefficients, is the position of the current solution, is the position of the current optimal solution, gradually decreases to 0 during the iteration process.

[0099] Based on the above-mentioned motion planning and control method for a robot to sort commodities, the present invention also provides a motion planning and control system for a robot to sort commodities.

[0100] As Figure 3 shown, a motion planning and control system for a robot to sort commodities includes a monocular camera 1, a robotic arm 2, and a toolbox 4 arranged in sequence from top to bottom; the monocular camera 1 is suspended above the toolbox 4 through a bent pipe 5; the robotic arm 2 is installed on the toolbox 4, and a suction and grasping integrated dexterous hand 3 is installed on the robotic arm 2; a display 6 is provided on the toolbox 4, four rolling pulleys 7 are provided at the bottom of the toolbox 4, and an industrial control computer is provided inside the toolbox 4;

[0101] The monocular camera 1 is used to photograph the commodities to be sorted in the sorting area to obtain an image of the commodities to be sorted;

[0102] The industrial control computer is used to map the image to be sorted based on a three-dimensional grid to obtain the initial features of each grid point in the three-dimensional grid; based on a feature correction method and a volume rendering method, calculate a camera depth map according to the initial features of each grid point in the three-dimensional grid; correct the camera depth map according to the current position information of the dexterous hand at the end of the robotic arm and the position information of the monocular camera to obtain a corrected depth map; adopt an adaptive technique to improve the whale optimization algorithm by introducing an improved potential field factor and virtual obstacles to obtain an improved whale optimization algorithm; use a pose estimation method to calculate the pose information of the commodity to be sorted from the image to be sorted; based on the improved whale optimization algorithm, plan the motion trajectory of the robotic arm 2 and the grasping posture of the dexterous hand 3 according to the corrected depth map and the pose information, so as to control the dexterous hand 3 at the end of the robotic arm 2 to sort the commodity to be sorted in the sorting area.

[0103] In a motion planning and control system for a robot to sort commodities according to the present invention, the monocular camera 1 identifies commodity information and transmits it to the display 6, and then the industrial control computer and the dexterous hand power supply inside the toolbox 4 control the robotic arm to sort the commodities on the conveyor belt to a suitable position. Four rolling pulleys 7 are provided at the bottom of the toolbox 4 to facilitate movement. The present invention integrates a camera, a robotic arm, a dexterous hand, and a display, and has the characteristics of compact structure, space saving, and convenient movement, and can complete various functions of taking and using goods in an unmanned supermarket, replacing manual labor and improving production efficiency.

[0104] For a motion planning and control method and system for a robot to sort commodities according to the present invention, by mapping a two-dimensional image to a three-dimensional space to obtain three-dimensional features, and then based on a feature correction method and a volume rendering method, a camera depth map can be accurately calculated. Since the dexterous hand and the camera are in different coordinate systems, correcting the camera depth map can obtain a corrected depth map relative to the dexterous hand, so that the dexterous hand can be controlled to quickly and accurately grasp the commodity; at the same time, the method of the present invention is simple, so it can save computer computing memory and improve computing speed, and thus greatly improve the grasping speed of the robot; in addition, the present invention adopts an improved whale optimization algorithm, which can overcome the problem that the trajectory planning algorithm is prone to fall into local optimization, thereby improving the planning accuracy and ensuring the accurate grasping of the target object.

[0105] The above are only the preferred embodiments of the present invention, and are not intended to limit the present invention. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A motion planning and control method for robot sorting goods, characterized in that: include: S1, using a monocular camera to photograph the goods to be sorted in the sorting area to obtain an image of the goods to be sorted; S2, mapping the image to be sorted based on a three-dimensional grid to obtain initial features of each grid point in the three-dimensional grid; S3, calculating a camera depth map according to initial features of each grid point in the three-dimensional grid based on a feature correction method and a volume rendering method; S4, correcting the camera depth map according to the current position information of the dexterous hand at the end of the robotic arm and the position information of the monocular camera to obtain a corrected depth map; S5, using adaptive technology, improves the whale optimization algorithm by introducing improved potential field factors and virtual obstacles to obtain an improved whale optimization algorithm; S6, calculating the posture information of the goods to be sorted from the image to be sorted using a posture estimation method; Based on the improved whale optimization algorithm, the motion trajectory of the robotic arm and the grasping posture of the dexterous hand are planned according to the corrected depth map and the posture information, so as to control the dexterous hand at the end of the robotic arm to sort the goods to be sorted in the sorting area.

2. The motion planning and control method for robot sorting commodities according to claim 1, characterized in that: The S2 is specifically: S21, extracting features from the image to be sorted to obtain a two-dimensional feature map; S22, using the internal and external parameters of the monocular camera, projecting each grid point in the three-dimensional grid onto the two-dimensional feature map to obtain initial features of each grid point in the three-dimensional grid.

3. The motion planning and control method for robot sorting commodities according to claim 2, characterized in that: In S21, a CNN convolutional neural network is specifically used to extract features of the image to be sorted.

4. The motion planning and control method for robot sorting commodities according to claim 1, characterized in that: The S3 is specifically: S31, inputting the initial features of each grid point in the three-dimensional grid into an occupancy probability prediction model to obtain the initial occupancy probability of each grid point in the three-dimensional grid; S32, calculating a correction parameter of an initial feature of each grid point in the three-dimensional grid according to the initial occupancy probability of each grid point in the three-dimensional grid; S33, correcting the initial features of each grid point in the three-dimensional grid based on the correction parameters of the initial features of each grid point in the three-dimensional grid, to obtain the corrected features of each grid point in the three-dimensional grid; S34, judging whether to continue the correction process according to the correction characteristics of each grid point in the three-dimensional grid; If yes, execute S35; if no, execute S36; S35, taking the correction feature of each grid point in the three-dimensional grid as the initial feature of each grid point in the three-dimensional grid and returning to S31 to execute iteratively in a loop until the correction feature of each grid point in the three-dimensional grid meets a preset condition; S36, inputting the correction feature of each grid point in the three-dimensional grid into the occupancy probability prediction model to obtain the occupancy probability of each grid point in the three-dimensional grid; Based on a volume rendering method, the image to be sorted is rendered according to the occupancy probability of each grid point in the three-dimensional grid to obtain a camera depth map.

5. The motion planning and control method for robot sorting commodities according to claim 4, characterized in that: In S31, the occupancy probability prediction model is specifically an Occupancy model.

6. The motion planning and control method for robot sorting commodities according to claim 4, characterized in that: In the S32, the calculation formula of the correction parameter is: ； in, Indicates the first The initial occupancy probability of grid points, Represents the preset refinement parameters, represents the correction parameter.

7. The motion planning and control method for robot sorting commodities according to claim 4, characterized in that: The S33 is specifically: Constructing a feature interpolation model, and adjusting model parameters of the feature interpolation model according to correction parameters of initial features of each grid point in the three-dimensional grid; Substituting the correction features of each grid point in the three-dimensional grid into the feature interpolation model after the model parameters are adjusted to output a feature function; The coordinates of each grid point in the three-dimensional grid are substituted into the characteristic function to obtain the correction features of each grid point in the three-dimensional grid.

8. The motion planning and control method for robot sorting commodities according to claim 4, characterized in that: The S34 is specifically: Calculating the average value of the correction features of all grid points in the three-dimensional grid to obtain the average correction feature; Performing difference processing on the correction feature of each grid point in the three-dimensional grid and the average correction feature respectively, to obtain the correction feature difference of each grid point in the three-dimensional grid; Determine whether the corrected feature difference of each grid point in the three-dimensional grid is within a preset feature difference threshold range, and count the point values ​​of the grid points whose corrected feature difference is not within the preset feature difference threshold range; Determine whether to continue the calibration process according to the point value; If yes, execute S35; if no, execute S36; If the point value is greater than or equal to the preset point value, S35 is executed; if the point value is less than the preset point value, S36 is executed.

9. The motion planning and control method for robot sorting commodities according to claim 1, characterized in that: The S4 is specifically: A camera coordinate system is constructed according to the position information and internal and external parameters of the monocular camera, and a dexterous hand coordinate system is constructed according to the current position information and posture of the dexterous hand at the end of the robotic arm; Calculating a correction matrix between the monocular camera and the dexterous hand according to a conversion relationship between the camera coordinate system and the dexterous hand coordinate system; The camera depth map is corrected according to the correction matrix to obtain a corrected depth map.

10. A motion planning and control system for a robot to sort goods, characterized in that: The invention comprises a monocular camera (1), a mechanical arm (2) and a tool box (4) which are arranged in sequence from top to bottom; the monocular camera (1) is suspended above the tool box (4) through a curved pipe (5); the mechanical arm (2) is mounted on the tool box (4), and a suction-grasping dexterous hand (3) is mounted on the mechanical arm (2); a display (6) is provided on the tool box (4), four rolling pulleys (7) are provided at the bottom of the tool box (4), and an industrial computer is provided inside the tool box (4); The monocular camera (1) is used to photograph the commodities to be sorted in the sorting area to obtain an image of the commodities to be sorted; The industrial computer is used to map the image to be sorted based on a three-dimensional grid to obtain initial features of each grid point in the three-dimensional grid; Based on the feature correction method and the volume rendering method, a camera depth map is calculated according to the initial features of each grid point in the three-dimensional grid; the camera depth map is corrected according to the current position information of the dexterous hand at the end of the robotic arm and the position information of the monocular camera to obtain a corrected depth map; Adaptive technology is used to improve the whale optimization algorithm by introducing improved potential field factors and virtual obstacles to obtain an improved whale optimization algorithm; Calculating the pose information of the goods to be sorted from the image to be sorted by using a pose estimation method; Based on the improved whale optimization algorithm, the motion trajectory of the robot arm (2) and the grasping posture of the dexterous hand (3) are planned according to the corrected depth map and the posture information, so as to control the dexterous hand (3) at the end of the robot arm (2) to sort the goods to be sorted in the sorting area.

Images