A cotton picking robot and a picking navigation method thereof

Abstract

The application provides a cotton picking robot, and relates to the field of intelligent picking, comprising two groups of multi-stage lifting frame mechanisms, a fixed horizontal plate (20), a steering walking mechanism, a connecting collecting box (40), two groups of end execution mechanisms and a supporting mechanism, the two groups of multi-stage lifting frame mechanisms are symmetrically arranged at the two ends of the lower side of the fixed horizontal plate (20) and are connected with the bottom surface of the fixed horizontal plate (20), the steering walking mechanism is arranged at the bottom of the multi-stage lifting frame mechanism, the connecting collecting box (40) is arranged between the two groups of multi-stage lifting frame mechanisms, the two groups of end execution mechanisms are symmetrically arranged at the middle part of the bottom surface of the fixed horizontal plate (20), and the front end of the fixed horizontal plate (20) is provided with the supporting mechanism. The picking robot can realize efficient, accurate and complete picking of cotton, and can avoid the problems of residual lint at the root of the cotton boll and crushing of the lodged stalks during the picking process.

Description

Technical Field

[0001] This invention relates to the field of intelligent harvesting technology, specifically to a cotton harvesting robot and its harvesting navigation method. Background Technology

[0002] Cotton is a vital global economic commodity, with over 90% of it currently used for clothing, home textiles, and industrial fabrics. Cotton is classified into long-staple, short-staple, and coarse-staple fibers, with long-staple cotton, known for its long and flexible fibers, often referred to as "gold" among cotton varieties. Currently, long-staple cotton is primarily harvested mechanically. However, the fruiting branch in long-staple cotton has a relatively short initial node (the node where the first flowering and fruiting branch appears on the main stem), making it easy for machines to miss the bottom bolls. This not only results in incomplete harvesting and reduced cotton yield but also increases the risk of pests and diseases, impacting subsequent cotton planting. Furthermore, the taller height of long-staple cotton plants makes them prone to lodging, affecting harvesting efficiency, increasing harvesting difficulty, and causing damage to the stems from harvesting machinery, reducing the yield and quality of future cotton harvests. In addition, the outward growth of long-staple cotton plants can cause rows to close together. Combined with the complex field environment, this makes it difficult for cotton harvesting machinery to move freely and accurately in the cotton field, which can easily damage the long-staple cotton plants and lead to problems such as reduced cotton yield and machinery jamming. Summary of the Invention

[0003] To address the problems existing in the prior art, the present invention aims to provide a cotton harvesting robot. This harvesting robot can not only achieve efficient, precise, and complete cotton harvesting, avoiding the problem of cotton bolls being left at the base during the harvesting process, but also lift lodged stems during the harvesting process, avoiding crushing of lodged stems, and effectively protecting cotton plants during the harvesting process.

[0004] Another objective of this invention is to provide a cotton picking robot navigation method that not only ensures accurate and efficient cotton picking, but also enables the picking robot to move autonomously in rows in cotton fields, thereby improving picking efficiency and ensuring the yield and quality of the picked cotton.

[0005] The objective of this invention is achieved through the following technical solution:

[0006] A cotton harvesting robot includes two sets of multi-stage lifting truss mechanisms, a fixed horizontal plate, a steering and walking mechanism, a connecting collection box, two sets of end effectors, and a seed-lifting mechanism. The two sets of multi-stage lifting truss mechanisms are symmetrically arranged at both ends of the lower side of the fixed horizontal plate and connected to the bottom surface of the fixed horizontal plate. The steering and walking mechanism is located at the bottom of the multi-stage lifting truss mechanisms, and two sets of steering and walking mechanisms are arranged at the bottom of one set of multi-stage lifting truss mechanisms. The connecting collection box is located between the two sets of multi-stage lifting truss mechanisms, and both ends of the connecting collection box are fixedly connected to the inner side of the corresponding multi-stage lifting truss mechanism. The connecting collection box is located at the rear end of the fixed horizontal plate. The two sets of end effectors are symmetrically arranged in the middle of the bottom surface of the fixed horizontal plate and located inside the two sets of multi-stage lifting truss mechanisms. The seed-lifting mechanism is arranged at the front end of the fixed horizontal plate.

[0007] Based on further optimization of the above scheme, the multi-stage lifting truss mechanism includes a positioning plate, two sets of first lifting components, a second lifting component, two sets of positioning support components, and a chassis crossbeam. The two sets of first lifting components are symmetrically arranged on both sides of the inner wall of the positioning plate, including a first cylinder and a first piston rod. The first cylinder is located on the inner wall of the positioning plate and its lower end is slidably connected to the first piston rod. The chassis crossbeam is located on the lower side of the corresponding positioning plate, and the end of the first piston rod away from the first cylinder is fixedly connected to the end face of the chassis crossbeam. The second lifting component is located in the middle of the inner wall of the positioning plate, including a second cylinder and a second piston rod. The second cylinder is located on the inner wall of the positioning plate and its upper end is slidably connected to the second piston rod. The end of the second piston rod away from the second cylinder is fixedly connected to the bottom surface of the fixed crossbeam. Positioning support components are respectively arranged between the second lifting component and the two sets of first lifting components. The positioning support component includes a sliding guide cylinder and a guide pillar. The sliding guide cylinder is located on the inner wall of the positioning plate, and the guide pillar is slidably arranged on its upper end. The end of the guide pillar away from the sliding guide cylinder is fixedly connected to the bottom surface of the fixed crossbeam.

[0008] Based on further optimization of the above scheme, the steering and walking mechanism includes a wheel fork, a steering bracket, a connecting shaft and a walking wheel. The wheel fork is fixedly connected to the bottom surface of the chassis crossbeam and the longitudinal section of the wheel fork is a Z-shaped structure. The steering bracket is rotatably installed at the bottom of the wheel fork and the connecting shaft is rotatably installed in the middle of the steering bracket. The walking wheel is fixedly sleeved on the outer wall of the connecting shaft.

[0009] Based on further optimization of the above scheme, the end effector includes a robotic arm assembly and an actuator assembly. The robotic arm assembly includes a rotating gimbal, a movable upper arm, an air supply pipe, an air source element, a movable lower arm, and a movable connecting arm. The rotating gimbal is mounted on the bottom surface of the fixed horizontal plate, and the movable upper arm is mounted at the bottom of the rotating gimbal. The end of the movable upper arm away from the rotating gimbal is rotatably connected to the movable lower arm, and the end of the movable lower arm away from the movable upper arm is rotatably connected to the movable connecting arm. One end of the air supply pipe is fixedly connected to the rotating gimbal via a valve, and the other end is fixedly connected to the air source element mounted on the movable upper arm via a valve. The actuator assembly includes a rear fixed plate, a front fixed plate, a fixed coupling shaft, a movable slide, a ball screw, a connecting rod, a rotating block, and three-jaw claws. The rear fixed plate is mounted on the movable connecting arm, and the front fixed plate is mounted on the side of the rear fixed plate away from the movable connecting arm. The front and rear fixed plates are arranged in parallel. Multiple fixed shafts are evenly arranged between the front and rear fixed plates and on their outer ring (connecting the front and rear fixed plates through the fixed shafts). A rotating ball screw is set in the middle of the front and rear fixed plates. A movable slide is set between the front and rear fixed plates and is passed through by the fixed shafts and the ball screw respectively. The movable slide is slidably connected to the fixed shafts and threadedly connected to the ball screw. Three rotating blocks are evenly arranged on the side of the front fixed plate away from the fixed shafts. The rotating blocks are rotatably connected to the front fixed plate. The movable slide is equipped with three connecting rods corresponding to the rotating blocks. One end of the connecting rod is rotatably connected to the movable slide and the other end is rotatably connected to the rotating block. Three claw fingers are set corresponding to the rotating blocks and one end of each claw finger is connected to the rotating block. Each claw finger has a protruding arc-shaped surface on its inner side.

[0010] Based on further optimization of the above scheme, the lifting mechanism includes a top ring, a hinged connecting rod, a lifting ring fixing disc, a fixed cylinder, a bottom disc, a rotating screw, a lifting ring, and levers. The top ring is located on the front end face of the fixed horizontal plate. One end of the hinged connecting rod is rotatably connected to the top ring, and the other end is rotatably connected to the lifting ring fixing disc. The fixed cylinder is coaxially located on the bottom surface of the lifting ring fixing disc, and the bottom disc is coaxially located at the end of the fixed cylinder away from the bottom surface of the lifting ring fixing disc. The rotating screw is coaxially located in the inner cavity of the fixed cylinder, and the upper end of the rotating screw passes through the lifting ring fixing disc and is rotatably connected. The lifting ring is sleeved on the outer wall of the fixed cylinder, and its inner wall is threadedly connected to the rotating screw located in the inner cavity of the fixed cylinder. Levers are symmetrically arranged on both sides of the lifting ring.

[0011] A cotton-harvesting robot navigation method, employing the aforementioned cotton-harvesting robot, includes:

[0012] Step S1: Adjusting the spatial height: When the picking robot reaches the pre-picking position, the spatial height of the picking robot is adjusted in real time through the multi-stage lifting trellis mechanism according to the growth of the long-staple cotton plants.

[0013] Step S2, Envelope Picking and Collection: The robotic arm assembly is activated, causing the actuator assembly to move to the target picking point. The actuator assembly envelops the cotton bolls of the long-staple cotton, and the robotic arm assembly moves to pick the cotton bolls and place them in the connecting collection box. Then the end effector resets, completing one picking cycle.

[0014] Step S3, Supporting the fallen stalks: Before the harvesting robot moves to the next harvestable space, it determines whether there are fallen stalks. If so, the supporting mechanism is activated to support the fallen long-staple cotton stalks.

[0015] Step S4, Field Ridge Finding Technology: Automatic ridge finding technology is used to enable the harvesting robot to find ridges in the cotton field and move the harvesting robot to the next harvestable space;

[0016] Step S5, Turning in the field: After completing the work on one row, turning in the narrow space of the field is achieved through four sets of steering and walking mechanisms;

[0017] Step S6, Cotton Field Navigation Method: By combining navigation methods, the harvesting robot can achieve efficient autonomous navigation in different environments.

[0018] Based on further optimization of the above scheme, step S4 specifically includes:

[0019] Step S41: Acquire images of the cotton field using a high-resolution camera to obtain the dataset;

[0020] Step S42: First, the collected image is converted to grayscale to obtain a grayscale image, and then the image is filtered to obtain a smooth image;

[0021] Step S43: Perform edge detection to obtain the ridge line edges in the image;

[0022] Step S44: First, extract the ridge lines in the image using the Hough transform method and determine the ridge line parameters:

[0023]

[0024] In the formula, This represents the distance from the straight line to the origin; Angle representing a straight line; x,y () represents the pixel coordinates;

[0025] Then, the least squares method is used to fit a straight line to the detected edge points, and the error is minimized. The position of the ridge line is described by the fitted straight line:

[0026]

[0027] In the formula: ( xi ,y i () represents the pixel coordinates of the edge point;

[0028] Step S45: Track the ridge line by measuring the lateral deviation and heading angle deviation between the harvesting robot and the target ridge line.

[0029]

[0030] In the formula: Y 1 indicates the current position of the harvesting robot. kx + b Represent the target ridge line equation; Indicates the current heading angle. The direction angle corresponding to the slope of the target ridge line; e Y Indicates heading deviation. Indicates the deviation of the heading angle;

[0031] Control commands are generated based on lateral and heading angle deviations. Turning commands are then generated based on the current position and target point of the harvesting robot. The robot's direction is adjusted using feedback and pre-aiming points.

[0032] Select a future point on the target path. x f ,y f ), as a pre-aiming point:

[0033]

[0034] In the formula: L d Indicates the aiming distance, i.e., the current point ( x,y The straight-line distance between the target and the aiming point;

[0035] The PID controller compensates for the heading angle deviation and generates the steering angle command. :

[0036]

[0037] In the formula: This represents the weighting coefficient, which takes the value [0, 1]. K p , K i , K d These represent the proportional, integral, and differential parameters, respectively.

[0038] Speed ​​adjustment command:

[0039]

[0040] In the formula: v 0 represents the base speed. K v This indicates the speed adjustment coefficient.

[0041] Based on further optimization of the above scheme, step S6 specifically includes:

[0042] Step S61: Initialize various navigation parameters, sampling frequency, and cutoff frequency; construct prediction models for position, velocity, and attitude;

[0043] Step S62: Obtain the covariance of the prediction error by calculating the prediction model. First, calculate the state transition matrix. F k :

[0044]

[0045] In the formula: I Represents the identity matrix; The cross product matrix representing the angular velocity of the gyroscope; The cross product matrix representing accelerometer measurements; R(q) This represents the rotation matrix corresponding to the attitude quaternion; This represents the time difference between the current moment and the next predicted moment.

[0046] Then, the process noise covariance is calculated. Q k :

[0047]

[0048] In the formula: Q ins Represents the inertial navigation noise covariance matrix; Q vio Represents the visual navigation noise covariance matrix; diag () This means arranging the input submatrices along the main diagonal to form a block diagonal matrix;

[0049] Finally, covariance prediction calculations are performed:

[0050]

[0051] In the formula: This represents the error covariance matrix of the prediction; This represents the error covariance matrix of the previous time step;

[0052] Step S63: Construct an observation model using Global Navigation Satellite System (GNSS) observations and visual observations:

[0053] Global Navigation Satellite System (GNSS) observation model:

[0054]

[0055] In the formula: Z GNSS This represents GNSS observations, which include a combination of position and velocity information; P GNSS This represents the three-dimensional position coordinates obtained through a GNSS sensor; V GNSS This represents the three-dimensional velocity coordinates obtained through a GNSS sensor; P INS This indicates the position predicted by the inertial navigation system through IMU integration; V INS This represents the velocity predicted by the inertial navigation system through IMU integration; This represents the position error, specifically the deviation between the GNSS observations and the INS predictions. This represents the velocity error, specifically the velocity deviation between GNSS observations and INS predictions. This indicates the measurement noise within a satellite navigation GNSS system.

[0056] The observation matrix of the GNSS observation model is:

[0057]

[0058] Visual observation model: Visual observation uses a method to calculate reprojection error, the first... i The observed coordinates of the feature points are Then its projection model is:

[0059]

[0060] In the formula: X i This represents the coordinates of the feature point in the global coordinate system; () represents the camera projection function; R VIO The rotation matrix represents visual navigation; p VIO Indicates the location of visual navigation; v vis Indicates visual observation noise;

[0061] The observation matrix of the visual observation model is:

[0062]

[0063] in:

[0064]

[0065] In the formula: Represents the camera projection function; R Represents the rotation matrix; Indicates the attitude error of inertial navigation; Indicates the attitude error in visual navigation; This indicates the positional error in visual navigation;

[0066] Kalman gain is used to dynamically adjust the weights between the predicted values ​​of the prediction model and the dynamic values ​​of the observation model to optimize the accuracy of state estimation.

[0067]

[0068]

[0069] In the formula: K k Represents the Kalman gain matrix. H k This represents the observation matrix obtained by the observation model, i.e., the fused observation source; R k Represents the observation noise covariance matrix;

[0070] Step S64: Update the attitude, position, velocity, and visual error states respectively to obtain the corrected states;

[0071] Step S65: First, an inertial measurement unit (IMU) is used to obtain the prediction models of attitude, position, and velocity during the prediction phase, and the corresponding covariance of the prediction error is also obtained. Then, the visual observation model is used for observation, and visual updates are performed, the state is continuously corrected, and the covariance is updated. When the Global Navigation Satellite System (GNSS) data is available (i.e., the GNSS module receives enough satellite signals to accurately calculate the position information), the position and velocity residuals are calculated, the error state of the GNSS observation model is updated again, and the covariance is updated again. The above prediction update steps are repeated to continuously fuse multi-sensor data and improve navigation and positioning accuracy.

[0072] Based on further optimization of the above scheme, the specific steps in step S61 for constructing the prediction model for position, velocity, and attitude are as follows:

[0073] By using gyroscope data to predict the attitude at the next moment, an attitude prediction model can be constructed:

[0074]

[0075] In the formula: Represents the predicted attitude quaternion (i.e. k (Predicted value at time). The attitude quaternion represents the previous time step; Indicates the time step (unit: s); This indicates the angular velocity measured by the gyroscope (unit: rad / s). Indicates zero bias of the gyroscope (unit: rad / s); This represents the quaternion multiplication operator; exp This represents a minute rotation, that is, converting angular velocity into an increment of quaternion;

[0076] Accelerometer data is used to predict the velocity at the next moment, thereby constructing a velocity prediction model:

[0077]

[0078] In the formula: Indicates the speed of prediction (i.e.) k (Predicted value at time) Represents the velocity at the previous moment. Rotation matrix corresponding to attitude quaternion; This indicates the value measured by the accelerometer; Indicates zero bias of the accelerometer (unit: m / s²); g Represents the gravitational acceleration vector;

[0079] By using velocity integrals to predict the position at the next moment, a position prediction model can be constructed:

[0080]

[0081] In the formula: Indicates the predicted location ( k (Predicted value at time) It indicates the position at the previous moment.

[0082] Based on further optimization of the above scheme, the specific steps for updating the attitude, position, velocity, and visual error states in step S64 are as follows:

[0083] Attitude correction:

[0084]

[0085] In the formula: express k The attitude error is updated in real time; it is achieved through... k At any given time, GNSS observations and INS-predicted attitude values ​​are obtained.

[0086] Position correction:

[0087]

[0088] In the formula: express k The position error is updated in real time; it is achieved through... k The position deviation between the satellite navigation GNSS observation and the inertial navigation INS prediction at a given time is obtained;

[0089] Speed ​​correction:

[0090]

[0091] In the formula: express k The speed error is updated in real time; it is achieved through... k The velocity deviation between the satellite navigation GNSS observation and the inertial navigation INS prediction at a given time is obtained.

[0092] Visual error correction:

[0093]

[0094] In the formula: express k Continuously update the visual positional error; express k Continuously update visual posture errors; express k The position of vision at time -1; express k Visual posture at time -1; express k The constantly updated visual position; express k The visual posture is constantly updated.

[0095] Based on further optimization of the above scheme, the specific steps for updating the covariance in step S65 are as follows:

[0096]

[0097] In the formula: This represents the updated error covariance matrix.

[0098] The following are the technical effects of the present invention:

[0099] This invention utilizes a three-claw actuator assembly to effectively adapt to the boll portion of long-staple cotton (i.e., the boll portion of long-staple cotton is primarily three-lobed), thereby solving the problem of cotton fibers remaining at the base of the boll during long-staple cotton harvesting and ensuring complete harvesting. Through the coordination of a multi-stage lifting truss mechanism with a fixed horizontal plate, a steering and walking mechanism, and a connecting collection box, the overall structure forms a gantry-like structure, effectively solving the problem of long-staple cotton rows clustering together and the chassis being unable to move normally in narrow spaces. Furthermore, the multi-stage lifting method facilitates height adjustment during harvesting, and in conjunction with the robotic arm assembly, achieves efficient long-staple cotton boll harvesting. It can also lower the entire boll during non-harvesting periods. The height of the robot body facilitates normal movement on field roads. The robotic arm assembly, consisting of a rotating gimbal, a large moving arm, an air supply pipe, an air source element, a moving forearm, and a connecting arm, uses pneumatic power to drive the rotating gimbal, providing strong load capacity and enabling precise and efficient long-staple cotton harvesting. The supporting mechanism, composed of a top ring, hinged connecting rod, hanging ring fixing disc, fixing cylinder, bottom disc, rotating screw, lifting ring, and lever, effectively solves the problem of long-staple cotton lodging, facilitates harvesting, prevents the long-staple cotton harvesting robot from crushing cotton stalks during movement, and reduces damage to surrounding unharvested long-staple cotton.

[0100] Furthermore, this invention employs row-finding technology to enable the long-staple cotton harvesting robot to move autonomously in rows within cotton fields, avoiding damage to the long-staple cotton plants. This invention also utilizes a navigation method combining BeiDou satellite navigation, visual navigation, and inertial navigation, enabling it to adapt to various navigation applications and improving the navigation accuracy of the long-staple cotton harvesting robot in different environments such as foliage obstruction, cloudy days, and strong light (for example, in cloudy conditions, BeiDou satellite navigation is significantly affected, while visual navigation performs better; in strong light, visual navigation is significantly affected, while BeiDou satellite navigation performs better; when foliage obstructs the view, visual navigation is more susceptible to interference, while BeiDou satellite navigation and inertial navigation perform better). This precisely leverages the advantages of different navigation methods, compensating for their weaknesses and improving positioning accuracy. Attached Figure Description

[0101] Figure 1 This is a schematic diagram of the overall structure of the harvesting robot in an embodiment of the present invention.

[0102] Figure 2 This is a front view of the harvesting robot in an embodiment of the present invention.

[0103] Figure 3 This is an internal view of the multi-stage lifting truss mechanism of the harvesting robot in an embodiment of the present invention.

[0104] Figure 4 This is a schematic diagram of the mechanical arm assembly of the harvesting robot in an embodiment of the present invention.

[0105] Figure 5This is a schematic diagram of the actuator assembly of the harvesting robot in an embodiment of the present invention.

[0106] Figure 6 This is a schematic diagram of the steering and walking mechanism of the harvesting robot in an embodiment of the present invention.

[0107] Figure 7 This is a schematic diagram of the harvesting robot's lifting mechanism in an embodiment of the present invention.

[0108] Among them, 11. Positioning plate; 121. First cylinder; 122. First piston rod; 131. Second cylinder; 132. Second piston rod; 141. Sliding guide cylinder; 142. Guide support column; 15. Chassis crossbeam; 20. Fixed cross plate; 31. Wheel fork holder; 32. Steering bracket; 33. Connecting shaft; 34. Travel wheel; 40. Connecting collection box; 511. Rotating gimbal; 512. Moving boom; 513. Air supply pipe; 514. Air source element 515. Moving forearm; 516. Moving connecting arm; 521. Rear fixed plate; 522. Front fixed plate; 523. Fixed coupling shaft; 524. Moving slide; 525. Ball screw; 526. Connecting rotating rod; 527. Rotating block; 528. Three-jaw claw; 61. Top ring; 62. Hinge connecting rod; 63. Hanging ring fixing disc; 64. Fixed cylinder; 65. Bottom disc; 66. Rotating screw; 67. Lifting ring; 68. Lever. Detailed Implementation

[0109] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments.

[0110] Example 1:

[0111] A cotton harvesting robot includes two sets of multi-stage lifting truss mechanisms, a fixed horizontal plate 20, a steering and walking mechanism, a connecting collection box 40, two sets of end effectors, and a cotton-lifting mechanism (such as...). Figure 1 As shown), two sets of multi-stage lifting truss mechanisms are symmetrically arranged at both ends of the lower side of the fixed horizontal plate 20 and connected to the bottom surface of the fixed horizontal plate 20. The multi-stage lifting truss mechanism includes a positioning plate 11, two sets of first lifting components, a second lifting component, two sets of positioning support components, and a chassis crossbeam 15 (as shown). Figure 2(As shown); Two sets of first lifting components are symmetrically arranged on both sides of the inner wall of the positioning plate 11, including a first cylinder 121 and a first piston rod 122. The first cylinder 121 is located on the inner wall of the positioning plate 11 and its lower end is slidably connected to the first piston rod 122. The chassis crossbeam 15 is located on the lower side of the corresponding positioning plate 11, and the end of the first piston rod 122 away from the first cylinder 121 (i.e., the bottom end) is fixedly connected to the end face of the chassis crossbeam 15. A second lifting component is arranged in the middle of the inner wall of the positioning plate 11, including a second cylinder 131 and a second piston rod 132. The second cylinder 131 is located on the inner wall of the positioning plate 11 and its upper end is slidably connected to the second piston rod. 132, the end of the second piston rod 132 away from the second cylinder 131 (i.e., the top end) is fixedly connected to the bottom surface of the fixed horizontal plate 20; positioning support components are respectively provided between the second lifting assembly and the two sets of first lifting assemblies. The positioning support components include a sliding guide cylinder 141 and a guide pillar 142. The sliding guide cylinder 141 is set on the inner wall of the positioning plate 11 and the guide pillar 142 is slidably set on its upper end. The end of the guide pillar 142 away from the sliding guide cylinder 141 (i.e., the top end) is fixedly connected to the bottom surface of the fixed horizontal plate 20 (the connection between the positioning plate 11 and the fixed horizontal plate 20 is achieved through the second piston rod 132 and the guide pillar 142).

[0112] The steering and traveling mechanism is located at the bottom of the multi-stage lifting truss mechanism, and two sets of steering and traveling mechanisms are located at the bottom of one set of multi-stage lifting truss mechanisms (e.g. Figure 1 As shown, in this embodiment, four sets of steering and traveling mechanisms are provided, and the steering and traveling mechanisms correspond to each other; the steering and traveling mechanism includes a wheel fork 31, a steering bracket 32, a connecting shaft 33, and a traveling wheel 34. The wheel fork 31 is fixedly connected to the bottom surface of the chassis crossbeam 15, and the longitudinal section of the wheel fork 31 is a Z-shaped structure (e.g., Figure 6 As shown), a steering bracket 32 ​​is rotatably mounted at the bottom of the wheel fork 31, and a connecting shaft 33 is rotatably mounted in the middle of the steering bracket 32. The outer wall of the connecting shaft 33 is fixedly sleeved with the traveling wheel 34 (the traveling wheel 34 is rotated by controlling the rotation of the connecting shaft 33).

[0113] The connecting collection box 40 is positioned between two sets of multi-stage lifting truss mechanisms, and both ends of the connecting collection box 40 are fixedly connected to the inner side of the corresponding multi-stage lifting truss mechanism (i.e., positioning plate 11). Figure 2 As shown), the connecting collection box 40 is located at the rear end of the fixed horizontal plate 20 (i.e. Figure 1 (The left end is shown).

[0114] Two sets of end effectors are symmetrically arranged in the middle of the bottom surface of the fixed horizontal plate 20 and located inside the two sets of multi-stage lifting truss mechanisms. The end effectors include a robotic arm assembly and an actuator assembly. The robotic arm assembly includes a rotating gimbal 511, a moving upper arm 512, an air supply pipe 513, an air source element 514, a moving lower arm 515, and a moving connecting arm 516 (e.g., Figure 4As shown), the rotating gimbal 511 is mounted on the bottom surface of the fixed horizontal plate 20, and a moving arm 512 is mounted on the bottom of the rotating gimbal 511. The end of the moving arm 512 away from the rotating gimbal 511 is rotatably connected to the moving arm 515, and the end of the moving arm 515 away from the moving arm 512 is rotatably connected to the moving connecting arm 516. One end of the air supply pipe 513 is fixedly connected to the rotating gimbal 511 through a valve, and the other end is fixedly connected to the air source element 514 mounted on the moving arm 512 through a valve. The actuator assembly includes a rear fixed plate 521, a front fixed plate 522, a fixed coupling shaft 523, a movable slide 524, a ball screw 525, a connecting rod 526, a rotating block 527, and a three-jaw claw 528 (as shown). Figure 5 (As shown); the rear fixing plate 521 is mounted on the motion connecting arm 516, and the front fixing plate 522 is mounted on the side of the rear fixing plate 521 away from the motion connecting arm 516 and is parallel to the rear fixing plate 521. Multiple fixed connecting shafts 523 are evenly arranged between the front fixing plate 522 and the rear fixing plate 521 and around their outer perimeter (the fixed connecting shafts 523 connect the front fixing plate 522 and the rear fixing plate 521; the number of fixed connecting shafts 523 is set according to actual conditions, but not less than two; in this embodiment, three are set). A rotating ball screw 525 is arranged in the middle of the front fixing plate 522 and the rear fixing plate 521 (the ball screw 525 is controlled to rotate by a drive motor located at one end of the rear fixing plate 521). The movable slide 524 is positioned between the front fixed plate 522 and the rear fixed plate 521, and is respectively penetrated by the fixed connecting shaft 523 and the ball screw 525. The movable slide 524 is slidably connected to the fixed connecting shaft 523 and threadedly connected to the ball screw 525. Three rotating blocks 527 are evenly arranged on the side of the front fixed plate 522 away from the fixed connecting shaft 523. The rotating blocks 527 are rotatably connected to the front fixed plate 522. The movable slide 524 is provided with three connecting rods 526 corresponding to the rotating blocks, with one end of the connecting rod 526 rotatably connected to the movable slide 524 and the other end rotatably connected to the rotating block 527. The three-jaw claws 528 are provided corresponding to the rotating blocks 527, with one end connected to the rotating block 527. Each claw of the three-jaw claws 528 has a protruding arc-shaped curved surface on its inner side (e.g., Figure 5 (As shown).

[0115] The front end of the fixed horizontal plate 20 (i.e. Figure 1 The right side (as shown) is equipped with a crop-lifting mechanism, which includes a top ring 61, a hinged connecting rod 62, a lifting ring fixing disc 63, a fixed cylinder 64, a bottom disc 65, a rotating screw 66, a lifting ring 67, and a lever 68. The top ring 61 is located on the front end face of the fixed horizontal plate 20. One end of the hinged connecting rod 62 is rotatably connected to the top ring 61, and the other end is rotatably connected to the lifting ring fixing disc 63 (as shown on the right side). Figure 1(As shown); the fixing cylinder 64 is coaxially mounted on the bottom surface of the lifting ring fixing disk 63, and a coaxial bottom disk 65 is mounted on the end of the fixing cylinder 64 away from the bottom surface of the lifting ring fixing disk 63 (as shown). Figure 7 (As shown); The rotating screw 66 is coaxially arranged in the inner cavity of the fixed cylinder 64, and the upper end of the rotating screw 66 passes through the lifting ring fixing disc 63 and is rotatably connected; The lifting ring 67 is sleeved on the outer wall of the fixed cylinder 64 (the inner wall of the lifting ring 67 is slidably connected to the outer wall of the fixed cylinder 64) and its inner wall is threadedly connected to the rotating screw 66 arranged in the inner cavity of the fixed cylinder 64 (for example: a lifting block is set on the outer wall of the rotating screw 66 and located on the inner wall of the fixed cylinder 64, the middle of the lifting block is passed through by the rotating screw 66 and threadedly connected, the outer wall of the lifting block is slidably connected to the inner wall of the fixed cylinder 64, and the side wall of the fixed cylinder 64 has no less than two vertical sliding grooves, which are evenly distributed around the central axis of the fixed cylinder 64. The inner wall of the lifting ring 67 and the outer wall of the lifting block are connected by a connecting slider, and the connecting slider is set corresponding to the vertical sliding groove); The lifting ring 67 is symmetrically provided with levers 68 on both sides (e.g. Figure 7 As shown, lever 68 consists of a slanted rod and a flat rod, with the slanted rod used to connect the lifting ring 67 and the flat rod.

[0116] Example 2:

[0117] A cotton picking robot navigation method, employing the cotton picking robot as described in Example 1, includes:

[0118] Step S1: Adjusting the spatial height: When the picking robot reaches the pre-picking position, the spatial height of the picking robot is adjusted in real time through the multi-stage lifting trellis mechanism according to the growth of the long-staple cotton plants.

[0119] Specifically, when the long-staple cotton plant is relatively tall and requires increased height, the first lifting component of the multi-stage lifting truss mechanism (i.e., the first cylinder 121 and the first piston rod 122) is activated. Under the action of the interaction force, it pushes the chassis beam 15 to lift the positioning plate 11, thereby causing the fixed horizontal plate 20, the connecting collection box 40, the two sets of end actuators, and the lifting mechanism to be lifted synchronously. If the lifting height is insufficient, the second lifting component (i.e., the second cylinder 131 and the second piston rod 132) is activated to further lift the fixed horizontal plate 20, causing the two sets of end actuators and the lifting mechanism to be lifted synchronously. During this process, the positioning support component (i.e., the sliding guide cylinder 141 and the guide pillar 142) extends and retracts as the second lifting component is lifted, thereby playing a role in fixing support and guiding. Ultimately, the overall height of the harvesting robot can meet the spatial height requirements for harvesting long-staple cotton bolls. When a descent is required, the second lifting assembly first initiates the descent, causing the fixed horizontal plate 20 to move downwards, which in turn causes the end effector and the lifting mechanism to move downwards synchronously. The first-stage lifting mechanism then initiates the descent, driving the positioning plate 11 downwards, which in turn causes the fixed horizontal plate 20, the connecting collection box 40, the two sets of end effectors and the lifting mechanism to move downwards synchronously.

[0120] Step S2, Envelope Picking and Collection: The robotic arm assembly is activated, causing the actuator assembly to move to the target picking point. The actuator assembly envelops the cotton bolls of the long-staple cotton, and the robotic arm assembly moves to pick the cotton bolls and place them in the connecting collection box. Then the end effector resets, completing one picking cycle.

[0121] Specifically, when the harvesting robot reaches the pre-harvesting position (the harvesting position can be determined by conventional image recognition, which can be directly adopted using conventional methods in this field, and will not be specifically discussed in this embodiment), the pneumatic-electric hybrid drive robotic arm assembly is activated. The mechanical arm 512 and the moving arm 515, driven by the motor, move the actuator assembly to a position 3-5 cm away from the upper part of the long-staple cotton boll. Then, the moving connecting arm 516 is driven to adjust the angle between the actuator assembly and the branches on the long-staple cotton boll to be parallel or not exceeding 30°. After that, the robotic arm assembly continues to drive the actuator assembly to move towards the root of the long-staple cotton boll until the arc-shaped protrusion at the front end of the claw of the actuator assembly can contact the gap between the roots of the cotton boll petals, at which point the movement of the robotic arm assembly stops. At this time, the drive motor starts to rotate forward, the ball screw 525 rotates, and the moving slide 524 moves forward. Through the connecting rod 526, it pushes the rotating block 527 to rotate towards the axis, thereby driving the three-jaw claw 528 to envelop the long-staple cotton bolls. Then, the air source element 514 starts, driving the rotating gimbal 511 pneumatically to rotate the actuator assembly and the robotic arm assembly above the connecting collection box 40. The drive motor of the actuator assembly reverses, the ball screw 525 rotates in the opposite direction, and the moving slide 524 moves backward. Through the connecting rod 526, it pushes the rotating block 527 to rotate away from the axis, thereby driving the three-jaw claw 528 to open and collect the long-staple cotton bolls in the connecting collection box 40. The rotating gimbal 511 resets, completing one long-staple cotton boll enveloping and harvesting process.

[0122] Step S3, Supporting Loose Stalks: Before the harvesting robot moves to the next harvestable area, it checks whether there are any lodged stalks. If so, the support mechanism is activated to support the lodged long-staple cotton stalks. The support mechanism can be disassembled depending on the actual working conditions. When the entire cotton field is growing well and there are no obviously lodged long-staple cotton plants, the support mechanism is not required. When there are many lodged long-staple cotton plants in the entire cotton field, the support mechanism needs to be installed.

[0123] The specific working principle of the cotton boll-lifting mechanism is as follows: When preparing to move to the next harvestable space, if the long-staple cotton stalks are lodged in the next harvestable space, the cotton boll-lifting mechanism is activated. The bottom disc 65 is moved into the cotton field and placed on the cotton field soil corresponding to the lodged long-staple cotton stalks by the up-and-down movement of the hinged connecting rod 62. The screw 66 is rotated in the forward direction, causing the lifting ring 67 to move upward on the outer wall of the fixed cylinder 64, which in turn drives the levers 68 on both sides to move upward, peeling off the lodged long-staple cotton stalks. Afterward, the hinged connecting rod 62 drives the entire mechanism consisting of the fixed disc 63, the fixed cylinder 64, the bottom disc 65, the rotating screw 66, the lifting ring 67, and the levers 68 to lift upward, thereby lifting the lodged long-staple cotton stalks and assisting in the harvesting of the cotton bolls from the lodged long-staple cotton plants. After the lodging is finished, rotate screw 66 in the opposite direction to move lifting ring 67 downward, which in turn moves levers 68 on both sides downward to the lowest point of the stroke. The lifting ring end face at the top of hinged connecting rod 62 and lifting ring fixing disc 63 swings up and down, returning the lodged long-staple cotton stalks to their original position. The stalk lifting mechanism resets, completing one stalk lifting process, and proceeding to the next harvestable area.

[0124] Step S4, Field Ridge Finding Technology: Automatic ridge finding technology is used to enable the harvesting robot to find ridges in the cotton field and move the harvesting robot to the next harvestable space;

[0125] Specifically:

[0126] Step S41: Acquire images of the cotton field using a high-resolution camera to obtain the dataset;

[0127] Step S42: First, perform grayscale processing on the collected images to obtain grayscale images:

[0128]

[0129] In the formula: R, G, and B represent the pixel values ​​of the red, green, and blue channels, respectively;

[0130] Then, the image is filtered to obtain a smooth image:

[0131]

[0132] In the formula: Represents the Gaussian kernel function. Indicates standard deviation;

[0133] Step S43: Perform edge detection to obtain the ridge edges in the image; use the Sobel operator to obtain the gradient of the image, and then obtain the gradient magnitude and gradient direction:

[0134] Gradient magnitude:

[0135]

[0136]

[0137] Gradient direction:

[0138]

[0139] By retaining edge points with the largest gradient magnitude and refining edges, the computational cost of edge detection is reduced, resulting in clearer edges. Furthermore, a high-low threshold method is used to filter edge points (i.e., setting a high threshold). T h With low threshold T l For the high threshold portion (i.e. edge points not less than the high threshold), strong edge points are retained, while for the low threshold portion (i.e. edge points located between the high and low thresholds), weak edge points connected to the strong edge points are retained, resulting in clearer and more realistic ridge line edges.

[0140] Step S44: First, extract the ridge lines in the image using the Hough transform method and determine the ridge line parameters:

[0141]

[0142] In the formula, This represents the distance from the straight line to the origin; Angle representing a straight line; x,y () represents the pixel coordinates;

[0143] Then, the least squares method is used to fit a straight line to the detected edge points, and the error is minimized. The position of the ridge line is described by the fitted straight line:

[0144]

[0145] In the formula: ( x i ,y i () represents the pixel coordinates of the edge point;

[0146] Step S45: Track the ridge line by measuring the lateral deviation and heading angle deviation between the harvesting robot and the target ridge line.

[0147]

[0148] In the formula: Y 1 indicates the current position of the harvesting robot. kx + b Represent the target ridge line equation; Indicates the current heading angle. The direction angle corresponding to the slope of the target ridge line; e Y Indicates heading deviation. Indicates the deviation of the heading angle;

[0149] Control commands are generated based on lateral and heading angle deviations. Turning commands are then generated based on the current position and target point of the harvesting robot. The robot's direction is adjusted using feedback and pre-aiming points.

[0150] Select a future point on the target path. x f ,y f ), as a pre-aiming point:

[0151]

[0152] In the formula: L d Indicates the aiming distance, i.e., the current point ( x,y The straight-line distance between the target and the aiming point;

[0153] The PID controller compensates for the heading angle deviation and generates the steering angle command. :

[0154]

[0155] In the formula: This represents the weighting coefficient, which takes the value [0, 1]. K p , K i , K d These represent the proportional, integral, and differential parameters, respectively.

[0156] Speed ​​adjustment command:

[0157]

[0158] In the formula: v 0 represents the base speed. K v This indicates the speed adjustment coefficient.

[0159] Step S5, Field Turning: After completing the task of one row, the robot turns in the narrow space of the field through four sets of steering and walking mechanisms. When the long-staple cotton picking robot needs to turn at the edge of the cotton field, the four sets of four-wheel steering mechanisms are activated. Since the steering method of the walking wheels 34 of the four sets of steering and walking mechanisms is the same, and the front and rear wheels turn in the same direction, the turning radius in the narrow space of the cotton field is reduced. After the field turn is completed, the steering bracket 32 ​​drives the connecting shaft 33 and the walking wheels 34 to rotate in opposite directions. The return speed of the rear wheel is 15% faster than that of the front wheel, which can quickly complete the return and reset of the walking wheels 34 and enter the stage of straight-line driving of the walking wheels 34, completing one turning process at the edge of the cotton field. The robot adopts an "S" shaped driving route to complete the field turn and self-propelled movement in the entire long-staple cotton field.

[0160] Step S6, Cotton Field Navigation Method: By combining navigation methods, the harvesting robot can achieve efficient autonomous navigation in different environments;

[0161] Specifically:

[0162] Step S61: Initialize various navigation parameters, sampling frequency, and cutoff frequency; construct prediction models for position, velocity, and attitude;

[0163] By using gyroscope data to predict the attitude at the next moment, an attitude prediction model can be constructed:

[0164]

[0165] In the formula: Represents the predicted attitude quaternion (i.e. k (Predicted value at time). The attitude quaternion represents the previous time step; Indicates the time step (unit: s); This indicates the angular velocity measured by the gyroscope (unit: rad / s). Indicates zero bias of the gyroscope (unit: rad / s); This represents the quaternion multiplication operator; exp This represents a minute rotation, that is, converting angular velocity into an increment of quaternion;

[0166] Accelerometer data is used to predict the velocity at the next moment, thereby constructing a velocity prediction model:

[0167]

[0168] In the formula: Indicates the speed of prediction (i.e.) k (Predicted value at time) Represents the velocity at the previous moment. Rotation matrix corresponding to attitude quaternion; This indicates the value measured by the accelerometer; Indicates zero bias of the accelerometer (unit: m / s²); g Represents the gravitational acceleration vector;

[0169] By using velocity integrals to predict the position at the next moment, a position prediction model can be constructed:

[0170]

[0171] In the formula: Indicates the predicted location ( k (Predicted value at time) It indicates the position at the previous moment.

[0172] Step S62: Obtain the covariance of the prediction error by calculating the prediction model. First, calculate the state transition matrix. F k :

[0173]

[0174] In the formula: I Represents the identity matrix; The cross product matrix representing the angular velocity of the gyroscope; The cross product matrix representing accelerometer measurements; R(q) This represents the rotation matrix corresponding to the attitude quaternion; This represents the time difference between the current moment and the next predicted moment.

[0175] Then, the process noise covariance is calculated. Q k :

[0176]

[0177] In the formula: Q ins Represents the inertial navigation noise covariance matrix; Q vio Represents the visual navigation noise covariance matrix; diag () This means arranging the input submatrices along the main diagonal to form a block diagonal matrix;

[0178] Finally, covariance prediction calculations are performed:

[0179]

[0180] In the formula: This represents the error covariance matrix of the prediction; This represents the error covariance matrix of the previous time step;

[0181] Step S63: Construct an observation model using Global Navigation Satellite System (GNSS) observations and visual observations:

[0182] Global Navigation Satellite System (GNSS) observation model:

[0183]

[0184] In the formula: Z GNSS This represents GNSS observations, which include a combination of position and velocity information; P GNSS This represents the three-dimensional position coordinates obtained through a GNSS sensor; V GNSS This represents the three-dimensional velocity coordinates obtained through a GNSS sensor; P INS This indicates the position predicted by the inertial navigation system through IMU integration; V INS This represents the velocity predicted by the inertial navigation system through IMU integration; This represents the position error, specifically the deviation between the GNSS observations and the INS predictions. This represents the velocity error, specifically the velocity deviation between GNSS observations and INS predictions. This indicates the measurement noise within a satellite navigation GNSS system.

[0185] in:

[0186]

[0187] In the formula: and These represent the position and velocity observations from GNSS satellite navigation, respectively. and These represent the position prediction and velocity prediction values ​​of the inertial navigation system (INS), respectively.

[0188] The observation matrix of the GNSS observation model is:

[0189]

[0190] Visual observation model: Visual observation uses a method to calculate reprojection error, the first... i The observed coordinates of the feature points are Then its projection model is:

[0191]

[0192] In the formula: Xi This represents the coordinates of the feature point in the global coordinate system; () represents the camera projection function; R VIO The rotation matrix represents visual navigation; p VIO Indicates the location of visual navigation; v vis Indicates visual observation noise;

[0193] The observation matrix of the visual observation model is:

[0194]

[0195] in:

[0196]

[0197] In the formula: R Represents the rotation matrix; Indicates the attitude error of inertial navigation; Indicates the attitude error in visual navigation; This indicates the positional error in visual navigation;

[0198] Kalman gain is used to dynamically adjust the weights between the predicted values ​​of the prediction model and the dynamic values ​​of the observation model to optimize the accuracy of state estimation.

[0199]

[0200]

[0201] In the formula: K k Represents the Kalman gain matrix. H k This represents the observation matrix obtained by the observation model, i.e., the fused observation source; R k Represents the observation noise covariance matrix;

[0202] Step S64: Update the attitude, position, velocity, and visual error states respectively to obtain the corrected states; specifically:

[0203] Attitude correction:

[0204]

[0205] In the formula: express k The attitude error is updated in real time; it is achieved through... k At that moment, the attitude values ​​obtained from GNSS observations and INS predictions are obtained, i.e. ;

[0206] Position correction:

[0207]

[0208] In the formula: express k The position error is updated in real time; it is achieved through... k The position deviation between the satellite navigation GNSS observation and the inertial navigation INS prediction at a given time is obtained;

[0209] Speed ​​correction:

[0210]

[0211] In the formula: express k The speed error is updated in real time; it is achieved through... k The velocity deviation between the satellite navigation GNSS observation and the inertial navigation INS prediction at a given time is obtained.

[0212] Visual error correction:

[0213]

[0214] In the formula: express k Constantly update the visual positional error, i.e. (Using visual ranging methods, a camera captures scene images, feature points are extracted from the captured images, and the camera's position in the scene is calculated by matching these feature points and using triangulation.) Using the accelerometer and gyroscope of the IMU, the predicted position is obtained by integrating acceleration and angular velocity. (Finally, the position error is obtained). express k Continuously update the visual pose error, i.e. (Using feature matching in the vision system, the pose is estimated from the feature points using a rotation matrix to obtain the observed pose at this time.) The predicted attitude can be directly calculated using a gyroscope. (Finally, the attitude error is obtained). express k The position of vision at time -1; express k Visual posture at time -1; express k The constantly updated visual position; express k The visual posture is constantly updated.

[0215] Step S65: First, an inertial measurement unit (IMU) is used to obtain a prediction model of attitude, position, and velocity during the prediction phase (i.e., step S61), and simultaneously, the covariance of the corresponding prediction error is obtained (i.e., step S62). Then, the visual observation model is used for observation (i.e., step S63), and visual updates and continuous state corrections are performed (i.e., step S64), and the covariance is updated. When the Global Navigation Satellite System (GNSS) data is available (i.e., the GNSS module receives enough satellite signals to accurately calculate the position information), the position and velocity residuals are calculated (i.e., ... and Then, the error state of the Global Navigation Satellite System (GNSS) observation model is updated (i.e., step S64 is used), and the covariance is updated again; the above prediction update steps are repeated to continuously fuse multi-sensor data and improve navigation and positioning accuracy;

[0216] The specific steps for updating the covariance are as follows:

[0217]

[0218] In the formula: This represents the updated error covariance matrix.

Claims

1. A picking navigation method of a cotton picking robot, characterized by: The cotton picking robot comprises two groups of multi-stage lifting frame mechanisms, a fixed horizontal plate, a steering walking mechanism, a connecting collecting box, two groups of end execution mechanisms and a supporting mechanism, the two groups of multi-stage lifting frame mechanisms are symmetrically arranged at the two ends of the lower side of the fixed horizontal plate and are connected with the bottom surface of the fixed horizontal plate, the steering walking mechanism is arranged at the bottom of the multi-stage lifting frame mechanism, and two groups of steering walking mechanisms are arranged at the bottom of one group of multi-stage lifting frame mechanisms; the connecting collecting box is arranged between the two groups of multi-stage lifting frame mechanisms and is fixedly connected with the inner sides of the corresponding multi-stage lifting frame mechanisms at the two ends of the connecting collecting box, and the connecting collecting box is located at the rear end of the fixed horizontal plate; the two groups of end execution mechanisms are symmetrically arranged at the middle part of the bottom surface of the fixed horizontal plate and are located on the inner sides of the two groups of multi-stage lifting frame mechanisms, and the front end of the fixed horizontal plate is provided with the supporting mechanism; the end execution mechanism comprises a mechanical arm assembly and an executor assembly; The specific picking navigation method comprises: Step S1, adjusting the spatial height: when the picking robot reaches the pre-picking position, the spatial height of the picking robot is adjusted in real time through the multi-stage lifting frame mechanism according to the growth conditions of the long-staple cotton plants; Step S2, envelope picking and collecting: the mechanical arm assembly is started, the executor assembly is moved to the target picking point position, the executor assembly envelopes the cotton bolls of the long-staple cotton, the mechanical arm assembly moves to pick the cotton bolls and place them in the connecting collecting box, then the end execution mechanism is reset, and one picking is completed; Step S3, supporting the fallen stalks: before the picking robot moves to the next pickable space, it is judged whether the stalks are fallen, if so, the supporting mechanism is started to support the fallen long-staple cotton stalks; Step S4, field ridge finding technology: the automatic ridge finding technology is adopted to realize the ridge finding and self-walking of the picking robot in the cotton field, and the picking robot is moved to the next pickable space; specifically: Step S41, the image of the cotton field is collected through a high-resolution camera to realize the acquisition of a data set; Step S42, the collected image is first processed to obtain a gray image, and then the image is filtered to obtain a smooth image; Step S43, edge detection is performed to obtain the ridge line edge in the image; Step S44, first, the ridge line in the image is extracted by the Hough transform method to determine the ridge line parameters: x, y; then, the least square method is used to fit a straight line to the detected edge points, and the minimum error is minimized, and the position of the ridge line is described by the fitting straight line: x, y; then, the lateral deviation and the heading angle deviation of the picking robot and the target ridge line are used to track the ridge line: kx; according to the lateral deviation and the heading angle deviation, a control instruction is generated, the current position of the picking robot and the target point generate a steering instruction, and the driving direction of the picking robot is adjusted through feedback and preview point: x, y; speed adjustment instruction: wherein represents the distance between the straight line and the origin; represents the angle of the straight line; Step S5, field head turning: after the work task of one ridge is completed, the four groups of steering walking mechanisms are used to realize the turning in the narrow space in the field; represents the pixel point coordinate; Step S6, cotton field navigation mode: through the combined navigation mode, the efficient autonomous navigation of the picking robot in different environments is realized. wherein: x i ,y i represents the edge point pixel coordinate; ​ In the formula: Y 1 represents the current position of the picking robot, ​ + b represents the target row line equation; represents the current heading angle, the direction angle corresponding to the target row line slope; e Y represents the heading deviation, represents the heading angle deviation; ​ Select a future point on the target path. x f ,y f ), as a pre-aiming point: In the formula: L d Indicates the aiming distance, i.e., the current point ( ​ The straight-line distance between the target and the aiming point; Compensate the heading angle deviation by the PID controller to generate the steering angle command : In the formula: represents a weight coefficient; K p , K i , K d respectively represent proportional, integral, and derivative parameters; ​ In the formulae: v 0 denotes a reference speed, K v denotes a speed adjustment coefficient; ​ ​ 2. The picking navigation method of the cotton picking robot according to claim 1, characterized in that: The multi-stage lifting frame mechanism comprises a positioning plate, two groups of first lifting assemblies, a second lifting assembly, two groups of positioning support assemblies and a chassis beam; the two groups of first lifting assemblies are symmetrically arranged on the inner wall of the positioning plate, and each first lifting assembly comprises a first cylinder and a first piston rod; the first cylinder is arranged on the inner wall of the positioning plate and is slidably connected to the first piston rod at the lower end thereof; the chassis beam is arranged on the lower side of the corresponding positioning plate, and the end of the first piston rod away from the first cylinder is fixedly connected to the end face of the chassis beam; the second lifting assembly is arranged in the middle of the inner wall of the positioning plate and comprises a second cylinder and a second piston rod; the second cylinder is arranged on the inner wall of the positioning plate and is slidably connected to the second piston rod at the upper end thereof; the end of the second piston rod away from the second cylinder is fixedly connected to the bottom surface of the fixed cross plate; the positioning support assemblies are arranged between the second lifting assembly and the two groups of first lifting assemblies; each positioning support assembly comprises a sliding guide cylinder and a guide support column; the sliding guide cylinder is arranged on the inner wall of the positioning plate and has the guide support column slidably arranged at the upper end thereof; and the end of the guide support column away from the sliding guide cylinder is fixedly connected to the bottom surface of the fixed cross plate.

3. The picking navigation method of a cotton picking robot according to claim 2, characterized in that: The turning walking mechanism comprises a wheel fork, a turning support, a connecting shaft and a walking wheel; the wheel fork is fixedly connected to the bottom surface of the chassis beam and has a Z-shaped structure in the longitudinal section; the turning support is rotatably arranged at the bottom of the wheel fork; and the connecting shaft is rotatably arranged in the middle of the turning support and has the walking wheel fixedly sleeved on the outer wall thereof.

4. The picking navigation method of a cotton picking robot according to claim 2 or 3, characterized in that: The mechanical arm assembly comprises a rotating holder, a moving large arm, a gas supply pipe, a gas source element, a moving small arm and a moving connecting arm; the rotating holder is arranged on the bottom surface of the fixed cross plate and has the moving large arm rotatably arranged at the bottom thereof; the end of the moving large arm away from the rotating holder is rotatably connected to the moving small arm; the end of the moving small arm away from the moving large arm is rotatably connected to the moving connecting arm; one end of the gas supply pipe is fixedly connected to the rotating holder through a valve, and the other end of the gas supply pipe is fixedly connected to the gas source element arranged on the moving large arm through a valve; the actuator assembly comprises a rear fixed plate, a front fixed plate, fixed connecting shafts, a moving slide, a ball screw, connecting rotating rods, rotating blocks and three-jaw fingers; the rear fixed plate is arranged on the moving connecting arm; the front fixed plate is arranged on the side of the rear fixed plate away from the moving connecting arm and is arranged in parallel with the rear fixed plate; a plurality of fixed connecting shafts are uniformly arranged between the front fixed plate and the rear fixed plate and at the outer periphery of the front fixed plate and the rear fixed plate; the ball screw is rotatably arranged in the middle of the front fixed plate and the rear fixed plate; the moving slide is arranged between the front fixed plate and the rear fixed plate and is penetrated by the fixed connecting shafts and the ball screw; the moving slide is slidably connected to the fixed connecting shafts and is threadedly connected to the ball screw; three rotating blocks are uniformly arranged on the side surface of the front fixed plate away from the fixed connecting shafts; the rotating blocks are rotatably connected to the front fixed plate; the moving slide is provided with three connecting rotating rods corresponding to the rotating blocks; one end of each connecting rotating rod is rotatably connected to the moving slide, and the other end of each connecting rotating rod is rotatably connected to the rotating block; and the three-jaw fingers are arranged corresponding to the rotating blocks and have one end connected to the rotating block; and each jaw of the three-jaw fingers has a convex arc-shaped curved surface arranged on the inner side thereof.

5. The picking navigation method of a cotton picking robot according to claim 2 or 3, characterized in that: The supporting mechanism comprises a top ring, a hinged connecting rod, a lifting ring fixing disc, a fixing cylinder, a bottom disc, a rotating screw rod, a lifting ring and a push rod, the top ring is arranged at the front end surface of the fixing horizontal plate, one end of the hinged connecting rod is rotationally connected with the top ring, and the other end is rotationally connected with the lifting ring fixing disc; the fixing cylinder is coaxially arranged at the bottom surface of the lifting ring fixing disc, and a bottom disc is coaxially arranged at the end of the fixing cylinder away from the bottom surface of the lifting ring fixing disc; the rotating screw rod is coaxially arranged in the inner cavity of the fixing cylinder, and the upper end of the rotating screw rod penetrates through the lifting ring fixing disc and is rotationally connected; the lifting ring is sleeved on the outer wall of the fixing cylinder, and the inner wall thereof is threadedly connected with the rotating screw rod arranged in the inner cavity of the fixing cylinder; the push rods are symmetrically arranged at the two sides of the lifting ring.

6. The picking navigation method of a cotton picking robot according to claim 1, characterized in that: The step S6 is specifically: Step S61, initializing each navigation parameter and sampling frequency, cutoff frequency; constructing a prediction model of position, velocity and attitude; Step S62, obtaining the covariance of the prediction error by calculation of the prediction model, first, calculate the state transition matrix F k : In the formula: I denotes a unit matrix; denotes a cross product matrix of the angular velocity of the gyroscope; denotes a cross product matrix of the accelerometer measurement value; R(q) denotes a rotation matrix corresponding to the attitude quaternion; denotes a time difference between the current time and the next predicted time; Then, the process noise covariance is calculated Q k : wherein: Q ins represents the inertial navigation noise covariance matrix; Q vio represents the visual navigation noise covariance matrix; diag() represents arranging the input sub-matrices along the main diagonal to form a block diagonal matrix; Finally, covariance prediction calculation is performed: In the formula: represents the predicted error covariance matrix; represents the error covariance matrix at the previous time Step S63, constructing an observation model through global navigation satellite system observation and visual observation: Global navigation satellite system observation model: wherein: Z GNSS represents the observation of the satellite navigation GNSS, containing a combination of position and velocity information; P GNSS represents the three-dimensional position coordinates obtained by the GNSS sensor; V GNSS represents the three-dimensional velocity coordinates obtained by the GNSS sensor; P INS represents the position obtained by the inertial navigation system through IMU integration prediction; V INS represents the velocity obtained by the inertial navigation system through IMU integration prediction; represents the position error, i.e. the position deviation between the satellite navigation GNSS observation and the inertial navigation INS prediction; represents the velocity error, i.e. the velocity deviation between the satellite navigation GNSS observation and the inertial navigation INS prediction; represents the measurement noise inside the satellite navigation GNSS system; The observation matrix of the GNSS observation model is: Visual observation model: Visual observation adopts the method of calculating re-projection error, the observation coordinate of the first i characteristic point is The projection model is: wherein: X i denotes the coordinates of the feature point in the global coordinate system; () denotes the camera projection function; R VIO denotes the rotation matrix of the visual navigation; p VIO denotes the position of the visual navigation; v vis denotes the visual observation noise; The observation matrix of the visual observation model is: Wherein: wherein: R represents a rotation matrix; represents an inertial navigation attitude error; represents a vision navigation attitude error; represents a vision navigation position error; The Kalman gain is used to dynamically adjust the weight between the predicted value of the prediction model and the dynamic value of the observation model to optimize the accuracy of state estimation: wherein: K k denotes the Kalman gain matrix, H k denotes the observation matrix obtained from the observation model, i.e. the fusion of the observation sources; R k denotes the observation noise covariance matrix; Step S64, respectively updating the attitude, position, velocity and visual error state to obtain the corrected state; Step S65, first, the inertial measurement unit is used to obtain the prediction model of attitude, position and velocity in the prediction stage, and the corresponding prediction error covariance is obtained; then, the visual observation model is observed, and the visual update is performed to continuously correct the state, and the covariance is updated; when the global navigation satellite system data reaches, the position and velocity residual error is calculated, and the error state of the global navigation satellite system observation model is updated, and the covariance is updated again; the above prediction update steps are repeated to continuously fuse the multi-sensor data and improve the navigation positioning accuracy.

7. The picking navigation method of a cotton picking robot according to claim 6, characterized in that: The covariance updating in the step S65 is specifically: In the formula, denotes the updated error covariance matrix.

Images