Cotton picking robot and picking navigation method thereof

Abstract

The invention provides a cotton picking robot, and relates to the field of intelligent picking, the cotton picking robot comprises two groups of multi-stage lifting truss mechanisms, a fixed transverse plate (20), a steering walking mechanism, a connection collection box (40), two groups of end execution mechanisms and a grain lifting mechanism, the two groups of multi-stage lifting truss mechanisms are symmetrically arranged at two ends of the lower side of the fixed transverse plate (20) and are connected with the bottom surface of the fixed transverse plate (20); the steering walking mechanism is arranged at the bottom of the multi-stage lifting truss mechanisms, the connecting collecting box (40) is arranged between the two multi-stage lifting truss mechanisms, the two tail end executing mechanisms are symmetrically arranged in the middle of the bottom face of the fixing transverse plate (20), and the grain lifting mechanism is arranged at the front end of the fixing transverse plate (20). According to the picking robot, cotton can be efficiently, accurately and completely picked, and the problems that in the picking process, cotton fibers are left at the roots of cotton bolls, and lodging stalks are rolled are solved.

Description

Technical Field

[0001] The present invention relates to the technical field of intelligent picking, and particularly relates to a cotton picking robot and a picking navigation method thereof. Background Art

[0002] Cotton is an important economic product globally. Currently, over 90% of cotton is used for clothing, home textiles, and industrial fabrics. Cotton includes long-staple cotton, fine-staple cotton, and coarse-staple cotton according to fiber types. Among them, long-staple cotton has long fibers and good fiber flexibility, and is known as the "gold" in cotton. Currently, the picking of long-staple cotton mainly relies on mechanical picking. However, the initial node of the fruiting branch of long-staple cotton (i.e., the node where the first flowering and fruiting branch on the main stem of cotton appears) is relatively short. During mechanical picking, it is easy to miss the cotton bolls at the bottom, which not only leads to incomplete picking and affects cotton yield, but also easily causes the breeding of pests and diseases and affects subsequent cotton planting. At the same time, the plant height of long-staple cotton plants is relatively high and prone to lodging problems, which not only affects the picking efficiency and increases the picking difficulty, but also easily causes the picking machinery to roll over the long-staple cotton stems, reducing the future cotton yield and quality. In addition, the long-staple cotton plants grow outward, resulting in the phenomenon of the rows closing together. Coupled with the complex field environment, it is difficult for cotton picking machinery to accurately walk freely in the cotton field, easily causing damage to long-staple cotton plants, and then leading to problems such as cotton yield reduction and picking machinery jamming. Summary of the Invention

[0003] Aiming at the above problems existing in the prior art, the purpose of the present invention is to provide a cotton picking robot, which can not only achieve efficient, accurate, and complete picking of cotton, avoid the problem of cotton floss remaining at the root of the cotton bolls during the picking process, but also lift the lodging stems during the picking process and avoid rolling over the lodging stems, effectively protecting the cotton plants during the picking process.

[0004] Another purpose of the present invention is to provide a picking navigation method for a cotton picking robot, which can not only ensure accurate and efficient cotton picking, but also enable the picking robot to walk in rows automatically in the cotton field, improving the picking efficiency and ensuring the cotton yield and quality of the picking.

[0005] The purpose of the present invention is achieved through the following technical solutions: A cotton picking robot, comprising two sets of multi-stage lifting truss mechanisms, a fixed cross plate, a steering and traveling mechanism, a connecting collection box, two sets of end effectors and a crop lifter. The two sets of multi-stage lifting truss mechanisms are symmetrically arranged at both ends of the lower side of the fixed cross plate and connected to the bottom surface of the fixed cross plate. The steering and traveling mechanism is arranged at the bottom of the multi-stage lifting truss mechanism, and two sets of steering and traveling mechanisms are arranged at the bottom of one set of multi-stage lifting truss mechanism; the connecting collection box is arranged between the two sets of multi-stage lifting truss mechanisms, and both ends of the connecting collection box are respectively fixedly connected to the inner sides of the corresponding multi-stage lifting truss mechanisms, and the connecting collection box is located at the rear end of the fixed cross plate; the two sets of end effectors are symmetrically arranged in the middle of the bottom surface of the fixed cross plate and are located inside the two sets of multi-stage lifting truss mechanisms, and a crop lifter is arranged at the front end of the fixed cross plate.

[0006] Based on the further optimization of the above solution, the multi-stage lifting truss mechanism includes a positioning plate, two sets of first lifting components, a second lifting component, two sets of positioning and supporting components and a chassis cross beam; 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 barrel and a first piston rod. The first cylinder barrel is arranged on the inner wall of the positioning plate and its lower end is slidably connected to the first piston rod. The chassis cross 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 barrel is fixedly connected to the end face of the chassis cross beam; a second lifting component is arranged in the middle of the inner wall of the positioning plate, including a second cylinder barrel and a second piston rod. The second cylinder barrel is arranged 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 barrel is fixedly connected to the bottom surface of the fixed cross plate; positioning and supporting components are respectively arranged between the second lifting component and the two sets of first lifting components. The positioning and supporting component includes a sliding guide cylinder and a guide pillar. The sliding guide cylinder is arranged on the inner wall of the positioning plate and its upper end slidably arranges the guide pillar. The end of the guide pillar away from the sliding guide cylinder is fixedly connected to the bottom surface of the fixed cross plate.

[0007] Based on the further optimization of the above solution, the steering and traveling mechanism includes a wheel fork, a steering bracket, a connecting shaft and a traveling wheel. The wheel fork is fixedly connected to the bottom surface of the chassis cross beam, and the longitudinal section of the wheel fork is a Z-shaped structure. The steering bracket is rotatably arranged at the bottom of the wheel fork, and the connecting shaft is rotatably arranged in the middle of the steering bracket. The traveling wheel is fixedly sleeved on the outer wall of the connecting shaft.

[0008] Based on the further optimization of the above solution, the end effector includes a robotic arm assembly and an actuator assembly. The robotic arm assembly includes a rotating turntable, a moving large arm, an air supply pipe, an air source component, a moving small arm, and a moving connecting arm. The rotating turntable is arranged on the bottom surface of the fixed cross plate, and the moving large arm is arranged at the bottom of the rotating turntable. One end of the moving large arm away from the rotating turntable is rotatably connected to the moving small arm, and one end of the moving small arm away from the moving large arm is rotatably connected to the moving connecting arm. One end of the air supply pipe is fixedly connected to the rotating turntable through a valve, and the other end is fixedly connected to the air source component arranged on the moving large arm through a valve. The actuator assembly includes a rear fixing plate, a front fixing plate, a fixed connecting shaft, a moving slide, a ball screw, a connecting rotating rod, a rotating block, and three-jaw fingers. The rear fixing plate is arranged on the moving connecting arm, the front fixing plate is arranged on the side of the rear fixing plate away from the moving connecting arm and is parallel to the rear fixing plate. A plurality of fixed connecting shafts are evenly arranged between the front fixing plate and the rear fixing plate and on their outer circles (the front fixing plate and the rear fixing plate are connected by the fixed connecting shafts). A rotating ball screw is arranged in the middle of the front fixing plate and the rear fixing plate. The moving slide is arranged between the front fixing plate and the rear fixing plate and is penetrated by the fixed connecting shafts and the ball screw respectively. The moving slide is slidably connected to the fixed connecting shafts and is threadedly connected to the ball screw. Three rotating blocks are evenly arranged on the side surface of the front fixing plate away from the fixed connecting shafts. The rotating blocks are rotatably connected to the front fixing plate. The moving slide is provided with three connecting rotating rods corresponding to the rotating blocks, and one end of each connecting rotating rod is rotatably connected to the moving slide and the other end is rotatably connected to the rotating block. The three-jaw fingers are arranged corresponding to the rotating blocks and one end of each is connected to the rotating block. A convex arc surface is arranged on the inner side of each jaw finger of the three-jaw fingers.

[0009] Based on the further optimization of the above solution, the lifting and supporting mechanism includes a top ring, a hinged connecting rod, a hanging ring fixing disc, a fixed cylinder, a bottom disc, a rotating screw rod, a lifting ring, and a dial rod. The top ring is arranged on the front end face of the fixed cross plate. One end of the hinged connecting rod is rotatably connected to the top ring and the other end is rotatably connected to the hanging ring fixing disc. The fixed cylinder is coaxially arranged on the bottom surface of the hanging ring fixing disc, and a coaxial bottom disc is arranged at one end of the fixed cylinder away from the bottom surface of the hanging ring fixing disc. The rotating screw rod is coaxially arranged in the inner cavity of the fixed cylinder, and the upper end of the rotating screw rod penetrates through the hanging 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 rod arranged in the inner cavity of the fixed cylinder. The dial rods are symmetrically arranged on both sides of the lifting ring.

[0010] A picking navigation method for a cotton picking robot, using the above cotton picking robot, includes: Step S1, adjusting the space height: When the picking robot reaches the pre-picking position, according to the growth situation of the long-staple cotton plants, the space height of the picking robot is adjusted in real time through the multi-stage lifting truss mechanism; 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 envelopes the cotton bolls of the long-staple cotton, and the robotic arm assembly moves to pick the cotton bolls and place them in the connected collection box. Then, the end effector returns to its original position, completing one picking operation. Step S3, Lodged Stem Rejuvenation: Before the picking robot moves to the next pickable space, it determines whether there is a situation of lodged stems. If so, the rejuvenation mechanism is activated to lift the lodged long-staple cotton stems. Step S4, Field Ridging Detection Technology: The automatic ridging detection technology is used to enable the picking robot to detect ridges and move forward automatically in the cotton field, moving the picking robot to the next pickable space. Step S5, Turning at the Field Edge: After completing the operation task of one ridge, the picking robot turns in the narrow space of the field through four sets of steering and walking mechanisms. Step S6, Cotton Field Navigation Method: Through a combined navigation method, the picking robot realizes efficient autonomous navigation in different environments.

[0011] Based on the further optimization of the above solution, Step S4 is specifically as follows: Step S41, Acquire images of the cotton field through a high-resolution camera to obtain a data set. Step S42, First, perform grayscale processing on the collected images to obtain grayscale images, and then perform filtering processing on the images to obtain smooth images. Step S43, Perform edge detection to obtain the ridge line edges in the images. Step S44, First, extract the ridge lines in the images by the method of Hough transform to determine the ridge line parameters:

[0012] In the formula, represents the distance between the straight line and the origin; represents the angle of the straight line; ([[]] x, y ) represents the pixel point coordinates; Then, use the least squares method to perform linear fitting on the detected edge points and minimize the error, and describe the ridge line position through the fitted straight line:

[0013] In the formula: ([[]] x i ,y i ) represents the pixel coordinates of the edge points; Step S45, Perform ridge line tracking based on the lateral deviation and heading angle deviation between the picking robot and the target ridge line:

[0014] In the formula: Y 1 represents the current position of the picking robot, kx + b represents the target ridge line equation; represents the current heading angle, the direction angle corresponding to the slope of the target ridge line; e Y represents the heading deviation, represents the heading angle deviation; Generate a control command based on the lateral deviation and the heading angle deviation, generate a steering command from the current position of the picking robot and the target point, and adjust the driving direction of the picking robot through feedback and preview points: Select a future point on the target path ( x f ,y f ) as the preview point:

[0015] In the formula: L d represents the preview distance, that is, the straight-line distance between the current point ( x, y ) and the preview point; Compensate for the heading angle deviation through a PID controller to generate a steering angle command :

[0016] In the formula: represents the weight coefficient, and its value range is [0, 1]; K p , K i , K d respectively represent the proportional, integral, and differential parameters; Speed adjustment command:

[0017] In the formula: v 0 represents the reference speed, K v represents the speed adjustment coefficient.

[0018] Based on the further optimization of the above solution, the specific step S6 is as follows: Step S61, Initialize each navigation parameter, sampling frequency, and cut-off frequency; construct a prediction model for position, speed, and attitude; Step S62, Obtain the covariance of the prediction error through the calculation of the prediction model. First, calculate the state transition matrix F k:

[0019] Where: I represents the identity matrix; represents the cross product matrix of the gyroscope angular velocity; represents the cross product matrix of the accelerometer measurement value; R(q) represents the rotation matrix corresponding to the attitude quaternion; represents the time difference between the current moment and the next predicted moment; Then, calculate the process noise covariance Q k :

[0020] Where: Q ins represents the inertial navigation noise covariance matrix; Q vio represents the visual navigation noise covariance matrix; diag () represents arranging the input sub - matrix along the main diagonal to form a block - diagonal matrix; Finally, perform covariance prediction calculation:

[0021] Where: represents the predicted error covariance matrix; represents the error covariance matrix of the previous moment; Step S63, construct an observation model through Global Navigation Satellite System (GNSS) observations and visual observations: Global Navigation Satellite System (GNSS) observation model:

[0022] Where: Z GNSS represents the observation value of satellite navigation GNSS, including a combination of position and velocity information; P GNSS represents the three - dimensional position coordinates obtained through the GNSS sensor; V GNSS represents the three - dimensional velocity coordinates obtained through the GNSS sensor; P INS represents the position predicted by the inertial navigation system through IMU integration; V INS represents the velocity predicted by the inertial navigation system through IMU integration; represents the position error, that is, the position deviation between the satellite navigation GNSS observation value and the inertial navigation INS predicted value; denotes the speed error, i.e., the speed deviation between the satellite navigation GNSS observation value and the inertial navigation INS predicted value; denotes the measurement noise within the satellite navigation GNSS system; The observation matrix of the GNSS observation model is:

[0023] Visual observation model: Visual observation adopts the method of calculating the reprojection error. The i observation coordinates of the th feature point are

[0024] where: X i denotes the coordinates of the feature point in the global coordinate system; () represents the camera projection function; R VIO denotes the rotation matrix of visual navigation; p VIO denotes the position of visual navigation; v vis denotes the visual observation noise; The observation matrix of the visual observation model is:

[0025] where:

[0026] where: represents the camera projection function; R represents the rotation matrix; denotes the inertial navigation attitude error; denotes the visual navigation attitude error; denotes the visual 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:

[0027]

[0028] where: K k denotes the Kalman gain matrix, H k denotes the observation matrix obtained from the observation model, i.e., the fused observation source; R k denotes the observation noise covariance matrix; Step S64: Update the attitude, position, velocity, and visual error states respectively to obtain the corrected state; Step S65: First, use the inertial measurement unit (IMU) to obtain the prediction models of attitude, position, and velocity during the prediction phase. At the same time, obtain the covariance of the corresponding prediction errors. Then, the visual observation model performs observations, conducts visual updates, continuously corrects the state, and updates the covariance. When the Global Navigation Satellite System (GNSS) data arrives (i.e., the GNSS module receives sufficient satellite signals to accurately calculate the position information), calculate the position and velocity residuals, and then perform the error state update of the Global Navigation Satellite System (GNSS) observation model and update the covariance again. Repeat the above prediction update steps to continuously fuse multi-sensor data and improve the navigation and positioning accuracy.

[0029] Based on the further optimization of the above solution, the specific construction of the prediction models of position, velocity, and attitude in step S61 is as follows: Use the data of the gyroscope to predict the attitude at the next moment, thereby constructing the prediction model of attitude:

[0030] In the formula: represents the predicted attitude quaternion (i.e., the predicted value at k moment); represents the attitude quaternion at the previous moment; represents the time step (unit: s); represents the angular velocity measured by the gyroscope (unit: rad / s); represents the gyroscope zero bias (unit: rad / s); represents the quaternion multiplication operator; exp represents the small rotation, that is, converting the angular velocity into the increment of the quaternion; Use the accelerometer data to predict the velocity at the next moment, thereby constructing the prediction model of velocity:

[0031] In the formula: represents the predicted velocity (i.e., the predicted value at k moment); represents the velocity at the previous moment the rotation matrix corresponding to the attitude quaternion; represents the value measured by the accelerometer; represents the accelerometer zero bias (unit: m / s²); g represents the gravitational acceleration vector; Use the velocity integration to predict the position at the next moment, thereby constructing the prediction model of position:

[0032] In the formula: represents the predicted position ( k predicted value at time represents the position at the previous time.

[0033] Based on the further optimization of the above solution, the specific update of attitude, position, speed, and visual error status in step S64 is as follows: Attitude correction:

[0034] In the formula: represents k the attitude error updated at time k which is obtained from the satellite navigation GNSS observation value and the inertial navigation INS predicted attitude value at time Position correction:

[0035] In the formula: represents k the position error updated at time k which is obtained from the position deviation between the satellite navigation GNSS observation value and the inertial navigation INS predicted value at time Speed correction:

[0036] In the formula: represents k the speed error updated at time k which is obtained from the speed deviation between the satellite navigation GNSS observation value and the inertial navigation INS predicted value at time Visual error correction:

[0037] In the formula: represents k the position error of the updated vision at time represents k the attitude error of the updated vision at time represents k the position of the vision at time -1; represents k the attitude of the vision at time -1; represents k the updated vision position at time represents k the updated vision attitude at time

[0038] Based on the further optimization of the above solution, the specific update of the covariance in step S65 is as follows:

[0039] In the formula: represents the updated error covariance matrix.

[0040] The following are the technical effects of the solution of the present invention: Through the actuator assembly of the three-jaw fingers of the present invention, it effectively adapts to the boll part of long-staple cotton (that is, the boll part of long-staple cotton is mainly three-petaled), thus solving the problem of cotton floss remaining at the root of the boll during the picking process of long-staple cotton and ensuring the completeness of picking; through the cooperation of the multi-stage lifting truss mechanism, the fixed cross plate, the steering and traveling mechanism, and the connecting collection box, the overall forms a gantry structure, effectively solving the problems of the rows of long-staple cotton being close together and the chassis being unable to drive normally in a narrow space, and adopting the multi-stage lifting method, which is convenient for adjusting the height during picking, and cooperating with the robotic arm assembly to achieve efficient picking of long-staple cotton bolls. It can also lower the height of the overall mechanism in non-picking situations, facilitating normal driving on field roads; through the robotic arm assembly composed of a rotating cloud platform, a moving large arm, an air pipe, a gas source component, a moving small arm, and a moving connecting arm, the rotating cloud platform is driven pneumatically, with strong load capacity, facilitating the realization of precise and efficient long-staple cotton picking actions; through the earthing-up mechanism composed of a top-mounted ring, a hinged connecting rod, a hanging ring fixing disc, a fixed cylinder, a bottom-mounted disc, a rotating screw rod, a lifting ring, and a lever, it effectively solves the problem of long-staple cotton lodging, facilitates the picking of long-staple cotton, and avoids the long-staple cotton picking robot from rolling over the long-staple cotton stalks during walking, while reducing the damage to the surrounding unpicked long-staple cotton.

[0041] In addition, the present invention adopts a ridge-following technology to realize the self-traveling of the long-staple cotton picking robot in the cotton field in the row direction, avoiding damage to the long-staple cotton plants. The present invention also uses a navigation method combining Beidou satellite navigation, visual navigation, and inertial navigation, which can adapt to navigation applications in multiple scenarios, improve the navigation accuracy of the long-staple cotton picking robot in different environments such as foliage occlusion, cloudy days, and strong light (for example: in cloudy days, Beidou satellite navigation is greatly affected, while visual navigation has a better effect; in strong light, visual navigation is greatly affected, while Beidou satellite navigation has a better effect; when blocked by leaves, visual navigation is more interfered, while Beidou satellite navigation and inertial navigation have a better effect), accurately leveraging the advantages of different navigations, complementing each other's strengths, and improving the positioning accuracy of navigation. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] Figure 1 It is a schematic diagram of the overall structure of the picking robot in the embodiment of the present invention.

[0043] Figure 2This is the front view of the picking robot in the embodiment of the present invention.

[0044] Figure 3 This is the internal view of the multi-stage lifting truss mechanism of the picking robot in the embodiment of the present invention.

[0045] Figure 4 This is the structural schematic diagram of the robotic arm assembly of the picking robot in the embodiment of the present invention.

[0046] Figure 5 This is the structural schematic diagram of the actuator assembly of the picking robot in the embodiment of the present invention.

[0047] Figure 6 This is the structural schematic diagram of the steering and traveling mechanism of the picking robot in the embodiment of the present invention.

[0048] Figure 7 This is the structural schematic diagram of the crop-holding mechanism of the picking robot in the embodiment of the present invention.

[0049] Among them, 11 is the positioning plate; 121 is the first cylinder barrel; 122 is the first piston rod; 131 is the second cylinder barrel; 132 is the second piston rod; 141 is the sliding guide cylinder; 142 is the guide pillar; 15 is the chassis cross beam; 20 is the fixed cross plate; 31 is the wheel fork; 32 is the steering bracket; 33 is the connecting shaft; 34 is the traveling wheel; 40 is the connecting collection box; 511 is the rotating cloud platform; 512 is the moving large arm; 513 is the air pipe; 514 is the air source component; 515 is the moving small arm; 516 is the moving connecting arm; 521 is the rear fixing plate; 522 is the front fixing plate; 523 is the fixed connecting shaft; 524 is the moving sliding table; 525 is the ball screw; 526 is the connecting rotating rod; 527 is the rotating block; 528 is the three-jaw finger; 61 is the top ring; 62 is the articulated connecting rod; 63 is the hanging ring fixing disc; 64 is the fixed cylinder; 65 is the bottom disc; 66 is the rotating screw; 67 is the lifting ring; 68 is the lever. Detailed implementation manners

[0050] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments.

[0051] Embodiment 1: A cotton picking robot includes two groups of multi-stage lifting truss mechanisms, a fixed cross plate 20, a steering and traveling mechanism, a connecting collection box 40, two groups of end effector mechanisms and a crop-holding mechanism (such as Figure 1As shown in the figure, two groups of multi-stage lifting truss mechanisms are symmetrically arranged at both lower ends of the fixed cross plate 20 and are connected to the bottom surface of the fixed cross plate 20. The multi-stage lifting truss mechanism includes a positioning plate 11, two groups of first lifting components, a second lifting component, two groups of positioning and supporting components, and a chassis cross beam 15 (as Figure 2 shown); two groups of first lifting components are symmetrically arranged on both sides of the inner wall of the positioning plate 11, including a first cylinder barrel 121 and a first piston rod 122. The first cylinder barrel 121 is arranged on the inner wall of the positioning plate 11 and its lower end is slidably connected to the first piston rod 122. The chassis cross beam 15 is arranged on the lower side of the corresponding positioning plate 11, and one end (i.e., the bottom end) of the first piston rod 122 away from the first cylinder barrel 121 is fixedly connected to the end surface of the chassis cross beam 15; a second lifting component is arranged in the middle of the inner wall of the positioning plate 11, including a second cylinder barrel 131 and a second piston rod 132. The second cylinder barrel 131 is arranged on the inner wall of the positioning plate 11 and its upper end is slidably connected to the second piston rod 132. One end (i.e., the top end) of the second piston rod 132 away from the second cylinder barrel 131 is fixedly connected to the bottom surface of the fixed cross plate 20; positioning and supporting components are respectively arranged between the second lifting component and the two groups of first lifting components. The positioning and supporting component includes a sliding guide cylinder 141 and a guide pillar 142. The sliding guide cylinder 141 is arranged on the inner wall of the positioning plate 11 and its upper end slidably arranges the guide pillar 142. One end (i.e., the top end) of the guide pillar 142 away from the sliding guide cylinder 141 is fixedly connected to the bottom surface of the fixed cross plate 20 (the connection between the positioning plate 11 and the fixed cross plate 20 is realized through the second piston rod 132 and the guide pillar 142).

[0052] The steering and traveling mechanism is arranged at the bottom of the multi-stage lifting truss mechanism, and two steering and traveling mechanisms are arranged at the bottom of one group of multi-stage lifting truss mechanisms (as Figure 1 shown. In this embodiment, four steering and traveling mechanisms are arranged 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 cross beam 15 and the longitudinal section of the wheel fork 31 is a Z-shaped structure (as Figure 6 shown). The steering bracket 32 is rotatably arranged at the bottom of the wheel fork 31 and the connecting shaft 33 is rotatably arranged in the middle of the steering bracket 32. The traveling wheel 34 is fixedly sleeved on the outer wall of the connecting shaft 33 (the rotation of the traveling wheel 34 is realized by controlling the rotation of the connecting shaft 33).

[0053] The connection and collection box 40 is arranged between the two groups of multi-stage lifting truss mechanisms, and both ends of the connection and collection box 40 are respectively fixedly connected to the inner sides of the corresponding multi-stage lifting truss mechanisms (i.e., the positioning plates 11) (as Figure 2 shown). The connection and collection box 40 is located at the rear end of the fixed cross plate 20 (i.e., Figure 1 the left end shown).

[0054] Two sets of end effectors are symmetrically arranged in the middle of the bottom surface of the fixed cross plate 20 and are located inside the two sets of multi-stage lifting truss mechanisms. The end effector includes a robotic arm assembly and an actuator assembly. The robotic arm assembly includes a rotary turntable 511, a moving boom 512, an air pipe 513, a gas source component 514, a moving forearm 515, and a moving connecting arm 516 (as Figure 4 shown). The rotary turntable 511 is arranged on the bottom surface of the fixed cross plate 20, and the moving boom 512 is arranged at the bottom of the rotary turntable 511. One end of the moving boom 512 away from the rotary turntable 511 is rotatably connected to the moving forearm 515, and one end of the moving forearm 515 away from the moving boom 512 is rotatably connected to the moving connecting arm 516; one end of the air pipe 513 is fixedly connected to the rotary turntable 511 through a valve, and the other end is fixedly connected to the gas source component 514 arranged on the moving boom 512 through a valve; the actuator assembly includes a rear fixing plate 521, a front fixing plate 522, a fixed connecting shaft 523, a moving slide 524, a ball screw 525, a connecting rotating rod 526, a rotating block 527, and a three-jaw finger 528 (as Figure 5 shown); the rear fixing plate 521 is arranged on the moving connecting arm 516, the front fixing plate 522 is arranged on the side of the rear fixing plate 521 away from the moving connecting arm 516, and the front fixing plate 522 is arranged in parallel with the rear fixing plate 521. A plurality of fixed connecting shafts 523 are evenly arranged between the front fixing plate 522 and the rear fixing plate 521 and on their outer circles (the front fixing plate 522 and the rear fixing plate 521 are connected by the fixed connecting shafts 523, and the number of the fixed connecting shafts 523 is set according to the actual situation, not less than two, and three are set in this embodiment). A rotating ball screw 525 is arranged in the middle of the front fixing plate 522 and the rear fixing plate 521 (at the same time, the ball screw 525 is controlled to rotate by a driving motor arranged at one end of the rear fixing plate 521). The moving slide 524 is arranged between the front fixing plate 522 and the rear fixing plate 521 and is respectively penetrated by the fixed connecting shafts 523 and the ball screw 525. The moving slide 524 is slidably connected to the fixed connecting shafts 523 and is threadedly connected to the ball screw 525. Three rotating blocks 527 are evenly arranged on the side surface of the front fixing plate 522 away from the fixed connecting shafts 523. The rotating blocks 527 are rotatably connected to the front fixing plate 522. The moving slide 524 is provided with three connecting rotating rods 526 corresponding to the rotating blocks. One end of the connecting rotating rod 526 is rotatably connected to the moving slide 524, and the other end is rotatably connected to the rotating block 527. The three-jaw finger 528 is arranged corresponding to the rotating block 527, and one end of it is connected to the rotating block 527. A convex arc surface is arranged on the inner side of each jaw finger of the three-jaw finger 528 (as Figure 5 shown).

[0055] The front end of the fixed cross plate 20 (i.e., Figure 1A grain lifter mechanism is provided on the right side (as shown on the right side in the figure). The grain lifter mechanism includes a top-mounted circular ring 61, a hinged connecting rod 62, a hanging ring fixing disc 63, a fixed cylinder 64, a bottom-mounted disc 65, a rotating screw 66, a lifting circular ring 67, and a shifting rod 68. The top-mounted circular ring 61 is provided on the front end face of the fixed cross plate 20. One end of the hinged connecting rod 62 is rotatably connected to the top-mounted circular ring 61, and the other end is rotatably connected to the hanging ring fixing disc 63 (as Figure 1 shown); The fixed cylinder 64 is coaxially provided on the bottom surface of the hanging ring fixing disc 63, and a coaxial bottom-mounted disc 65 is provided at one end of the fixed cylinder 64 away from the bottom surface of the hanging ring fixing disc 63 (as Figure 7 shown); The rotating screw 66 is coaxially provided in the inner cavity of the fixed cylinder 64. The upper end of the rotating screw 66 penetrates through the hanging ring fixing disc 63 and is rotatably connected; The lifting circular ring 67 is sleeved on the outer wall of the fixed cylinder 64 (the inner wall of the lifting circular 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 provided in the inner cavity of the fixed cylinder 64 (for example: a lifting block is provided 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 penetrated and threadedly connected by the rotating screw 66. The outer wall of the lifting block is slidably connected to the inner wall of the fixed cylinder 64. No less than two vertical sliding grooves are provided on the side wall of the fixed cylinder 64, and the vertical sliding grooves are evenly distributed around the central axis of the fixed cylinder 64. The inner wall of the lifting circular ring 67 is connected to the outer wall of the lifting block through a connecting slider, and the connecting slider corresponds to the vertical sliding groove); Shifting rods 68 are symmetrically provided on both sides of the lifting circular ring 67 (as Figure 7 shown. The shifting rod 68 is composed of an inclined rod and a flat rod, and the inclined rod is used to connect the lifting circular ring 67 and the flat rod).

[0056] Embodiment 2: A picking navigation method for a cotton picking robot, using the cotton picking robot as described in Embodiment 1, includes: Step S1, adjusting the spatial height: When the picking robot reaches the pre-picking position, according to the growth situation of the long-staple cotton plants, the spatial height of the picking robot is adjusted in real time through a multi-stage lifting truss mechanism; Specifically, when the height of the long-staple cotton plant is relatively high and the space height needs to be lifted, the first lifting component of the multi-stage lifting truss mechanism (i.e., the first cylinder barrel 121 and the first piston rod 122) is activated. Under the action of the mutual force, by pushing the chassis cross beam 15, the positioning plate 11 is lifted, so that the fixed cross plate 20, the connecting collection box 40, the two sets of end effectors and the threshing mechanism are lifted synchronously; if the lifting height is insufficient, the second lifting component (i.e., the second cylinder barrel 131 and the second piston rod 132) is activated, so as to further lift the fixed cross plate 20, making the two sets of end effectors and the threshing mechanism lifted synchronously. During this process, the positioning and supporting component (i.e., the sliding guide cylinder 141 and the guide pillar 142) expands and contracts following the lifting of the second lifting component, thus playing the role of fixed support and guiding; finally, the overall height of the picking robot can meet the space height requirements for picking the cotton bolls of long-staple cotton. When the height needs to be lowered, the second lifting component is first activated to descend, causing the fixed cross plate 20 to move downward, and then prompting the end effector and the threshing mechanism to move downward synchronously; then the first-stage lifting mechanism is activated to descend, driving the positioning plate 11 to descend, and then prompting the fixed cross plate 20, the connecting collection box 40, the two sets of end effectors and the threshing mechanism to move downward synchronously.

[0057] Step S2, envelope picking and collection: The robotic arm assembly is activated to move the actuator assembly to the target picking point position. The actuator assembly envelopes the cotton bolls of the long-staple cotton. The robotic arm assembly moves to pick the cotton bolls and places them in the connecting collection box, and then the end effector resets to complete one picking. Specifically: when the picking robot reaches the pre-picking position (the picking position can be judged by conventional image recognition, and conventional means in this field can be directly adopted, which will not be specifically described in this embodiment), the pneumatic-electric hybrid drive robotic arm assembly is activated. The large robotic arm 512 and the moving small arm 515 are driven by motors to move the actuator assembly to a position 3-5 cm above 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 branch of 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 convex part at the front end of the claw finger of the actuator assembly can touch the gap at the root between the boll petals, and then the movement of the robotic arm assembly stops. At this time, the driving motor starts to rotate forward, the ball screw 525 rotates, the moving slide 524 moves forward, and the rotating block 527 is pushed to rotate towards the axis through the connecting rotating rod 526, so as to drive the three-claw claw finger 528 to achieve the enveloping of the long-staple cotton boll. Then, the air source component 514 is activated, and the rotating cloud platform 511 is driven pneumatically to rotate the actuator assembly and the robotic arm assembly above the connecting collection box 40. The driving motor of the actuator assembly rotates in reverse, the ball screw 525 rotates in the reverse direction, the moving slide 524 moves backward, and the rotating block 527 is pushed to rotate away from the axis through the connecting rotating rod 526, so as to drive the three-claw claw finger 528 to achieve the opening action, collect the long-staple cotton boll in the connecting collection box 40, and the rotating cloud platform 511 resets, completing a process of enveloping, picking and collecting the long-staple cotton boll.

[0058] Step S3, lodging stem reaping: Before the picking robot moves to the next pickable space, it is judged whether there is a situation of stem lodging. If so, the reaping mechanism is activated to lift the lodged long-staple cotton stems. The reaping mechanism can be considered for disassembly according to the actual working conditions. When the growth condition of the whole cotton field is good and there are no obvious lodged long-staple cotton plants, the reaping mechanism does not need to be installed. When there are many lodged long-staple cotton plants in the whole cotton field, the reaping mechanism needs to be installed.

[0059] The specific working principle of the cotton supporting mechanism is as follows: When preparing to move to the next pickable space, if the long-staple cotton stalks in the next pickable space are lodged, the cotton supporting mechanism is activated. The bottom disk 65 is moved into the cotton field and placed on the cotton field land corresponding to the lodged long-staple cotton stalks by the up-and-down movement of the articulated connecting rod 62. The screw 66 is rotated forward, so that the lifting ring 67 moves upward along the outer wall of the fixed cylinder 64, and then drives the two side lever rods 68 to move upward, stripping the lodged long-staple cotton stalks; After that, the articulated connecting rod 62 drives the mechanism composed of the entire hanging ring fixed disk 63, the fixed cylinder 64, the bottom disk 65, the rotating screw 66, the lifting ring 67 and the lever rod 68 to lift upward, thereby lifting the lodged long-staple cotton stalks and assisting in completing the picking of the cotton bolls of the lodged long-staple cotton plants. When the picking is over, the screw 66 is rotated reversely, so that the lifting ring 67 moves downward, driving the two side lever rods 68 to move downward to the lowest point of the stroke. The articulated connecting rod 62 and the hanging ring end face part at the top of the hanging ring fixed disk 63 swing up and down, placing the lodged long-staple cotton stalks back to the original position, and the cotton supporting mechanism is reset, completing one cotton supporting process and proceeding to the picking work of the next pickable space.

[0060] Step S4, Field ridgeline finding technology: The automatic ridgeline finding technology is adopted to enable the picking robot to find the ridgeline and move automatically in the cotton field, and move the picking robot to the next pickable space; Specifically: Step S41, Collect images of the cotton field through a high-resolution camera to obtain a data set; Step S42, First, perform grayscale processing on the collected images to obtain grayscale images:

[0061] In the formula: R, G, and B respectively represent the pixel values of the red, green, and blue channels; Then, perform filtering processing on the images to obtain smooth images:

[0062] In the formula: represents the Gaussian kernel function, represents the standard deviation; Step S43, Perform edge detection to obtain the ridgeline edges in the images; Adopt the Sobel operator method to obtain the gradients of the images, and then obtain the gradient magnitudes and gradient directions of the images: Gradient magnitude:

[0063]

[0064] Gradient direction:

[0065] Retain the edge points with the largest gradient magnitude, thin the edges, reduce the computational complexity of edge detection, obtain clearer edges, and use the high and low threshold method to screen edge points (that is, set a high threshold T h and a low threshold T l ). For the part with the high threshold (i.e., the edge points not less than the high threshold), retain the strong edge points, while for the part with the low threshold (i.e., the edge points between the high threshold and the low threshold), retain the weak edge points connected to the strong edge points to obtain clearer and more realistic ridge line edges; Step S44: First, extract the ridge lines in the image by the method of Hough transform to determine the ridge line parameters:

[0066] In the formula, represents the distance from the straight line to the origin; represents the angle of the straight line; ( x, y ) represents the pixel coordinates; Then, use the least squares method to fit a straight line to the detected edge points and minimize the error, and describe the ridge line position by the fitted straight line:

[0067] In the formula: ( x i ,y i ) represents the pixel coordinates of the edge points; Step S45: Perform ridge line tracking based on the lateral deviation and heading angle deviation between the picking robot and the target ridge line:

[0068] In the formula: Y 1 represents the current position of the picking robot, kx + b represents the target ridge line equation; represents the current heading angle, the direction angle corresponding to the slope of the target ridge line; e Y represents the heading deviation, represents the heading angle deviation; Generate a control command according to the lateral deviation and the heading angle deviation, generate a steering command for the current position and the target point of the picking robot, and adjust the driving direction of the picking robot through feedback and preview points: Select a future point on the target path ( x f ,yf ) As the preview point:

[0069] In the formula: L d represents the preview distance, that is, the straight-line distance between the current point ( x, y ) and the preview point; Compensate the heading angle deviation through the PID controller to generate a steering angle command :

[0070] In the formula: represents the weight coefficient, and its value range is [0, 1]; K p , K i , K d respectively represent the proportional, integral, and differential parameters; Speed adjustment command:

[0071] In the formula: v 0 represents the reference speed, K v represents the speed adjustment coefficient.

[0072] Step S5, Turning at the field edge: After completing the operation task of one ridge, turn in the narrow space of the field through four groups of steering walking mechanisms; when the long-staple cotton picking robot moves to the edge of the cotton field and needs to turn, the four groups of four-wheel steering mechanisms are activated. Since the steering methods of the walking wheels 34 of the four groups of steering walking mechanisms are the same, and the front wheels and the rear wheels turn in the same direction, the turning radius in the narrow space of the cotton field is reduced; when the field turning is completed, the steering bracket 32 drives the connecting shaft 33 and the walking wheel 34 to rotate in the opposite direction, and the return speed of the rear wheels is 15% faster than that of the front wheels, which can quickly complete the return reset of the walking wheel 34 and enter the straight-line driving stage of the walking wheel 34, completing one turning process at the edge of the cotton field; adopt an "S"-shaped driving route to complete the field U-turn and in-row self-driving of the entire long-staple cotton field.

[0073] Step S6, Cotton field navigation method: Realize the efficient autonomous navigation of the picking robot in different environments through a combined navigation method; Specifically: Step S61, Initialize each navigation parameter, sampling frequency, and cut-off frequency; construct a prediction model for position, speed, and attitude; Use the data of the gyroscope to predict the attitude at the next moment, so as to construct a prediction model for attitude:

[0074] In the formula: represents the predicted attitude quaternion (i.e., k the predicted value at the represents the attitude quaternion at the previous moment; represents the time step (unit: s); represents the angular velocity measured by the gyroscope (unit: rad / s); represents the gyroscope zero bias (unit: rad / s); represents the quaternion multiplication operator; exp represents a small rotation, that is, converting the angular velocity into an increment of the quaternion; Predict the velocity at the next moment using the accelerometer data, and thus construct a prediction model for the velocity:

[0075] In the formula: represents the predicted velocity (i.e., k the predicted value at the represents the velocity at the previous moment the rotation matrix corresponding to the attitude quaternion; represents the value measured by the accelerometer; represents the accelerometer zero bias (unit: m / s²); g represents the gravitational acceleration vector; Predict the position at the next moment by integrating the velocity, and thus construct a prediction model for the position:

[0076] In the formula: represents the predicted position ( k the predicted value at the represents the position at the previous moment.

[0077] Step S62, obtain the covariance of the prediction error through the calculation of the prediction model. First, calculate the state transition matrix F k :

[0078] In the formula: I represents the identity matrix; represents the cross product matrix of the gyroscope angular velocity; represents the cross product matrix of the accelerometer measurement value; R(q) represents the rotation matrix corresponding to the attitude quaternion; represents the time difference between the current moment and the next prediction moment; Then, calculate the process noise covariance Qk :

[0079] 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-matrix along the main diagonal to form a block diagonal matrix; Finally, perform covariance prediction calculation:

[0080] Wherein: represents the predicted error covariance matrix; represents the error covariance matrix at the previous moment; Step S63, construct an observation model through Global Navigation Satellite System (GNSS) observations and visual observations: Global Navigation Satellite System (GNSS) observation model:

[0081] Wherein: Z GNSS represents the observation value of satellite navigation GNSS, including a combination of position and velocity information; P GNSS represents the three-dimensional position coordinates obtained through the GNSS sensor; V GNSS represents the three-dimensional velocity coordinates obtained through the GNSS sensor; P INS represents the position predicted by the inertial navigation system through IMU integration; V INS represents the velocity predicted by the inertial navigation system through IMU integration; represents the position error, that is, the position deviation between the satellite navigation GNSS observation value and the inertial navigation INS predicted value; represents the velocity error, that is, the velocity deviation between the satellite navigation GNSS observation value and the inertial navigation INS predicted value; represents the measurement noise inside the satellite navigation GNSS system; Wherein:

[0082] Wherein: and respectively represent the position observation value and the velocity observation value of satellite navigation GNSS; and respectively represent the position prediction value and the speed prediction value of the inertial navigation INS; The observation matrix of the GNSS observation model is:

[0083] Vision observation model: Vision observation adopts the method of calculating the reprojection error. The observation coordinates of the i th feature point are , then its projection model is:

[0084] In the formula: X i represents the coordinates of the feature point in the global coordinate system; () represents the camera projection function; R VIO represents the rotation matrix of vision navigation; p VIO represents the position of vision navigation; v vis represents the vision observation noise; The observation matrix of the vision observation model is:

[0085] Among them:

[0086] In the formula: R represents the rotation matrix; represents the inertial navigation attitude error; represents the vision navigation attitude error; represents the 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:

[0087]

[0088] In the formula: K k represents the Kalman gain matrix, H k represents the observation matrix obtained from the observation model, that is, the fusion observation source; R k represents the observation noise covariance matrix; Step S64: Update the attitude, position, speed, and vision error states respectively to obtain the corrected state; specifically: Attitude correction:

[0089] Wherein: represents k the attitude error updated at time ; which is obtained from the GNSS observations of satellite navigation and the predicted attitude value of inertial navigation INS at time , that is k ; ; Position correction:

[0090] Wherein: represents k the position error updated at time ; which is obtained from the position deviation between the GNSS observations of satellite navigation and the predicted value of inertial navigation INS at time k ; Velocity correction:

[0091] Wherein: represents k the velocity error updated at time ; which is obtained from the velocity deviation between the GNSS observations of satellite navigation and the predicted value of inertial navigation INS at time k ; Visual error correction:

[0092] Wherein: represents k the position error of the updated vision at time , that is (By the visual ranging method, using a camera to capture the scene image, extracting feature points from the captured image, calculating the position of the camera in the scene through the matched feature points and using the triangulation method, that is , and using the accelerometer and gyroscope of the IMU to obtain the predicted position at this time by integrating the acceleration and angular velocity, that is ; finally obtaining the position error); represents k the attitude error of the updated vision at time , that is (Using feature matching in the vision system, estimating the attitude from the feature points using the rotation matrix to obtain the observed attitude at this time ; the predicted attitude at this time can be directly calculated through the gyroscope ; finally obtaining the attitude error); represents k the position of the vision at time -1; represents k the attitude of the vision at time -1; represents k the updated vision position at time Indicates k The visual pose after the moment update.

[0093] Step S65: First, use an inertial measurement unit (IMU) to obtain a prediction model of pose, position, and velocity during the prediction phase (i.e., use step S61), and at the same time, obtain the covariance of the corresponding prediction error (i.e., use step S62); then, the visual observation model performs observations (i.e., use step S63), and performs visual updates to continuously correct the state (i.e., use step S64), and update the covariance; when the Global Navigation Satellite System (GNSS) data arrives (i.e., the GNSS module receives sufficient satellite signals to accurately calculate the position information), calculate the position and velocity residuals (i.e., And ), and then perform an error state update of the Global Navigation Satellite System (GNSS) observation model (i.e., use step S64), and update the covariance again; repeat the above prediction update steps to continuously fuse multi-sensor data and improve the navigation and positioning accuracy; The specific update of the covariance is as follows:

[0094] In the formula: Indicates the updated error covariance matrix.

Claims

1. A cotton picking robot, characterized in that: It includes two groups of multi-stage lifting truss mechanisms, a fixed horizontal plate, a steering walking mechanism, a connecting collection box, two groups of terminal actuators and a straw-supporting mechanism. The two groups of multi-stage lifting truss mechanisms are symmetrically arranged at the two ends of the lower side of the fixed horizontal plate and are connected to the bottom surface of the fixed horizontal plate. The steering walking mechanism is arranged at the bottom of the multi-stage lifting truss mechanism and two groups of steering walking mechanisms are arranged at the bottom of one group of multi-stage lifting truss mechanisms; the connecting collection box is arranged between the two groups of multi-stage lifting truss mechanisms and the two ends of the connecting collection box are respectively fixedly connected to the inner side of the corresponding multi-stage lifting truss mechanisms, and the connecting collection box is located at the rear end of the fixed horizontal plate; the two groups of terminal actuators are symmetrically arranged in the middle of the bottom surface of the fixed horizontal plate and are located on the inner side of the two groups of multi-stage lifting truss mechanisms, and the straw-supporting mechanism is arranged at the front end of the fixed horizontal plate.

2. A cotton picking robot according to claim 1, characterized in that: The multi-stage lifting truss mechanism includes a positioning plate, two groups of first lifting components, a second lifting component, two groups of positioning support components and a chassis crossbeam; the two groups 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 arranged on the inner wall of the positioning plate and its lower end is slidably connected to the first piston rod, the chassis crossbeam 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 crossbeam; a second lifting component is arranged in the middle of the inner wall of the positioning plate, including a second cylinder and a second piston rod, the second cylinder is arranged on the inner wall of the positioning plate and its upper end is slidably connected to the second piston rod, and the end of the second piston rod away from the second cylinder is fixedly connected to the bottom surface of the fixed cross plate; positioning support components are respectively arranged between the second lifting component and the two groups of first lifting components, the positioning support components include a sliding guide cylinder and a guide pillar, the sliding guide cylinder is arranged on the inner wall of the positioning plate and a guide pillar is slidably arranged on its upper end, and the end of the guide pillar away from the sliding guide cylinder is fixedly connected to the bottom surface of the fixed cross plate.

3. A cotton picking robot according to claim 1 or 2, characterized in that: The steering travel mechanism includes a wheel fork, a steering bracket, a connecting shaft and a travel 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 arranged at the bottom of the wheel fork and the connecting shaft is rotatably arranged in the middle of the steering bracket. The outer wall of the connecting shaft is fixedly sleeved with the travel wheel.

4. A cotton picking robot according to claim 2 or 3, characterized in that: The end actuator includes a mechanical arm assembly and an actuator assembly. The mechanical arm assembly includes a rotating platform, a moving arm, an air pipe, an air source element, a moving arm and a moving connecting arm. The rotating platform is arranged on the bottom surface of the fixed horizontal plate and the moving arm is arranged at the bottom of the rotating platform. The end of the moving arm away from the rotating platform is rotatably connected to the moving arm, and the end of the moving arm away from the moving arm is rotatably connected to the moving connecting arm; one end of the air pipe is fixedly connected to the rotating platform through a valve, and the other end is fixedly connected to the air source element arranged on the moving arm through a valve; the actuator assembly includes a rear fixed plate, a front fixed plate, a fixed coupling, a movable slide, a ball screw, a connecting rod, a rotating block and a three-claw claw finger; the rear fixed plate is arranged on the moving connecting arm, and the front fixed plate is arranged at the rear fixed plate away from the moving connecting arm One side of the front fixed plate is arranged parallel to the rear fixed plate, a plurality of fixed couplings are evenly arranged between the front fixed plate and the rear fixed plate and located at their outer circles, a rotating ball screw is arranged in the middle of the front fixed plate and the rear fixed plate, the movable slide is arranged between the front fixed plate and the rear fixed plate and are respectively penetrated by the fixed coupling and the ball screw, the movable slide is slidably connected to the fixed coupling and is threadedly connected to the ball screw, three rotating blocks are evenly arranged on the side surface of the front fixed plate away from the fixed coupling, the rotating block is rotatably connected to the front fixed plate, three connecting rotating rods are arranged corresponding to the rotating block of the movable slide, and one end of the connecting rotating rod is rotatably connected to the movable slide, and the other end is rotatably connected to the rotating block, the three-claw claw finger is arranged corresponding to the rotating block and one end thereof is connected to the rotating block, and a convex arc surface is arranged on the inner side of each claw finger of the three-claw claw finger.

5. A cotton picking robot according to claim 2 or 3, characterized in that: The rice-supporting mechanism comprises an overhead ring, an articulated connecting rod, a lifting ring fixed disc, a fixed cylinder, a bottom disc, a rotating screw, a lifting ring and a shifting rod. The overhead ring is arranged on the front end face of the fixed horizontal plate, one end of the articulated connecting rod is rotatably connected to the overhead ring, and the other end is rotatably connected to the lifting ring fixed disc; the fixed cylinder is coaxially arranged on the bottom surface of the lifting ring fixed disc, and a coaxial bottom disc is arranged on the end of the fixed cylinder away from the bottom surface of the lifting ring fixed disc; the rotating screw is coaxially arranged in the inner cavity of the fixed cylinder, and the upper end of the rotating screw penetrates the lifting ring fixed 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 arranged in the inner cavity of the fixed cylinder; the shifting rods are symmetrically arranged on both sides of the lifting ring.

6. The cotton picking robot navigation method according to claim 1, characterized in that: include: 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 a multi-level lifting gantry mechanism according to the growth of the long-staple cotton plants; Step S2, enveloping, picking and collecting: the mechanical arm assembly is started, and the actuator assembly is moved to the target picking point position, the actuator assembly envelops the cotton bolls of the long-staple cotton, the mechanical arm assembly moves to pick the cotton bolls and place them in the connected collection box, and then the end actuator is reset to complete one picking; Step S3, lifting the fallen stalks: before the picking robot moves to the next picking space, it determines whether the stalks have fallen. If so, the lifting mechanism is activated to lift up the fallen long-staple cotton stalks; Step S4, field ridge-finding technology: using automatic ridge-finding technology to enable the picking robot to self-walk in the cotton field and move the picking robot to the next available picking space; Step S5, turning in the field: after completing the operation task of one ridge, the turning in the narrow space of the field is realized through four sets of steering walking mechanisms; Step S6, cotton field navigation method: through the combined navigation method, efficient autonomous navigation of the picking robot in different environments is achieved.

7. The cotton picking robot navigation method according to claim 6, characterized in that: The step S4 is specifically as follows: Step S41, collecting images of the cotton field by using a high-resolution camera to obtain a data set; Step S42, first grayscale the collected image to obtain a grayscale image, and then filter the image to obtain a smoothed image; Step S43, performing edge detection to obtain the ridge edge in the image; Step S44: First, the ridge lines in the image are extracted by using the Hough transform method to determine the ridge line parameters: In the formula, Represents the distance from the straight line to the origin; Represents the angle of a straight line; ( x,y ) represents the pixel coordinates; Then, the least squares method is used to fit the detected edge points to a straight line and minimize the error. The ridge position is described by the fitted straight line: Where: x i ,y i ) represents the pixel coordinates of edge points; Step S45, tracking the ridge line by the lateral deviation and heading angle deviation between the picking robot and the target ridge line: Where: Y 1 indicates the current position of the picking robot, kx + b represents 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 the heading deviation, Indicates the heading angle deviation; The control instructions are generated based on the lateral deviation and heading angle deviation, and the steering instructions are generated based on the current position and target point of the picking robot. The driving direction of the picking robot is adjusted through feedback and preview points: Select a future point on the target path ( x f ,y f ), as the preview point: Where: L d Indicates the preview distance, that is, the current point ( x,y ) and the straight-line distance between the preview point; The PID controller is used to compensate for the heading angle deviation and generate the steering angle command. : Where: represents the weight coefficient; K p , K i , K d Represent the proportional, integral, and differential parameters respectively; Speed ​​adjustment command: Where: v 0 means the base speed, K v Indicates the speed adjustment factor.

8. The cotton picking robot navigation method according to claim 6, characterized in that: The step S6 is specifically as follows: Step S61, initializing various navigation parameters, sampling frequency, and cutoff frequency; constructing a prediction model of position, speed, and posture; Step S62: Obtain the covariance of the prediction error by calculating the prediction model. First, calculate the state transfer matrix F k : Where: I represents the identity matrix; The cross product matrix representing the gyroscope angular velocity; A cross product matrix representing the accelerometer measurements; R(q) Represents the rotation matrix corresponding to the attitude quaternion; Indicates the time difference between the current moment and the next predicted moment; Then, calculate the process noise covariance Q k : Where: Q ins represents the inertial navigation noise covariance matrix; Q vio represents the visual navigation noise covariance matrix; diag() It means that the input sub-matrices are arranged along the main diagonal to form a block diagonal matrix; Finally, the covariance prediction calculation is performed: Where: represents the predicted error covariance matrix; Represents the error covariance matrix of the previous moment; Step S63: construct an observation model through global navigation satellite system observation and visual observation: GNSS Observation Model: Where: Z GNSS Represents the observation value of satellite navigation GNSS, which contains 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 Indicates the position predicted by the inertial navigation system through IMU integration; V INS It indicates the speed predicted by the inertial navigation system through IMU integration; Represents the position error, that is, the position deviation between the satellite navigation GNSS observation value and the inertial navigation INS prediction value; Represents the speed error, that is, the speed deviation between the satellite navigation GNSS observation value and the inertial navigation INS prediction value; Represents the measurement noise inside the satellite navigation GNSS system; The observation matrix of the GNSS observation model is: Visual observation model: Visual observation uses the method of calculating reprojection error. i The observed coordinates of the feature points are , then its projection model is: Where: X i Represents the coordinates of the feature point in the global coordinate system; () represents the camera projection function; R VIO Represents the rotation matrix for visual navigation; p VIO Indicates the location of visual navigation; v vis represents visual observation noise; The observation matrix of the visual observation model is: in: Where: R represents the rotation matrix; Represents the inertial navigation attitude error; Represents the attitude error of visual navigation; Represents the position error of visual navigation; 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 the state estimation: Where: K k represents the Kalman gain matrix, H k Represents the observation matrix obtained by the observation model, that is, the fused observation source; R k represents the observation noise covariance matrix; Step S64, respectively updating the posture, position, speed, and visual error states to obtain a corrected state; Step S65: First, an inertial measurement unit is used to obtain a prediction model of attitude, position, and velocity in the prediction stage, and at the same time, the covariance of the corresponding prediction error is 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 data arrives, the position and velocity residuals are calculated, and then the error state of the global navigation satellite system observation model is updated, and the covariance is updated again; the above prediction and update steps are repeated, and the multi-sensor data is continuously integrated to improve the navigation and positioning accuracy.

9. The cotton picking robot navigation method according to claim 8, characterized in that: The updating of the covariance in step S65 is specifically as follows: Where: represents the updated error covariance matrix.

Images