A picking robot with learning ability and a picking method

Abstract

A picking robot with learning ability and a picking method belong to the technical field of bionic picking equipment. The present invention consists of a mounting base, a stem support device, a clamping device, a visual device, an edge learning module, a control cabinet, and an air pump. As an end effector adapted for a picking robot, the present invention can minimize damage to the fruit during the actual picking operation. The concept of bionics is implemented in the design, and the picking behavior can simulate the bionics of human hands, the clamping equipment can simulate the bionics of octopus suction cups, and the wrist joints can simulate the bionics of elephant trunks. The present invention can autonomously learn the picking movements of human hands before the picking behavior begins, thereby improving the adaptability of the robot to different growth states. The wrist joints of the bionic elephant trunk are flexible, the bionic octopus suction cups are firmly clamped, and the behavior is stable, which can well achieve non-destructive picking.

Description

Technical Field

[0001] The present invention belongs to the technical field of bionic picking equipment, and specifically relates to a picking manipulator with learning ability and a picking method Background Art

[0002] The high fruit damage rate is currently the key factor restricting the industrialization of picking robots. Therefore, solving the problem of picking damage is a key and difficult issue in the research and development of picking robots.

[0003] The main reasons that cause damage to fruits when picked by picking robots are as follows: 1. The picking posture of the picking robot is unreasonable, causing collision and squeezing damage to the machine, stems and branches and fruits; 2. The wrist joint posture is stiff, with poor flexibility and insufficient coordination and cooperation with the visual system; 3. The picking hand lacks flexible adsorption parts and the contact with the fruit is too hard. Summary of the Invention

[0004] The purpose of the present invention is to provide a picking manipulator and a picking method with learning ability. The picking manipulator is installed on a picking robot and can learn the picking movements of human hands before the picking behavior begins. It has strong adaptability to fruits in different environments; the picking manipulator is installed with a stem supporting device, which can support the stem while picking fruits; to address the defect that the current picking manipulator movement is too rigid, the present invention is based on the concept of bionics and innovatively designs a bionic elephant trunk wrist joint component, which can move more freely under the action of air force, is installed with angle measurement sensors and distance measurement sensors, and the movement behavior is more precise; the finger component adopts a bionic octopus suction cup structure, which can better provide support force and improve the stability of fruit picking.

[0005] A picking robot with learning ability of the present invention is composed of a mounting base A, a stem supporting device B, a clamping device C, a visual device D, an edge learning module E, a control cabinet F and an air pump G, wherein the front plate of the cover 1 of the mounting base A is fixedly connected to the rear plate of the slide rail 14 of the stem supporting device B by bolts; the driving wheel 5 of the mounting base A is movably connected to the base 16 of the wrist joint component C1 in the clamping device C through a transmission shaft; the depth camera 43 of the visual device D is fixed to the support plate 2 of the mounting base A; the edge learning module E is communicatively connected to the visual device D; the control cabinet F is communicatively connected to the mounting base A, the stem supporting device B, the clamping device C, the visual device D, the edge learning module E and the air pump G respectively; the air pump G is connected to the pneumatic push rod 10 of the stem supporting device B, the air hole group 17 and the airway 22 of the wrist joint component C1, the airway hole pair Ⅱ36 of the pneumatic component C2, and the splint airway pair 42a of the finger component C3 through an air pipe.

[0006] The mounting seat A is composed of a cover shell 1, a support plate 2, a bearing 3, a motor I4, a driving wheel 5, a synchronous belt 6 and a driven wheel 7. The support plate 2 is fixed to the upper end of the cover shell 1, and the motor I4 is fixed to the cover shell 1; the driving wheel 5, the synchronous belt 6 and the driven wheel 7 are located inside the cover shell 1, and the driving wheel 5 is connected to the driven wheel 7 via the synchronous belt 6; the outer ring of the bearing 3 is fixed to the lower part of the rear plate of the cover shell 1; the motor I4 is fixed to the lower part of the front plate of the cover shell 1, the middle part of the output shaft of the motor I4 is connected to the center key of the driving wheel 5, and the rear end of the output shaft of the motor I4 is interference connected with the inner ring of the bearing 3.

[0007] The stem supporting device B is composed of fingers 8, rod I9, pneumatic push rod 10, rod II11, screw 12, slider 13, slide rail 14 and motor II15. The fingers 8, rod I9, pneumatic push rod 10, rod II11 and slider 13 are arranged in sequence from front to back and fixedly connected; the middle part of the screw 12 is threadedly connected to the center of the slider 13, the right end of the screw 12 is movably connected to the center of the right plate of the slide rail 14, and the left end of the screw 12 is fixedly connected to the output shaft of the motor II15 through the center of the left plate of the slide rail 14; the slider 13 is slidably connected to the slide rail 14.

[0008] The clamping device C is composed of a wrist joint assembly C1, a pneumatic assembly C2 and a finger assembly pair C3, wherein the wrist joint assembly C1 is composed of a base 16, an air hole group 17, a flexible telescopic assembly C1a, a connecting seat group 18, a connecting head group 19 and a front plate 20, wherein the flexible telescopic member group C1a is composed of five flexible telescopic members with the same structure, each flexible telescopic member is composed of a distance measurer 21, an air duct 22, a hollow plate group 23 and an angle measurer 24, the hollow plate group 23 is composed of 8-12 hollow plates, the 8-12 hollow plates are arranged front and back, the hollow plates are disc-shaped, and the air duct 22 is located at the center of the hollow plate group 23; the distance measurer 21 is fixed to the front of the front end plate of the hollow plate group 23; the angle measurer The device 24 is fixed to the rear end plate of the hollow plate group 23; the pneumatic assembly C2 consists of a shell 25, a connecting plate Ⅰ26, a connecting plate Ⅱ27, a left plate 32, a rod pair 33, a threaded hole pair Ⅳ34, a cylinder 35, an airway hole pair Ⅱ36, a right plate 37, and a threaded hole pair Ⅴ38. The shell 25 is a front-open rectangular box, on which the upper plate is provided with a threaded hole pair Ⅰ28, the right plate is provided with an airway hole pair Ⅰ29, the rear plate is provided with a threaded hole pair Ⅱ30, and the lower plate is provided with a threaded hole pair Ⅲ31; the rear of the cylinder 35 is provided with an airway hole pair Ⅱ36, the lower plate is provided with a threaded hole pair Ⅳ34, and the upper plate is provided with a threaded hole pair Ⅴ38. The left plate 32, the cylinder 35 and the right plate 37 are arranged in sequence from right to left and are connected by the rod pair 33; the connecting plate Ⅰ26 is connected by the horizontal plate Ⅰ26 a and vertical plate Ⅰ 26b are fixedly connected in a T-shape; the rear end of vertical plate Ⅰ 26b of connecting plate Ⅰ 26 is fixedly connected to the front end of the left plate 32 of cylinder 35; connecting plate Ⅱ 27 is fixedly connected in a T-shape by horizontal plate Ⅱ 27a and vertical plate Ⅱ 27b; the rear end of vertical plate Ⅱ 27b of connecting plate Ⅱ 27 is fixedly connected to the front end of the right plate 37 of cylinder 35; the threaded hole pair Ⅳ 34 and threaded hole pair Ⅴ 38 of cylinder 35 are threadedly connected to the threaded hole pair Ⅰ 28 and threaded hole pair Ⅱ 30 of housing 25 by bolts; finger assembly pair C3 is composed of right finger assembly C3a and left finger assembly C3b. The right finger assembly C3a and the left finger assembly C3b have the same structure but opposite directions. Taking the right finger assembly C3a as an example, it is composed of lower plate 39, side plate pair 40, support body 41 and a splint 42, which is a rectangular parallelepiped, the lower end of which is fixedly connected to the splint airway pair 42a, and the inner surface of the splint 42 is provided with six suction cups of suction cup group I 42e, eight suction cups of suction cup group II 42d, eight suction cups of suction cup group III 42c, and four suction cups of suction cup group IV 42b; and the suction cup group I 42e, suction cup group II 42d, suction cup group III 42c and suction cup group IV 42b are arranged radially outward from the center of the splint 42; the lower plate 39, the support body 41 and the splint 42 are arranged in sequence from back to front, and the lower plate 39 and the support body 41 are fixedly connected by bolts to the side plate pair 40; the lower plate 39 is formed by the vertical plate III 39a and the horizontal plate III 39b being fixed in a T-shape; the right side of the splint 42 is fixedly connected to the upper left side of the support body 41.

[0009] The visual device D is composed of a depth camera 43 and an image transmission module 44 , and the depth camera 43 and the image transmission module 44 are communicatively connected.

[0010] The edge learning module E is composed of a high-performance edge computing processor 45, a depth image receiving unit 46 and a signal output unit 47, and the high-performance edge computing processor 45, the depth image receiving unit 46 and the signal output unit 47 are connected by lines.

[0011] For the suction cups on the inner surface of the clamping plate 42, take the suction cups of the suction cup group IV 42b as an example. The bottom end of the inner contour line is the origin, the left side of the suction cup is the X-axis, and the direction perpendicular to the X-axis is the Y-axis. The equation of the curve is:

[0012] Y1=0.7691-3.203X1+5.355X1 2

[0013] Y2=0.4657+0.866X2-7.318X2 2

[0014] Among them, when the values ​​of X1 are 3.975-5.083mm, 2.683-3.658mm, 1.115-2.287mm, and 0.782-1.082, they correspond to suction cup group I 42e, suction cup group II 42d, suction cup group III 42c, and suction cup group IV 42b, respectively;

[0015] When the values ​​of X2 are 5.437-6.782mm, 4.562-5.257mm, 3.053mm-4.491mm, and 2.752-2.959, they correspond to the sizes of suction cup group I 42e, suction cup group II 42d, suction cup group III 42c, and suction cup group IV 42b, respectively.

[0016] When the base 16 of the wrist joint assembly C1 remains stationary, the angle between the bending plane of the end of the flexible telescopic component group C1a and the initial plane is θ, the curvature radius R of the flexible telescopic component group C1a when bent, the installation radius r of the flexible telescopic component, the telescopic length △L of the flexible telescopic component, and the torsion angle around the Z axis are Initial length L and working length L of flexible telescopic component Ⅰ The relationship between them is:

[0017]

[0018] Li=θRi

[0019] in: The range of the angle is -10°-10°, and the range of the θ angle is -50°—50°.

[0020] A picking method based on a picking robot with learning ability of the present invention comprises the following steps:

[0021] 1) For unstructured environments, the visual device D collects depth image information of the technician picking. Using a deep learning method based on key point detection, it collects and processes the key motion features of the human hand. The edge learning module E analyzes the key motion features of the human hand, analyzing the combined pulling, twisting, bending, and dragging behaviors, and saves them.

[0022] 2) The visual device D collects image information of the target fruit. The edge learning module E, equipped with a lightweight neural network, can quickly recognize the picking object and automatically calculate the position between the picking target and the depth camera, and obtain the relative position relationship between the picking target and the arm;

[0023] 3) The edge learning module E analyzes the depth image information collected by the visual device D during the technician's picking, transmits the analyzed behavioral data to the control cabinet F, and drives the clamping device C, stem support device B, and air pump G to complete the specified picking behavior;

[0024] 4) After receiving the signal from the control cabinet F, the air pump G controls the movement of the clamping device C and the stem supporting device B by changing the airflow.

[0025] The beneficial effect of the present invention is that, driven by the elephant trunk wrist joint assembly, the picking robot moves more freely, can adopt a variety of different behaviors to complete the picking operation, and adapt to more complex non-structural environments; the fingers designed under the bionic octopus concept have stronger adsorption force and can better grasp the fruit; they have learning ability and can learn the picking movements of human hands before the picking behavior begins, so as to complete the picking operation more efficiently. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] Figure 1 Axonometric drawing of a learning-capable picking robot.

[0027] Figure 2 This is a partial appearance diagram of the mounting base A;

[0028] Figure 3 It is a partial cross-sectional view of the mounting seat A;

[0029] Figure 4 Schematic diagram of the internal structure of the mounting base A;

[0030] Figure 5 This is the axonometric view of the stem support device B;

[0031] Figure 6 is an axonometric view of the clamping device C;

[0032] Figure 7Axonometric drawing of wrist joint component C1

[0033] Figure 8 is an axonometric view of the flexible telescopic component C1a;

[0034] Figure 9 Appearance of the pneumatic component C2 Figure 1 ;

[0035] Figure 10 is an axonometric view of the housing 25;

[0036] Figure 11 It is an axonometric view of cylinder 35;

[0037] Figure 12 It is an axonometric view of the cylinder 35, connecting plate I 26, and connecting plate II 27;

[0038] Figure 13 It is an axonometric drawing of connecting plate I 26 and connecting plate II 27;

[0039] Figure 14 Axonometric drawing of C3 for the finger assembly;

[0040] Figure 15 This is the axonometric view of the right finger assembly C3a;

[0041] Figure 16 It is an axonometric view of the lower plate 39;

[0042] Figure 17 It is the 40 degree axonometric drawing of the side panel;

[0043] Figure 18 It is an axonometric view of the support body 41;

[0044] Figure 19 It is an axonometric view of the splint 42;

[0045] Figure 20 A schematic diagram of the composition of the visual device D;

[0046] Figure 21 Schematic diagram of the edge learning module E;

[0047] Figure 22 This is a half-section view of the bionic suction cup assembly;

[0048] Figure 23 This is a schematic diagram of the circumferential distribution of the flexible telescopic component C1a;

[0049] Figure 24 This is a schematic diagram of the motion trajectory of wrist joint component C1;

[0050] Figure 25 A flowchart of a picking method for a picking robot with learning ability;

[0051] Wherein: A. Mounting base B. Stem support device C. Clamping device C1. Wrist joint assembly C1a. Flexible telescopic assembly C2. Pneumatic assembly C3. Finger assembly pair C3a. Right finger assembly C3b. Left finger assembly D. Vision device E. Edge learning module F. Control cabinet G. Air pump 1. Cover 2. Support plate 3. Bearing 4. Motor I 5. Driving pulley 6. Synchronous belt 7. Driven pulley 8. Finger 9. Rod I 10. Pneumatic push rod 11. Rod II 12. Screw 13. Slider 14. Slide rail 15. Motor II 16. Base 17. Air hole assembly 18. Connecting seat assembly 19. Connecting head assembly 20. Front plate 21. Distance measuring device 22. Air duct 23. Hollow plate assembly 24 Angle measuring device 25. Housing 26. Connecting plate Ⅰ26a. Horizontal plate Ⅰ26b. Vertical plate Ⅰ27. Connecting plate Ⅱ27a. Horizontal plate Ⅱ27b. Vertical plate Ⅱ28. Threaded hole pair Ⅰ29. Airway hole pair Ⅰ30. Threaded hole pair Ⅱ31. Threaded hole pair Ⅲ32. Left plate 33. Rod pair 34. Threaded hole pair Ⅳ35. Cylinder 36. Airway hole pair Ⅱ37. Right plate 38. Threaded hole pair Ⅴ39. Lower plate 39a. Vertical plate Ⅲ39b. Horizontal plate Ⅲ40. Side plate pair 41. Support body 42. Clamp 42a. Clamp airway pair 42e. Suction cup group Ⅰ42d. Suction cup group Ⅱ42c. Suction cup group III42b. Suction cup group IV43. Depth camera 44. Image transmission module 45. High-performance edge computing processor 46. Depth image receiving unit 47. Signal output unit DETAILED DESCRIPTION

[0052] The present invention will be described below with reference to the accompanying drawings.

[0053] like Figure 1 As shown, a picking robot with learning ability of the present invention is composed of a mounting base A, a stem supporting device B, a clamping device C, a visual device D, an edge learning module E, a control cabinet F and an air pump G, wherein the front plate of the cover 1 of the mounting base A is fixedly connected to the rear plate of the slide rail 14 of the stem supporting device B by bolts; the driving wheel 5 of the mounting base A is movably connected to the base 16 of the wrist joint component C1 in the clamping device C through a transmission shaft; the depth camera 43 of the visual device D is fixed to the support plate 2 of the mounting base A; the edge learning module E is communicated with the visual device D; the control cabinet F is communicated with the mounting base A, the stem supporting device B, the clamping device C, the visual device D, the edge learning module E and the air pump G respectively; the air pump G is connected to the pneumatic push rod 10 of the stem supporting device B, the air hole group 17 and the airway 22 of the wrist joint component C1, the airway hole pair Ⅱ36 of the pneumatic component C2, and the splint airway pair 42a of the finger component C3 through the air pipe.

[0054] like Figures 2 to 4As shown, the mounting seat A is composed of a cover shell 1, a support plate 2, a bearing 3, a motor I4, a driving wheel 5, a synchronous belt 6 and a driven wheel 7. The support plate 2 is fixed to the upper end of the cover shell 1, and the motor I4 is fixed to the cover shell 1; the driving wheel 5, the synchronous belt 6 and the driven wheel 7 are located inside the cover shell 1, and the driving wheel 5 is connected to the driven wheel 7 via the synchronous belt 6; the outer ring of the bearing 3 is fixed to the lower part of the rear plate of the cover shell 1; the motor I4 is fixed to the lower part of the front plate of the cover shell 1, the middle part of the output shaft of the motor I4 is connected to the center key of the driving wheel 5, and the rear end of the output shaft of the motor I4 is interference connected to the inner ring of the bearing 3.

[0055] like Figure 5 As shown, the stem supporting device B is composed of a finger 8, a rod I9, a pneumatic push rod 10, a rod II11, a screw 12, a slider 13, a slide rail 14 and a motor II15. The finger 8, the rod I9, the pneumatic push rod 10, the rod II11 and the slider 13 are arranged in sequence from front to back and fixedly connected; the middle part of the screw 12 is threadedly connected to the center of the slider 13, the right end of the screw 12 is movably connected to the center of the right plate of the slide rail 14, and the left end of the screw 12 is fixedly connected to the output shaft of the motor II15 through the center of the left plate of the slide rail 14; the slider 13 is slidably connected to the slide rail 14.

[0056] like Figures 6 to 19As shown, the clamping device C is composed of a wrist joint component C1, a pneumatic component C2 and a finger component pair C3, wherein the wrist joint component C1 is composed of a base 16, an air hole group 17, a flexible telescopic component C1a, a connecting seat group 18, a connecting head group 19 and a front plate 20, wherein the flexible telescopic component group C1a is composed of five flexible telescopic components with the same structure, each flexible telescopic component is composed of a distance measurer 21, an air duct 22, a hollow plate group 23 and an angle measurer 24, the hollow plate group 23 is composed of 8-12 hollow plates, the 8-12 hollow plates are arranged front and back, the hollow plates are disc-shaped, and the air duct 22 is arranged at the center of the hollow plate group 23; the distance measurer 21 is fixed to the front of the front end plate of the hollow plate group 23; the angle The degree measuring device 24 is fixed to the rear end plate of the hollow plate group 23; the pneumatic assembly C2 consists of a housing 25, a connecting plate Ⅰ26, a connecting plate Ⅱ27, a left plate 32, a rod pair 33, a threaded hole pair Ⅳ34, a cylinder 35, an airway hole pair Ⅱ36, a right plate 37, and a threaded hole pair Ⅴ38. The housing 25 is a front-open rectangular box, on which the upper plate is provided with a threaded hole pair Ⅰ28, the right plate is provided with an airway hole pair Ⅰ29, the rear plate is provided with a threaded hole pair Ⅱ30, and the lower plate is provided with a threaded hole pair Ⅲ31; the rear of the cylinder 35 is provided with an airway hole pair Ⅱ36, the lower plate is provided with a threaded hole pair Ⅳ34, and the upper plate is provided with a threaded hole pair Ⅴ38. The left plate 32, the cylinder 35 and the right plate 37 are arranged in sequence from right to left and are connected by the rod pair 33; the connecting plate Ⅰ26 is connected by the horizontal plate Ⅰ 26a and vertical plate I 26b are fixedly connected in a T-shape; the rear end of vertical plate I 26b of connecting plate I 26 is fixedly connected to the front end of the left plate 32 of the cylinder 35; connecting plate II 27 is fixedly connected in a T-shape by horizontal plate II 27a and vertical plate II 27b; the rear end of vertical plate II 27b of connecting plate II 27 is fixedly connected to the front end of the right plate 37 of the cylinder 35; the threaded hole pair IV 34 and threaded hole pair V 38 of the cylinder 35 are threadedly connected to the threaded hole pair I 28 and threaded hole pair II 30 of the housing 25 by bolts; the finger assembly pair C3 is composed of a right finger assembly C3a and a left finger assembly C3b. The right finger assembly C3a and the left finger assembly C3b have the same structure but opposite directions. Taking the right finger assembly C3a as an example, it is composed of a lower plate 39, a side plate pair 40, and a support body 4 1 and a splint 42, the splint 42 is a rectangular parallelepiped, the lower end of the splint 42 is fixed to the splint airway pair 42a, the inner surface of the splint 42 is provided with six suction cups of suction cup group I 42e, eight suction cups of suction cup group II 42d, eight suction cups of suction cup group III 42c, and four suction cups of suction cup group IV 42b; and the suction cup group I 42e, suction cup group II 42d, suction cup group III 42c and suction cup group IV 42b are arranged radially outward from the center of the splint 42; the lower plate 39, the support body 41 and the splint 42 are arranged in sequence from back to front, and the lower plate 39 and the support body 41 are fixedly connected by bolts to the side plate pair 40; the lower plate 39 is formed by the vertical plate III 39a and the horizontal plate III 39b being fixed in a T-shape; the right side of the splint 42 is fixedly connected to the upper left side of the support body 41.

[0057] like Figure 20As shown, the visual device D is composed of a depth camera 43 and an image transmission module 44, and the depth camera 43 and the image transmission module 44 are communicatively connected.

[0058] like Figure 21 As shown, the edge learning module E is composed of a high-performance edge computing processor 45, a depth image receiving unit 46 and a signal output unit 47, and the high-performance edge computing processor 45, the depth image receiving unit 46 and the signal output unit 47 are connected by lines.

[0059] like Figure 22 As shown, the suction cups on the inner surface of the clamping plate 42, taking the suction cups of the suction cup group IV 42b as an example, the bottom end of the inner contour line is the origin, the left side of the suction cup is the X-axis, and the direction perpendicular to the X-axis is the Y-axis. The equation of the curve is:

[0060] Y1=0.7691 -3.203X1+5.355X1 2

[0061] Y2=0.4657+0.866X2-7.318X2 2

[0062] Among them, when the values ​​of X1 are 3.975-5.083mm, 2.683-3.658mm, 1.115-2.287mm, and 0.782-1.082, they correspond to suction cup group I 42e, suction cup group II 42d, suction cup group III 42c, and suction cup group IV 42b, respectively;

[0063] When the values ​​of X2 are 5.437-6.782mm, 4.562-5.257mm, 3.053mm-4.491mm, and 2.752-2.959, respectively, they correspond to the sizes of suction cup group I 42e, suction cup group II 42d, suction cup group III 42c, and suction cup group IV 42b.

[0064] like Figure 23 and Figure 24 As shown, when the base 16 of the wrist joint assembly C1 remains stationary, the angle between the bending plane of the end of the flexible telescopic component group C1a and the initial plane is θ, the curvature radius R of the flexible telescopic component group C1a when bent, the installation radius r of the flexible telescopic component, the telescopic length △L of the flexible telescopic component, and the torsion angle around the Z axis are Initial length L and working length L of flexible telescopic component Ⅰ The relationship between them is:

[0065]

[0066] Li=θRi

[0067] in: The range of the angle is -10° to -10°, and the range of the θ angle is -50° to 50°.

[0068] like Figure 25 As shown, a picking method based on a picking robot with learning ability of the present invention includes the following steps:

[0069] 1) For unstructured environments, the visual device D collects depth image information of the technician picking. Using a deep learning method based on key point detection, it collects and processes the key motion features of the human hand. The edge learning module E analyzes the key motion features of the human hand, analyzing the combined pulling, twisting, bending, and dragging behaviors, and saves them.

[0070] 2) The visual device D collects image information of the target fruit. The edge learning module E, equipped with a lightweight neural network, can quickly recognize the picking object and automatically calculate the position between the picking target and the depth camera, and obtain the relative position relationship between the picking target and the arm;

[0071] 3) The edge learning module E analyzes the depth image information collected by the visual device D during the technician's picking, transmits the analyzed behavioral data to the control cabinet F, and drives the clamping device C, stem support device B, and air pump G to complete the specified picking behavior;

[0072] 4) After receiving the signal from the control cabinet F, the air pump G controls the movement of the clamping device C and the stem supporting device B by changing the airflow.

Claims

1. A picking robot with learning ability, characterized by It consists of a mounting base (A), a stem supporting device (B), a clamping device (C), a visual device (D), an edge learning module (E), a control cabinet (F) and an air pump (G). The mounting base (A) consists of a cover (1), a support plate (2), a bearing (3), a motor I (4), a driving wheel (5), a synchronous belt (6) and a driven wheel (7); the stem supporting device (B) consists of a finger (8), a rod I (9), a pneumatic push rod (10), a rod II (11), a screw (12), a slider (13), a slide rail (14) and a motor II (15); the clamping device (C) consists of a wrist joint assembly (C1), a pneumatic assembly (C2) and a pair of finger assemblies (C3); the wrist joint assembly (C1) consists of a base (16) , air hole group (17), flexible telescopic component (C1a), connecting seat group (18), connecting head group (19) and front plate (20), pneumatic component (C2) is composed of housing (25), connecting plate I (26), connecting plate II (27), left plate (32), rod pair (33), threaded hole pair IV (34), cylinder (35), airway hole pair II (36), right plate (37), threaded hole pair V (38), finger assembly pair (C3) is composed of right finger assembly (C3a) and left finger assembly (C3b), the right finger assembly (C3a) and the left finger assembly (C3b) are of the same structure but opposite directions. Taking the right finger assembly (C3a) as an example, it is composed of lower plate (39), side plate pair (40), support body (41) and splint (42 ); the visual device (D) is composed of a depth camera (43) and an image transmission module (44); the edge learning module (E) is composed of a high-performance edge computing processor (45), a depth image receiving unit (46) and a signal output unit (47); the front plate of the cover (1) of the mounting seat (A) is fixed to the rear plate of the slide rail (14) of the stem support device (B) by bolts; the driving wheel (5) of the mounting seat (A) is movably connected to the base (16) of the wrist joint component (C1) in the clamping device (C) through a transmission shaft; the flexible telescopic component (C1a) in the clamping device (C) is composed of five flexible telescopic parts with the same structure, each of which is composed of a distance measuring device (21), an airway (22), a hollow plate group ( 23) and an angle measuring device (24), the hollow plate group (23) is composed of 8-12 hollow plates, the 8-12 hollow plates are arranged in a front-to-back manner, the hollow plates are disc-shaped, and the airway (22) is located at the center of the hollow plate group (23); the distance measuring device (21) is fixed to the front of the front end plate of the hollow plate group (23); the angle measuring device (24) is fixed to the back of the rear end plate of the hollow plate group (23); the depth camera (43) of the visual device (D) is fixed to the support plate (2) of the mounting seat (A); the edge learning module (E) is connected to the visual device (D); the control cabinet (F) is connected to the mounting seat (A), the stem supporting device (B), the clamping device (C), the visual device (D), the edge learning module (E) and the air pump (G) respectively;The air pump (G) is connected to the pneumatic push rod (10) of the stem support device (B), the air hole group (17) and the airway (22) of the wrist joint component (C1), the airway hole pair II (36) of the pneumatic component (C2), and the splint airway pair (42a) of the finger component pair (C3) through the air tube.

2. The learning-capable picking robot according to claim 1 is characterized in that The support plate (2) in the mounting seat (A) is fixed to the upper end of the housing (1), and the motor I (4) is fixed to the housing (1); the driving wheel (5), the synchronous belt (6) and the driven wheel (7) are located in the housing (1), and the driving wheel (5) is connected to the driven wheel (7) through the synchronous belt (6); the outer ring of the bearing (3) is fixed to the lower part of the rear plate of the housing (1); the motor I (4) is fixed to the lower part of the front plate of the housing (1), the middle part of the output shaft of the motor I (4) is connected to the center key of the driving wheel (5), and the rear end of the output shaft of the motor I (4) is interference-connected to the inner ring of the bearing (3).

3. The learning-capable harvesting robot according to claim 1, characterized in that In the stem supporting device (B), the finger (8), rod I (9), pneumatic push rod (10), rod II (11) and slider (13) are arranged in sequence from front to back and fixedly connected; the middle part of the screw (12) is threadedly connected to the center of the slider (13), the right end of the screw (12) is movably connected to the center of the right plate of the slide rail (14), and the left end of the screw (12) is fixedly connected to the output shaft of the motor II (15) through the center of the left plate of the slide rail (14); the slider (13) is slidably connected to the slide rail (14).

4. The learning-capable harvesting robot according to claim 1, characterized in that The housing (25) of the pneumatic assembly (C2) in the clamping device (C) is a front-open rectangular box, wherein the upper plate is provided with a threaded hole pair I (28), the right plate is provided with an airway hole pair I (29), the rear plate is provided with a threaded hole pair II (30), and the lower plate is provided with a threaded hole pair III (31); the rear of the cylinder (35) is provided with an airway hole pair II (36), the lower plate is provided with a threaded hole pair IV (34), the upper plate is provided with a threaded hole pair VV (38), and the left plate (32), the cylinder (35) and the right plate (37) are arranged from right to left in sequence. The connecting plate Ⅰ (26) is formed by the horizontal plate Ⅰ (26a) and the vertical plate Ⅰ (26b) being fixedly connected in a T-shape; the rear end of the vertical plate Ⅰ (26b) of the connecting plate Ⅰ (26) is fixedly connected to the front end of the left plate (32) of the cylinder (35); the connecting plate Ⅱ (27) is formed by the horizontal plate Ⅱ (27a) and the vertical plate Ⅱ (27b) being fixedly connected in a T-shape; the rear end of the vertical plate Ⅱ (27b) of the connecting plate Ⅱ (27) is fixedly connected to the front end of the right plate (37) of the cylinder (35); the threaded hole pair Ⅳ of the cylinder (35) is fixedly connected to the front end of the right plate (37) of the cylinder (35). (34) and threaded hole pair V (38) are threadedly connected to threaded hole pair I (28) and threaded hole pair II (30) of the housing (25) through bolts; the splint (42) of the finger assembly pair (C3) is a rectangular parallelepiped, the lower end of the splint (42) is fixedly connected to the splint airway pair (42a), and the inner surface of the splint (42) is provided with six suction cups of suction cup group I (42e), eight suction cups of suction cup group II (42d), eight suction cups of suction cup group III (42c), and four suction cups of suction cup group IV (42b); and the suction cup group Ⅰ (42e), suction cup group II (42d), suction cup group III (42c) and suction cup group IV (42b) are arranged radially outward from the center of the clamping plate (42); the lower plate (39), the support body (41) and the clamping plate (42) are arranged in sequence from back to front, and the lower plate (39) and the support body (41) are fixedly connected by bolts of side plate pair (40); the lower plate (39) is formed by T-shaped fixing of vertical plate III (39a) and horizontal plate III (39b); the right side of the clamping plate (42) is fixedly connected to the upper left side of the support body (41).

5. The learning-capable harvesting robot according to claim 1, characterized in that The depth camera (43) and the image transmission module (44) of the visual device (D) are communicatively connected; the high-performance edge computing processor (45), the depth image receiving unit (46) and the signal output unit (47) of the edge learning module (E) are connected by lines.

6. The learning-capable harvesting robot according to claim 4, characterized in that The suction cup on the inner surface of the clamping plate (42) is taken as an example of the suction cup of the suction cup group IV (42b). The lower end of the inner contour line is the origin, the left side of the suction cup is the X axis, and the direction perpendicular to the X axis is the Y axis. The equation of the curve is: Y1=0.7691-3.203X1+5.355X1 2 Y2=0.4657+0.866X2-7.318X2 2 Among them: when the values ​​of X1 are 3.975-5.083mm, 2.683-3.658mm, 1.115-2.287mm, and 0.782-1.082, they correspond to the sizes of suction cup group I (42e), suction cup group II (42d), suction cup group III (42c), and suction cup group IV (42b), respectively; when the values ​​of X2 are 5.437-6.782mm, 4.562-5.257mm, 3.053mm-4.491mm, and 2.752-2.959, they correspond to the sizes of suction cup group I (42e), suction cup group II (42d), suction cup group III (42c), and suction cup group IV (42b), respectively.

7. The picking robot with learning ability according to claim 4, characterized in that: When the base (16) of the wrist joint assembly (C1) remains stationary, the angle between the bending plane at the end of the flexible telescopic assembly (C1a) and the initial plane is θ, the curvature radius R of the flexible telescopic assembly (C1a) when bent, the installation radius r of the flexible telescopic assembly, the telescopic length △L of the flexible telescopic assembly, and the torsion angle around the Z axis are Initial length L and working length L of flexible telescopic component i The relationship between them is: Li=θRi in: The range of the angle is -10°-10°, and the range of the θ angle is -50°—50°.

8. A picking method based on the picking robot with learning ability according to claim 1, characterized in that: The steps include: 1) For unstructured environments, the visual device (D) collects depth image information of technicians picking. Using a deep learning method based on key point detection, it collects and processes key hand motion features. The edge learning module (E) analyzes these key hand motion features, analyzing the combined pulling, twisting, bending, and dragging behaviors of the hand, and saves them. 2) The vision device (D) collects image information of the target fruit. The edge learning module (E) is equipped with a lightweight neural network to quickly identify the picking object and automatically calculate the position between the picking target and the depth camera, thereby obtaining the relative position relationship between the picking target and the arm. 3) The edge learning module (E) analyzes the depth image information collected by the visual device (D) while the technician is picking, transmits the analyzed behavioral data to the control cabinet (F), and drives the clamping device (C), stem support device (B), and air pump (G) to complete the specified picking behavior; 4) The air pump (G) controls the movement of the clamping device (C) and the stem support device (B) by changing the air flow after receiving the signal from the control cabinet (F).

Images