Fruit picking equipment based on modular design

Abstract

The invention relates to fruit picking equipment based on modular design, the fruit picking equipment comprises a mechanical arm and an end effector, the mechanical arm comprises a connecting arm and a plurality of bionic joints, each bionic joint comprises a front joint module and a rear joint module which rotate relatively, and the rear joint module is driven by a rotation driving mechanism arranged on the front joint module to rotate; each joint module comprises an L-shaped connecting plate composed of two orthogonally-arranged fixing plates. The rotary driving mechanism comprises a motor, a speed reducer and an output flange which are in transmission connection; one of two fixing plates of the L-shaped connecting plate is connected with the output flange and rotates by taking a motor output shaft as a rotating shaft, and the other fixing plate is connected with the end effector and the connecting arm or is provided with a motor of an adjacent joint module; or one is connected with the connecting arm, and the other is provided with the motor of the adjacent joint module. The problems that an existing picking device cannot be modularly installed, the degree of freedom is fixed, and the application scene is single are solved.

Description

Technical Field

[0001] This invention relates to the field of harvesting robotic arms, and in particular to a fruit harvesting device based on modular design. Background Technology

[0002] Harvesting of fruits and vegetables such as tomatoes has long relied on manual labor, resulting in high labor intensity, low efficiency, and high costs. Furthermore, uneven handling during manual harvesting can easily damage the fruit, affecting its commercial value. With the modernization of agriculture, harvesting robots have become a key solution to these problems.

[0003] However, existing robotic arms for tomato harvesting suffer from a lack of versatility and modularity in their joint design, resulting in fixed degrees of freedom in operation and limiting their application scenarios, failing to fully meet the needs of various work environments. Furthermore, the end effector is prone to damaging the tomato skin and picking non-target tomatoes, leading to poor adaptability and low harvesting accuracy. Summary of the Invention

[0004] To address the shortcomings of existing technologies, this invention provides a fruit picking device based on modular design, which solves the technical problems of existing picking devices having fixed degrees of freedom and limited applicable scenarios due to their inability to be modularly installed.

[0005] The technical solution adopted in this invention is as follows: The present invention provides a fruit picking device based on modular design, including a robotic arm and an end effector. The robotic arm includes a connecting arm and several bionic joints. The bionic joints include two relatively rotating joint modules, one in front and one behind, and the latter joint module is driven to rotate by a rotation drive mechanism disposed on the former joint module. The joint module includes an L-shaped connecting plate consisting of two orthogonally arranged fixing plates; The rotary drive mechanism includes a motor, a reducer, and an output flange that are connected by a transmission. The L-shaped connecting plate has two fixing plates, one of which is connected to the output flange and rotates with the motor output shaft as the rotation axis, and the other is connected to the end effector, connecting arm, or motor with adjacent joint modules; or, one is connected to the connecting arm and the other is connected to the motor with adjacent joint modules. The connecting arm includes a first connecting arm and a second connecting arm. The first connecting arm is connected to a first bionic joint and a second bionic joint at its two ends, respectively. The first and second bionic joints simulate the shoulder joint and the elbow joint, respectively. The first bionic joint has two joint modules, one of which is connected to the robot body and the other of which is connected to the first connecting arm; the second bionic joint has two joint modules, one of which is connected to the first connecting arm and the other of which is connected to one end of the second connecting arm. The other end of the second connecting arm is connected to a third bionic joint for simulating a wrist joint, which is connected to the end effector through an end joint module to form a six-degree-of-freedom mechanism; or the second connecting arm is directly connected to the end effector through the end joint module to form a four-degree-of-freedom mechanism. The robotic arm is equipped with a vision module, which is used to collect image information of the target fruit and transmit it to the host computer, providing a basis for the host computer to control the posture of the robotic arm and the movement of the end effector.

[0006] The preferred technical solution is as follows: The host computer outputs control commands, including: The fruit position is obtained based on the target fruit image, and the expected values ​​of the angle and rotational angular velocity of each joint module are calculated based on the fruit position. The actual values ​​of the angle and rotational angular velocity of each joint module are read and compared with the corresponding expected values ​​to obtain the tracking error; A time-varying performance function is set for the tracking error to define the upper and lower bounds of the tracking error. The host computer runs a fixed-time inversion adaptive control algorithm to perform hierarchical calculations. The outer loop generates a virtual angular velocity command that ensures the angle error converges within a fixed time. The inner loop, based on the robotic arm dynamics model and online adaptive disturbance estimation, solves for the actual control torque required to drive the motors of each joint module. The host computer outputs the actual control torque command required by the motor to the corresponding motor controller and ensures that all tracking errors are constrained by the performance function to achieve trajectory tracking.

[0007] The expected values ​​of the angles and rotational angular velocities of each joint module calculated based on the fruit's position include: The target detection algorithm based on deep neural networks detects the bounding box and depth image of the target fruit, calculates the three-dimensional spatial position of the center point of the fruit image, and then calculates the position of the fruit's center by combining the camera coordinates and the known average radius of the fruit. Identify the connection point between the fruit stalk and the tomato. P p Based on the position of the fruit's center and P p The vector formed serves as the orientation of the target fruit. Based on the position of the fruit's center and the orientation of the target fruit, the robot's end-effector grasping posture is determined. Using the inverse kinematics model of the robotic arm, the angles that each joint module of the robotic arm needs to achieve are calculated, i.e., the expected values ​​of the joint angles. By combining the planned trajectory from the initial position to the grasping position, the expected values ​​of the angular velocities of each joint are obtained by differentiating the expected values ​​of the joint angles.

[0008] The host computer outputs control commands, which also include: When the end effector has moved to the pre-grabbing position and the error between the actual pose and the expected grasping pose is less than the preset threshold, the host computer controls the end effector to trigger and complete the grasping action.

[0009] The performance function specifies the upper and lower bounds of the allowable tracking error, including: Divide the original tracking error by the performance function to obtain the standardized error variable; The constrained problem is transformed into a stabilization problem by using the standardized variables, that is, the standardized error variables converge to the set value, thereby achieving the predetermined steady-state and transient performance.

[0010] The end effector includes a cylinder, the cylinder body of which is connected to the output flange of the joint module; the piston output end of the cylinder is connected to a turntable to drive its rotation. The bionic palm part of the turntable has three connectors evenly arranged around its upper circumference. The upper two ends of the connectors are respectively hinged to the cylinder body and the turntable, and the lower side of the connectors is hinged to one end of the bionic fingertip. The other end of the bionic fingertip is provided with a bionic fingertip, and a roller is provided on its inner side. The bionic fingertip is equipped with a force sensor that collects pressure data, and the host computer is connected to the force sensor to control the opening and closing of the end effector.

[0011] The opening and closing control methods of the end effector include: The host computer uses a model predictive control algorithm to compare the received real-time pressure data with a preset pressure damage threshold, and generates a prediction model based on the state space model of the dynamic system. Taking the real-time pressure data as the initial state, it predicts the pressure evolution trajectory under different control strategies in the future finite time domain, and then calculates the optimal adjustment command for the finger opening and closing angle through online optimization. The finger opening and closing angle adjustment command is sent to the cylinder's drive actuator, which adjusts the angle of the bionic fingertip so that while the roller is continuously in contact with the surface of the target fruit, the contact pressure between the two is controlled below a safe threshold. If the coverage area exceeds the preset single fruit size or abnormal contact pressure is detected, the host computer will control the interruption of the closing action.

[0012] The two joint modules of the third bionic joint are connected, with the first one connected to the second connecting arm and the second one connected to the end joint module, and they rotate relative to each other along two orthogonal rotation axes.

[0013] The first joint module of the first bionic joint is connected to the robot body via a base; Based on the six-degree-of-freedom mechanism, a rotation module is set between the robot body and the base to form a seven-degree-of-freedom mechanism.

[0014] The structure of the joint module also includes a housing, which is connected to the L-shaped connecting plate and covers the outside of the L-shaped connecting plate.

[0015] The beneficial effects of this invention are as follows: The joint module structure of this invention adopts a modular structure based on an L-shaped connecting plate, which facilitates assembly according to harvesting needs and usage scenarios to form different types of bionic joints, thereby enabling flexible adjustment of the robotic arm's degrees of freedom. It is easy to assemble and disassemble and has a wide range of applications.

[0016] In this invention, the two joint modules of each bionic joint rotate relative to each other through a rotary drive mechanism. The rotary drive mechanism is simple in structure, easy to process, easy to install, low in cost, and easy to adjust.

[0017] This invention performs path planning based on the target fruit's location, enabling robotic arm posture adjustment and end effector control. The vision module, along with the host computer's trajectory planning and end effector control, forms a closed-loop collaborative system. The vision module outputs fruit information in real time, calculates the optimal grasping angle and path using the robotic arm's kinematic model, and dynamically corrects the path through force sensor feedback, thus forming a closed-loop feedback control. Compared to existing harvesting robots that perform single-path planning after recognizing fruit information, this invention, through closed-loop feedback control, can automatically correct the grasping path in cases of fruit occlusion or posture deviation, thereby significantly improving harvesting accuracy and success rate.

[0018] The end effector of this invention adopts an enveloping structure, achieving flexible harvesting of tomato fruits through biomimetic design. The end effector employs a collaborative strategy of force sensors and visual recognition: the vision system locks onto a single target, while the force sensor of the end effector detects the contact state in real time. If the enveloping area exceeds the size of a single fruit or abnormal contact pressure is detected, the closing action is automatically interrupted. This achieves adaptive adhesion to the fruit surface, overcomes interference from non-target fruits, and effectively solves the problems of existing tomato harvesting end effectors being easily interfered with and swallowing multiple fruits.

[0019] Other features and advantages of the invention will be set forth in the following description or may be learned by practicing the invention. Attached Figure Description

[0020] Figure 1 This is a three-dimensional structural diagram of Embodiment 1 of the present invention.

[0021] Figure 2 for Figure 1 A schematic diagram of the structure after the outer shell is disassembled.

[0022] Figure 3 for Figure 2 A further exploded diagram of the first joint module.

[0023] Figure 4 for Figure 2 A further breakdown diagram.

[0024] Figure 5 This is a schematic diagram of the end effector in an embodiment of the present invention.

[0025] Figure 6 This is a schematic diagram of the structure of Embodiment 2 of the present invention.

[0026] In the diagram: 1. Robotic arm; 2. End effector; 3. Base; 4. First housing; 5. Second housing; 6. First connecting arm; 7. Third housing A; 8. Third housing B; 9. Second connecting arm; 10. Fourth housing; 11. Fifth housing; 12. Sixth housing; 13. Visual recognition camera; 14. First motor; 15. First harmonic reducer; 16. First output flange; 17. First L-shaped connecting plate; 18. Second motor; 19. Second harmonic reducer; 20. Second output flange; 21. Second L-shaped connecting plate; 22. Third L-shaped connecting plate A; 23. 24. Third motor; 25. Third harmonic reducer; 26. Third output flange; 27. Third L-shaped connecting plate B; 28. Fourth L-shaped connecting plate; 29. ​​Fourth harmonic reducer; 30. Fourth output flange; 31. Fifth L-shaped connecting plate; 32. Fifth motor; 33. Fifth harmonic reducer; 34. Fifth output flange; 35. Sixth motor; 36. Sixth harmonic reducer; 37. Sixth output flange; 38. Sixth L-shaped connecting plate; 39. Cylinder; 40. Piston; 41. Connector; 42. Roller; 43. Bionic fingertip; 44. Bionic finger pad. Detailed Implementation

[0027] The specific embodiments of the present invention are described below with reference to the accompanying drawings.

[0028] The present invention provides a fruit picking device based on modular design, including a robotic arm and an end effector. The robotic arm includes a connecting arm and several bionic joints. The bionic joints include two relatively rotating joint modules, one in front and one behind, and the latter joint module is driven to rotate by a rotation drive mechanism disposed on the former joint module. The joint module includes an L-shaped connecting plate consisting of two orthogonally arranged fixing plates; The rotary drive mechanism includes a motor, a reducer, and an output flange that are connected by a transmission. The L-shaped connecting plate has two fixing plates, one of which is connected to the output flange and rotates with the motor output shaft as the rotation axis, and the other is connected to the end effector, connecting arm, or motor with adjacent joint modules; or, one is connected to the connecting arm and the other is a motor with adjacent joint modules.

[0029] The robotic arm of this invention employs a combination of joint modules to form a detachable, modular bionic joint structure. By adjusting the number of connecting arms and bionic joints, it can meet the application requirements of robotic arms with different degrees of freedom, enabling pose changes during the harvesting process. The end effector of this invention is used as a bionic hand to achieve flexible grasping of the target fruit.

[0030] Preferably, the motor, reducer, and output flange are arranged coaxially.

[0031] Preferably, the motor is a miniature AC optomotor; the reducer is a harmonic reducer; the output shaft of the miniature AC optomotor is a D-type milled flat shaft, and it is keyless interference fit with the output hole of the harmonic reducer.

[0032] The technical solution of the present invention will be further illustrated below with specific embodiments. Example 1:

[0033] See Figures 1 to 4 This embodiment provides a fruit picking device based on modular design, including a robotic arm 1 and an end effector 2. The structure of the robotic arm 1 includes a first connecting arm 6, with a bionic joint connected to each end, namely a first bionic joint and a second bionic joint; a second connecting arm 9, with a bionic joint connected to each end, namely a second bionic joint and a third bionic joint; and a sixth joint module, which serves as the end joint module and is used to connect the end effector 2. The above components constitute a six-degree-of-freedom robotic arm.

[0034] Among them, the first, second and third bionic joints are used to simulate the shoulder joint, elbow joint and wrist joint respectively, forming the bionic shoulder, bionic elbow and bionic wrist of the robotic arm respectively; the first connecting arm 6 and the second connecting arm 9 are used to simulate the upper arm and forearm respectively; and the end effector 2 is used to simulate the hand.

[0035] The first bionic joint comprises two joint modules, namely a first joint module and a second joint module. Preferably, the structure of the first joint module is as follows: It includes a base 3, a first rotary drive mechanism consisting of a first motor 14, a first harmonic reducer 15, a first output flange 16, and a first L-shaped connecting plate 17; the first motor 14 is mounted on the base 3, and the first output flange 16 is connected to the first L-shaped connecting plate 17 to drive it to rotate around a first axis, which is coaxial with the output shaft of the first motor 14, thereby realizing the rotation of the first joint module.

[0036] Specifically, the base 3 can be directly connected to the robot body to form a six-degree-of-freedom mechanism, or connected to the robot body through a rotation module to form a seven-degree-of-freedom mechanism.

[0037] Specifically, the first L-shaped connecting plate 17 has two orthogonally arranged fixing plates, one of which is rotatably engaged with the base 2 and connected to the first output flange 16, and the other is used to set the second motor 18 of the second joint module.

[0038] Preferably, the first motor 14 is fixed to the base 3 by screws, and the robot body is preferably a mobile platform or AGV trolley. The bottom of the base 3 can be fixed to the mobile platform or AGV trolley by screws. Specifically, both the robot carrier connecting base and the circular outer shell have through holes for wiring of the motor and control system.

[0039] Preferably, a first outer shell 4 is connected to the first L-shaped connecting plate 17, which covers the entire first joint module, serving both protective and aesthetic purposes.

[0040] The structure of the second joint module is as follows: It includes a second rotary drive mechanism consisting of a second motor 18, a second harmonic reducer 19, a second output flange 20, and a second L-shaped connecting plate 21; the second output flange 20 is connected to the second L-shaped connecting plate 21 to drive it to rotate around a second axis, which is coaxial with the output shaft of the second motor 18 and orthogonal to the first axis.

[0041] Specifically, the two orthogonally arranged fixing plates of the second L-shaped connecting plate 21 are used to connect the second output flange 20 and the other to connect one end of the first connecting arm 6.

[0042] Preferably, a second outer shell 5 is connected to the second L-shaped connecting plate 21, which encloses the entire second joint module.

[0043] The second bionic joint comprises two joint modules, namely the third joint module A and the third joint module B. The connection structure between the two is similar to the connection structure of the first and second joint modules. As a preferred embodiment, the specific connection structure is as follows: The structure of the third joint module A is as follows: it includes a third rotary drive mechanism A consisting of a third motor A23, a third harmonic reducer A24, and a third output flange 25, as well as a third L-shaped connecting plate A22; of the two orthogonal fixed plates of the third L-shaped connecting plate A22, one is connected to the other end of the first connecting arm 6, and the other is used to set the third motor A23; the structure of the third joint module B includes a third L-shaped connecting plate B26, the third output flange 25 is connected to one fixed plate of the third L-shaped connecting plate B26 to drive it to rotate around the second axis, and the other fixed plate of the third L-shaped connecting plate B26 is used to connect one end of the second connecting arm 9.

[0044] Preferably, the third L-shaped connecting plate A22 is externally connected to the third outer shell A7 to cover the third joint module A; the third L-shaped connecting plate B26 is externally connected to the third outer shell B8 to cover the third joint module B.

[0045] The two joint modules of the third bionic joint are the fourth joint module and the fifth joint module, respectively. The connection structure between them is the same as the connection structure of the third joint modules A and B. As a preferred embodiment, the specific connection structure is as follows: The fourth joint module includes a third rotary drive mechanism consisting of a fourth motor 28, a fourth harmonic reducer 29, and a fourth output flange 30, as well as a fourth L-shaped connecting plate 27; one fixing plate of the fourth L-shaped connecting plate 27 is connected to the second connecting arm 9, and the fourth motor 28 is mounted on the other fixing plate. The fifth joint module includes a fifth L-shaped connecting plate 31 and a fifth rotary drive mechanism consisting of a fifth motor 32, a fifth harmonic reducer 33, and a fifth output flange 34; the fourth output flange 30 is connected to one fixed plate of the fifth L-shaped connecting plate 31 to drive it to rotate around the second axis, and the fifth motor 32 is mounted on the other fixed plate.

[0046] As a preferred embodiment, the sixth joint module includes a sixth L-shaped connecting plate 38 and a sixth rotary drive mechanism consisting of a sixth motor 35, a sixth harmonic reducer 36, and a sixth output flange 37; one fixed plate of the sixth L-shaped connecting plate 38 is connected to the fifth output flange 34 and thus rotates around the first axis, the other fixed plate is provided with the sixth motor 35, and the sixth output flange 37 is used to connect with the end effector 2 to drive it to rotate around the second axis.

[0047] Preferably, the fourth outer shell 10, the fifth outer shell 11 and the sixth outer shell 12 are respectively connected to the L-shaped connecting plates of the fourth to sixth joint modules, thereby covering the corresponding joint modules inside.

[0048] Preferably, each of the above-mentioned outer shells is a quarter-sphere shell; each shell is preferably connected to the corresponding L-shaped connecting plate by screws.

[0049] Preferably, the first connecting arm 6 and the second connecting arm 9 are connected to the corresponding L-shaped connecting plates at both ends by screws.

[0050] Preferably, each L-shaped connecting plate is connected to the output flange that drives its rotation by screws.

[0051] Specifically, the lengths of the first and second connecting arms can be adjusted according to actual application requirements, and appropriate motors can be selected based on the length of the connecting arms and the load of the joints.

[0052] The robotic arm in this embodiment is also equipped with a vision module. The vision module preferably includes a vision recognition camera 13, which is preferably disposed on the sixth housing 12 of the sixth joint module. It is used to collect target fruit image information and transmit it to the host computer, so as to provide a basis for the host computer to control the posture of the robotic arm and the action of the end effector 2.

[0053] The six-DOF modular robotic arm in this embodiment possesses three major joints similar to the human arm—shoulder, elbow, and wrist—allowing for arbitrary spatial posture adjustment. It is more suitable for vertical trellis planting and scenarios with densely intertwined branches and leaves, effectively ensuring harvesting accuracy and fruit integrity. The seven-DOF mechanism in this embodiment provides additional redundant degrees of freedom, which can be used for path optimization and obstacle avoidance, making it more suitable for open orchard environments and areas with irregularly distributed tomato plants. Through redundant degrees of freedom, the robotic arm's posture can be flexibly adjusted in complex terrain, improving the working range and efficiency.

[0054] Preferably, the host computer outputs control commands in the following manner: S1. Obtain the fruit position based on the target fruit image, and calculate the expected values ​​of the angles and rotational angular velocities of each joint module based on the fruit position. Specifically, this includes: Fruit localization and center estimation: Based on the YOLO-v8 model, the bounding box and depth image of the target fruit are detected, and the three-dimensional spatial position of the center point of the fruit image is calculated. P b ( x b , y b , z b ), and then combined with camera coordinates P c ( x c , y c , z c ) and the known average radius of the fruit r The three-dimensional spatial coordinates of the fruit's center are calculated using the following formula. P o ( x o , y o , z o ):

[0055] Picking point location and posture determination: Identifying the connection point between the fruit stalk and the tomato. P p Connecting the center of the fruit P oand connection point P p The direction of this vector is taken as the orientation of the tomato. Inverse kinematics solution: Based on the calculated position of the fruit's center and the tomato's orientation vector, the robot's end effector grasping posture is determined using the right-hand rule. Then, using the inverse kinematics model of the robotic arm, the required angles of each joint are calculated, i.e., the expected joint angle values. q d ; Expected angular velocity generation: By combining the planned trajectory from the initial position to the grasping position, the expected value of the angular velocity of each joint is obtained by differentiating the expected joint angle.

[0056] S2. Read the actual values ​​of the angle and rotational angular velocity of each joint module, and compare them with the corresponding expected values ​​to obtain the tracking error.

[0057] S3. Set a time-varying performance function for the tracking error. ρ ( t ), used to specify the upper and lower limits of the allowable tracking error.

[0058] As a preferred method, the original tracking error e Divide by performance function ρ ( t ), to obtain the standardized error variable The standardized variables are used to transform the constrained problem into a stabilization problem, i.e. Converging to a set value (e.g., 0), the original error satisfies This achieves the intended steady-state and transient performance.

[0059] S4. The host computer runs a fixed-time inversion adaptive control algorithm and performs hierarchical calculations. The outer loop generates a virtual angular velocity command that ensures the angle error converges within a fixed time. The inner loop, based on the robotic arm dynamics model and online adaptive disturbance estimation, solves for the actual control torque required to drive the motors of each joint module.

[0060] S5. The host computer outputs these torque commands to the motor controller and ensures that all tracking errors are constrained by the performance function boundary to achieve trajectory tracking.

[0061] S5. When the end effector of the robotic arm has moved to the pre-grabbing position, and the error between the actual pose and the expected grasping pose is less than the preset threshold, the end effector is triggered to complete the grasping action.

[0062] As a preferred method, see Figure 5The end effector 2 includes a cylinder 39, the cylinder body of which is fixed to the sixth output flange 37 on the sixth joint module; the piston 40 of the cylinder 39 is connected to the output end of the turntable to drive its rotation. The bionic palm part of the turntable has three connectors 41 evenly arranged around its upper circumference. The upper two ends of the connectors 41 are respectively hinged to the cylinder body of the cylinder 39 and the turntable. The lower side of the connectors 41 is hinged to one end of the bionic fingertip 44. The other end of the bionic fingertip 44 is provided with a bionic fingertip 43. A roller 42 is provided on its inner side, thus forming three sets of bionic fingers evenly distributed at 120° along the circumference. A force sensor for collecting pressure data is provided on the inner side of the bionic fingertip. The host computer is connected to the force sensor to control the opening and closing of the end effector.

[0063] It is understood that the end effector comprises a two-stage mechanism: a primary fixed mechanism consisting of cylinders 39, and a secondary movable mechanism consisting of the piston 40 of cylinders 39 and three bionic fingers. The two stages are connected by three connectors 41. Cylinders 39 serve as the power source, driving the piston 40 to move. The connectors 41 rotate with the movement of the piston 40, controlling the three bionic fingers to synchronously open or close (envelop the target fruit inwards). When closed, the target fruit can be covered collaboratively by a turntable and the bionic fingers.

[0064] Preferably, the inner surface of the roller is a smooth arc shape, which is beneficial for separating non-target fruits during the clamping process.

[0065] Preferably, the inner surface of the bionic fingertip and the inner surface of the turntable are both covered with flexible cushioning pads to prevent damage to the tomatoes.

[0066] Preferably, the cylinder body is connected to the sixth output flange of the sixth joint module by screws.

[0067] Preferably, the cylinder is a stroke-adjustable cylinder, which can achieve precise control of the cylinder stroke to achieve a precise closed or open position.

[0068] As a preferred embodiment, the working process and principle of the end effector in this embodiment include: The end effector, driven by the robotic arm, moves to the pre-harvesting position, at which point it is not in contact with the surface of the target tomato fruit. It then moves continuously from the pre-gripping position to the gripping position. The reciprocating motion of the piston in the cylinder controls the opening of the bionic finger. The bionic fingertip first contacts the surface of the target tomato fruit, and the roller rolls along the surface of the fruit during the wrapping process, reducing the friction between the bionic finger and the tomato and preventing damage to the tomato skin. When the target fruit is completely wrapped, the end effector stops moving, and the cylinder drives the end effector to close. During this process, the gripping force gradually increases, and when the gripping force reaches a threshold, the cylinder stops working. The rotation of the sixth joint module of the robotic arm causes the end effector to rotate along the central axis to separate the fruit from the stem. After the tomato is picked, the end effector moves to the appropriate position and places the tomato. During the picking process, the flexible cushioning pads on the bionic fingertips and the inside of the turntable effectively prevent damage to the tomato fruit.

[0069] As a preferred embodiment, the control method for opening and closing the end effector includes: The host computer uses a model predictive control (MPC) algorithm to compare the received real-time pressure data with a preset pressure damage threshold, and generates a prediction model based on the state space model of the dynamic system. Using the real-time pressure data as the initial state, it predicts the pressure evolution trajectory under different control strategies in the future finite time domain, and then calculates the optimal adjustment command for the finger opening and closing angle through online optimization. The command to adjust the finger opening and closing angle is sent to the cylinder's drive actuator, which adjusts the angle of the bionic fingertip so that while the roller continuously adheres to the surface of the target fruit, the contact pressure between the two is controlled below a safe threshold. If the coverage area exceeds the preset single fruit size or abnormal contact pressure is detected, the control system automatically interrupts the closing action.

[0070] The modular harvesting robotic arm of the present invention can achieve various configurations ranging from low to high degrees of freedom by combining different numbers and types of joint modules to adapt to different working environments and harvesting needs. Another embodiment is described below for further explanation. Example 2:

[0071] See Figure 6 This embodiment provides a fruit picking device based on modular design. The difference from embodiment 1 is that the third bionic joint is not set (the wrist joint is canceled). The second connecting arm 9 is directly connected to the fourth joint module. Its function is to connect the end joint module to the end effector 2, thereby configuring the robotic arm's degrees of freedom to four degrees of freedom. It can complete forward and backward swinging, up and down pitching and end grasping actions, but it lacks the flexible adjustment of the wrist joint.

[0072] The four-degree-of-freedom device in this embodiment is suitable for planting in greenhouses where row spacing is wide and plant height is relatively uniform. It is more suitable for situations with a simple task and relatively concentrated target fruits, and can significantly reduce hardware costs and computational burden.

[0073] It will be understood by those skilled in the art that the above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments or make equivalent substitutions for some of the technical features. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A fruit picking apparatus based on modular design, characterized by, The mechanical arm comprises a connecting arm and a plurality of bionic joints, the bionic joint comprises two joint modules rotating relative to each other, and the latter joint module is driven to rotate by a rotary driving mechanism arranged on the former joint module; The joint module comprises an L-shaped connecting plate composed of two fixed plates arranged orthogically; The rotary driving mechanism comprises a motor, a speed reducer and an output flange connected in transmission; The two fixed plates of the L-shaped connecting plate are connected with the output flange and rotate around the output shaft of the motor, and the other fixed plate is connected with the end effector, the connecting arm or the motor of the adjacent joint module; or one is connected with the connecting arm, and the other is connected with the motor of the adjacent joint module; The connecting arm comprises a first connecting arm and a second connecting arm, the two ends of the first connecting arm are connected with a first bionic joint and a second bionic joint respectively, and the first and second bionic joints simulate a shoulder joint and an elbow joint respectively; The two joint modules of the first bionic joint are connected with the robot body and the first connecting arm respectively, and the two joint modules of the second bionic joint are connected with the first connecting arm and the second connecting arm respectively; The other end of the second connecting arm is connected with a third bionic joint for simulating a wrist joint, and the third bionic joint is connected with the end effector through an end joint module to form a six-degree-of-freedom mechanism, or the second connecting arm is directly connected with the end effector through the end joint module to form a four-degree-of-freedom mechanism; A vision module is arranged on the mechanical arm, which is used to collect target fruit image information and transmit the information to an upper computer, so as to provide a basis for the upper computer to control the position of the mechanical arm and the movement of the end effector.

2. The fruit picking apparatus based on modular design according to claim 1, characterized in that, The upper computer outputs a control instruction, which comprises: Based on the target fruit image, the position of the fruit is obtained, and the expected value of the angle and the rotational angular velocity of each joint module is calculated based on the position of the fruit; The actual value of the angle and the rotational angular velocity of each joint module is read and compared with the corresponding expected value to obtain a tracking error; A time-varying performance function is set for the tracking error, which is used to define the upper and lower limits of the tracking error allowed; The upper computer runs a fixed-time inversion adaptive control algorithm to perform hierarchical calculation, an outer ring generates a virtual angular velocity instruction to ensure that the angle error converges within a fixed time, and an inner ring calculates the actual control torque required by the motor driving each joint module based on the dynamics model of the mechanical arm and online adaptive disturbance estimation; The upper computer outputs the actual control torque required by the motor to the corresponding motor controller, and ensures that all tracking errors are constrained by the performance function, so that the trajectory tracking is realized.

3. The fruit picking apparatus based on modular design as claimed in claim 2, wherein, The expected value of the angle and the rotational angular velocity of each joint module calculated based on the position of the fruit comprises: A target detection algorithm based on a deep neural network is used to detect the boundary box and the depth image of the target fruit, calculate the three-dimensional space position of the center point of the fruit image, and then calculate the position of the center of the fruit ball in combination with the camera coordinates and the known average radius of the fruit. Identifying the connection point between the stem and the tomato P p The vector formed by the fruit ball center position and P p The vector formed by the fruit ball center position and According to the fruit ball center position and the target fruit posture direction, a robot end grasping posture is determined, and an inverse kinematics model of the mechanical arm is used to solve angles that each joint module of the mechanical arm needs to reach, i.e. joint angle expected values; In combination with the planned trajectory from the initial position to the grasping position, the joint angle expected values are differentiated to obtain expected values of each joint angular velocity.

4. The fruit picking apparatus based on modular design as claimed in claim 2, wherein, The host computer output control instruction further comprises: When the end effector has moved to the pre-grasping position, and the errors of the actual posture and the expected grasping posture are both less than a preset threshold, the host computer controls the end effector to trigger to complete the grasping action.

5. The fruit picking apparatus based on modular design as claimed in claim 2, wherein, The performance function specifies the upper and lower bounds of the tracking error, comprising: The original tracking error is divided by the performance function to obtain a standardized error variable; The standardized variable is used to convert the constraint problem into a stabilized problem, i.e. the standardized error variable converges to a set value, achieving the predetermined steady-state and transient performance.

6. The modular design based fruit picking apparatus as claimed in claim 1, wherein, The structure of the end effector comprises a cylinder, the cylinder body of which is connected with the output flange of the joint module; the piston output end of the cylinder is connected with a rotating disc to drive the rotation thereof; The rotating disc simulates the palm center, three connectors are uniformly arranged on the rotating disc in the circumferential direction, the upper two ends of the connectors are respectively hinged with the cylinder body and the rotating disc, the lower side of the connector is hinged with one end of the bionic finger pad, and the other end of the bionic finger pad is provided with a bionic finger tip, and the inner side of the bionic finger tip is provided with a roller; The inner side of the bionic finger pad is provided with a force sensor for collecting pressure data, and the host computer is connected with the force sensor to control the opening and closing of the end effector.

7. The fruit picking apparatus based on modular design as claimed in claim 6, wherein, The opening and closing control mode of the end effector comprises: The host computer adopts a model predictive control algorithm, compares the received real-time pressure data with a preset pressure damage threshold, generates a prediction model based on a state space model of a dynamic system, takes the real-time pressure data as an initial state, predicts the pressure evolution trajectory under different control strategies in a future limited time domain, and then calculates an optimal adjustment instruction of the finger opening and closing angle through online optimization; The finger opening and closing angle adjustment instruction is sent to the driving actuator of the cylinder, which adjusts the angle of the bionic finger pad, so that the roller continuously adheres to the surface of the target fruit while ensuring that the contact pressure between the two is controlled below a safety threshold; If the coating range exceeds the preset single fruit size or an abnormal contact pressure is detected, the host computer controls the closing action to be interrupted.

8. The fruit picking apparatus based on modular design as claimed in claim 1 wherein, The two joint modules of the third bionic joint are connected with the second connecting arm and the end joint module respectively and rotate relative to each other along two orthogonal rotation axes.

9. The modular design based fruit picking apparatus as claimed in claim 1, wherein, The first bionic joint is connected with the robot body through the base; The robot body and the base are connected through a rotating module to form a seven-degree-of-freedom mechanism.

10. The modular design based fruit picking apparatus as claimed in claim 1, wherein, The structure of the joint module further comprises an outer shell connected with the L-shaped connecting plate and covering the outer side of the L-shaped connecting plate.

Images