Non-destructive harvesting of water-grown narcissus bulbs using autonomous robotic apparatus and methods

Abstract

The application discloses a kind of water-cultivated narcissus bulb non-destructive harvesting self-determination robot operation device and method, belong to agricultural robot technical field.Device includes mobile chassis, six-axis robot arm, flexible dexterous hand, 3D binocular vision module, buffer collection platform and edge computing controller;Flexible dexterous hand clamps the upper gripping convex rib of planting basket, and 3D binocular vision module gathers depth image.Controller built-in segmented flexible overturning, height linkage following and residual material detection vibration cleaning program: overturning program drives planting basket segmented variable speed progressive overturning;Linkage following program is synchronous in overturning Adjustment robot arm height, make basket mouth and platform fall difference keep preset small drop interval;Vibration cleaning program is identified after overturning stationary, drive planting basket to do small angle reciprocating vibration.The application realizes narcissus bulb grabbing, draining, discharging, cleaning, empty basket storage whole process non-destructive automation operation, solve bulb breakage, clamping slip, nutrient solution pollution and residual material blockage and other problems.

Description

Technical Field

[0001] This invention belongs to the field of agricultural robot harvesting equipment technology, specifically relating to an autonomous robot operation device and method for non-destructive harvesting of hydroponic daffodil bulbs, applicable to non-destructive automated harvesting of bulb crops in hydroponic pipeline planting baskets. Background Technology

[0002] Hydroponic daffodils, as an important ornamental flower and economic crop, rely heavily on manual labor or simple auxiliary equipment for bulb harvesting. While automated harvesting technology for hydroponic leafy vegetables and bulbous crops has seen some development, existing publicly available solutions primarily focus on improving the hydroponic cultivation structure, the planting basket itself, or the cleaning device. However, a fully automated robotic arm system for positioning and gripping the upper part of the planting basket and achieving intelligent, non-destructive harvesting throughout the entire process is still lacking. Analysis reveals several technical deficiencies in existing automated harvesting equipment:

[0003] First, the gripping structure has poor adaptability, making samples easily damaged and prone to slipping. Existing equipment mostly uses direct clamping of the bulbs or clamping of the entire planting basket. However, the surface of hydroponic daffodil bulbs is slippery and covered with epiphytic algae, making direct clamping extremely prone to slipping and detachment; at the same time, hydroponic roots are delicate and dense, easily causing root compression and breakage during clamping. If the entire planting basket is clamped, the grippers need to extend into the hydroponic tube to operate, which is a confined space and easily interferes with the tube wall, affecting operational stability.

[0004] Secondly, the tilting discharge method is prone to causing bulb damage and equipment failure. Most existing equipment uses a single, large-angle tilting discharge method. The bulbs fall from a height and generate impact, which can easily scratch the skin. As the tilt angle increases, the clamping friction is insufficient to balance the component of the basket's weight, which can easily lead to malfunctions such as the planting basket slipping and losing material. In addition, the discharge drop is fixed and cannot be adjusted, making it impossible to reduce impact damage according to actual conditions.

[0005] Third, the lack of residual material detection and force feedback mechanisms exacerbates damage under jammed conditions. When roots become entangled or bulbs become stuck in the basket wall, the existing system cannot effectively identify the jammed state and can only force discharge by increasing the turning angle and turning speed, causing large-scale tearing of roots and crushing of bulbs, which seriously affects the harvest quality.

[0006] Fourth, the lack of a draining process makes the nutrient solution prone to contaminating precision sensor components. If the nutrient solution attached to the roots is not drained before harvesting, the liquid will splash and adhere to the lenses of sensors such as binocular cameras and lidar during the turning process, causing image damage, reduced positioning and detection accuracy, and increasing the frequency of equipment maintenance and operating costs.

[0007] In summary, there is an urgent need for an integrated automated harvesting solution that takes into account stable grasping, sample protection, residue cleaning, and equipment protection. Summary of the Invention

[0008] To address the problems of existing hydroponic daffodil harvesting robots, such as easy damage to bulbs, slippage of planting baskets, difficulty in cleaning stuck residue, and easy contamination of sensors by nutrient solution, this invention provides an autonomous robot operating device and method for non-destructive harvesting of hydroponic daffodil bulbs, which can achieve integrated non-destructive automated operation of grasping, pulling, draining, flexible flipping, low drop discharge, visual cleaning, and empty basket storage.

[0009] To achieve the above objectives, the present invention adopts the following technical solution: an autonomous robot operation device for non-destructive harvesting of hydroponic daffodil bulbs, comprising a mobile chassis, a six-axis robotic arm, a flexible dexterous hand, a 3D binocular vision module, a buffer collection platform, and an edge computing controller.

[0010] The six-axis robotic arm is mounted on a mobile chassis. A flexible, dexterous hand is attached to the end of the six-axis robotic arm and is used to grip the upper part of the planting basket and grasp the protruding edges. The 3D binocular vision module is used to acquire depth images of the planting basket and the bulbs inside. The buffer collection platform is used to receive the dislodged daffodil bulbs. The edge computing controller is electrically connected to the six-axis robotic arm, the flexible, dexterous hand, and the 3D binocular vision module.

[0011] The edge computing controller has a built-in segmented flexible flipping control program, a highly linked follow control program, and a residual material detection and vibration cleaning program.

[0012] The segmented flexible flipping control program is configured to drive a six-axis robotic arm to gradually flip the planting basket in segments with varying speeds, limiting the maximum flipping angle to ≤22°. By limiting the maximum flipping angle to within 22°, slippage is avoided due to insufficient clamping force to balance the weight component of the basket caused by an excessively large angle.

[0013] The height-linked tracking control program is configured to synchronously drive the six-axis robotic arm to move vertically during the flipping operation, maintaining the vertical height difference between the planting basket opening and the buffer collection platform within a preset small drop range. Through height-linked tracking during the flipping process, a low-drop, flexible transition of the bulb from the basket opening to the collection platform is achieved, avoiding impact damage caused by falling from a height.

[0014] The residual material detection and vibration cleaning program is configured as follows: after the plant has been turned over and allowed to stand, the system determines whether there are any remaining bulbs or roots in the basket based on images captured by the 3D binocular vision module. If any residue is found, the planting basket is driven to vibrate at a small angle around the maximum tilt angle. By combining visual recognition with small-amplitude vibration, residual material can be effectively removed without increasing the tilt angle, thus effectively clearing stuck roots and residual bulbs.

[0015] Furthermore, the flexible and dexterous hand integrates a pressure sensing module; the edge computing controller is configured to use a first clamping force range during the vertical extraction stage of the planting basket, and switch to a second clamping force range greater than the first clamping force range during the flipping operation stage; the flexible and dexterous fingertips are equipped with elastic anti-slip buffer pads. Through condition-specific adaptive clamping force adjustment, a lower clamping force is used during the extraction stage to avoid damaging the protruding edges of the planting basket or squeezing the bulbs inside the basket, while a higher clamping force is used during the flipping stage to balance the weight component of the basket and prevent slippage.

[0016] Furthermore, the preset small drop range is 20mm to 30mm; the segmented flexible flipping control program executes three actions sequentially: the pre-tilting section flips from 0° to 10° using a first angular velocity; the discharge sliding section flips from 10° to the maximum flipping angle using a second angular velocity greater than the first angular velocity; the stationary discharge section maintains the maximum flipping angle for 2s to 5s. The pre-tilting section flips slowly at a lower angular velocity, gradually separating the root system from the basket wall; the discharge sliding section completes the main body discharge at a higher angular velocity; the stationary discharge section relies on gravity to allow the bulb to slide out fully.

[0017] Furthermore, the edge computing controller is also configured to drive a six-axis robotic arm to vertically lift the planting basket at low speed. After the planting basket detaches from the hydroponic planting unit, it hovers to drain water before being lifted to the turning height. By adding a hovering and draining process, the risk of nutrient solution splashing and contaminating the sensors is reduced, thus lowering the frequency of equipment maintenance.

[0018] Furthermore, the mobile chassis is equipped with an empty basket storage rack; the edge computing controller is configured to drive a six-axis robotic arm to move the empty planting basket to the empty basket storage rack and release it after harvesting. By integrating the empty basket storage station, the empty basket return pipeline process is eliminated, making it suitable for multiple batches of one-time mass harvesting.

[0019] Furthermore, the buffer collection platform is composed of an upper flexible buffer layer, a lower liquid collection box, and a surrounding flexible enclosure. The upper flexible buffer layer absorbs the impact of the falling bulbs, the lower liquid collection box collects residual nutrient solution, and the surrounding flexible enclosure prevents the bulbs from rolling off.

[0020] Furthermore, the mobile chassis is equipped with a LiDAR and SLAM positioning module, and the mobile chassis completes secondary docking and locking based on the visual calibration marks at the workstation. The combination of SLAM autonomous navigation and visual calibration secondary positioning ensures the accuracy of the robotic arm's working reference position.

[0021] This invention also provides a method for harvesting hydroponic daffodil bulbs using an autonomous robot without damage, employing the aforementioned device and including the following steps:

[0022] Grasping step: The flexible and dexterous hand grips the upper part of the planting basket and grasps the protruding edge;

[0023] Removal Step: The six-axis robotic arm lifts the planting basket vertically at low speed, causing the planting basket to detach from the hydroponic planting unit;

[0024] Draining process: After the planting basket is detached from the hydroponic planting unit, it is suspended to drain water, and then raised to the height for turning over.

[0025] The flipping and unloading step: The segmented flexible flipping control program drives the six-axis robotic arm to gradually flip the planting basket in segments with variable speed, limiting the maximum flipping angle to ≤22°; at the same time, the height linkage follow control program synchronously drives the six-axis robotic arm to move vertically, so that the vertical height difference between the opening of the planting basket and the buffer collection platform is maintained within the preset small drop range.

[0026] Residual material detection and cleaning steps: After the plant is turned over and left to stand, the 3D binocular vision module acquires images of the inside of the planting basket. The residual material detection and cleaning program determines whether there are any remaining bulbs or roots in the basket based on the images. If there are any residues, the planting basket is driven to vibrate back and forth at a small angle with the maximum tilt angle as the center.

[0027] Empty basket storage steps: Move the empty planting basket to the empty basket storage rack and release it.

[0028] Furthermore, in the grasping step, the edge computing controller switches the clamping force according to the operation stage: the first clamping force range is used in the vertical pulling stage of the planting basket, and the second clamping force range, which is greater than the first clamping force range, is switched to in the flipping operation stage.

[0029] Furthermore, in the flipping and discharging step, the segmented flexible flipping control program sequentially executes three actions: the pre-tilting section flips from 0° to 10° using a first angular velocity; the discharging sliding section flips from 10° to the maximum flipping angle using a second angular velocity greater than the first angular velocity; and the stationary discharging section maintains the maximum flipping angle for 2s to 5s.

[0030] Compared with the prior art, the present invention has the following beneficial effects:

[0031] 1. Three-stage variable speed low-angle flipping combined with height-linked low-drop tracking provides dual cushioning against bulb impact through both motion trajectory and drop height. Compared to traditional discharge methods with a single large angle greater than 45° and a drop of over 100mm, this significantly reduces the bulb skin damage rate and ensures sample integrity.

[0032] 2. The adaptive clamping force adjustment for different working conditions, combined with the special gripping protrusions on the upper part of the planting basket, uses a lower clamping force during the vertical extraction stage to avoid damaging the basket or squeezing the bulbs, and switches to a higher clamping force during the flipping stage to balance the weight component of the basket, effectively improving the clamping stability in the tilted state and preventing failures such as slippage and material loss when the planting basket is flipped.

[0033] 3. After pulling out the roots, a suspension and drainage process is added to ensure that the nutrient solution attached to the roots is fully drained before the turning operation, thereby reducing the splashing of nutrient solution and contaminating precision sensors such as lidar and binocular cameras, and reducing the frequency of equipment maintenance and operation and maintenance costs.

[0034] 4. 3D vision-based residual material recognition is linked to small-amplitude, low-frequency vibration for material removal. This effectively cleans stuck roots and residual bulbs without increasing the tilt angle, avoiding root tearing and bulb compression damage caused by forced discharge.

[0035] 5. Full-process multi-sensor closed-loop control, motion parameters, clamping force parameters, and timing parameters can all be quantified and stored, adapting to smart agriculture experimental data collection and scientific research and teaching scenarios in universities.

[0036] 6. The chassis integrates an empty basket temporary storage station, eliminating the empty basket return pipeline process, which is suitable for multiple batches of one-time batch harvesting, and the operation process is simple and efficient. Attached Figure Description

[0037] Figure 1 This is a schematic diagram of the overall structure of the autonomous robot operation device for non-destructive harvesting of hydroponic daffodil bulbs provided in an embodiment of the present invention;

[0038] Figure 2 This is a schematic diagram of a partial structure of the grasping protrusion on the upper part of the planting basket in an embodiment of the present invention;

[0039] Figure 3 This is a schematic diagram of the three flexible flipping angle-angular velocity variation curves in an embodiment of the present invention;

[0040] Figure 4 This is a comparative diagram of highly coordinated following operations in an embodiment of the present invention;

[0041] Figure 5 This is a schematic diagram of the ±5° reciprocating vibration action for cleaning residual material in an embodiment of the present invention;

[0042] Figure 6 This is a schematic cross-sectional view of the flexible buffer collection platform in an embodiment of the present invention;

[0043] Figure 7 This is a flowchart illustrating the implementation of the autonomous robot operation method for non-destructive harvesting of hydroponic daffodil bulbs provided in an embodiment of the present invention. Detailed Implementation

[0044] The present invention will be further described below with reference to the accompanying drawings and embodiments.

[0045] It should be noted that the following detailed descriptions are exemplary and intended to provide further explanation of this application. Unless otherwise specified, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application pertains.

[0046] It should be noted that the terminology used herein is for the purpose of describing particular embodiments only and is not intended to limit the exemplary embodiments according to this application. As used herein, the singular form is intended to include the plural form as well, unless the context clearly indicates otherwise. Furthermore, it should be understood that when the terms "comprising" and / or "including" are used in this specification, they indicate the presence of features, steps, operations, devices, components, and / or combinations thereof.

[0047] I. System Overall Structure and Hardware Deployment

[0048] like Figure 1 As shown, the autonomous robot operation device for non-destructive harvesting of hydroponic daffodil bulbs provided in this embodiment has the following hardware components: a four-wheel differential speed mobile chassis (using a 5G smart garden vehicle chassis), a six-axis collaborative robotic arm, a multi-degree-of-freedom flexible dexterous hand, a 16-line LiDAR, a global shutter 3D binocular camera (i.e., a 3D binocular vision module), an edge computing controller (edge ​​computing power ≥21 TOPS); matching 160mm~200mm PVC hydroponic planting pipes, PP material planting baskets with upper gripping protrusions, a double-layer sponge buffer collection platform, and a vehicle-mounted empty basket temporary storage rack.

[0049] like Figure 1 As shown, a six-axis collaborative robotic arm is fixed to one side of the mobile chassis. A flexible, dexterous hand is mounted at the end of the robotic arm to grip the upper part of the planting basket and grasp the protruding edges. A 3D binocular camera is mounted at the end of the robotic arm or above the side of the robotic arm base, with its field of view covering the working area of ​​the planting basket, used to acquire depth images of the planting basket and its internal bulbs. An edge computing control box is fixed to the mobile chassis and is electrically connected to the six-axis collaborative robotic arm, the flexible, dexterous hand, the 3D binocular camera, and the 16-line LiDAR. A vehicle-mounted empty basket storage rack is installed at the rear of the mobile chassis for bulk storage of empty planting baskets after harvesting.

[0050] like Figure 1 As shown, the buffer collection platform is mounted on the mobile chassis and located within the operating radius of the robotic arm, used to collect the dumped daffodil bulbs. A 16-line LiDAR is installed on the top front of the mobile chassis, working in conjunction with the SLAM positioning module to achieve autonomous navigation and workstation identification.

[0051] In this invention, the planting basket adopts a one-piece molded structure made of PP material with annular anti-slip gripping protrusions. For example... Figure 2As shown, the upper outer wall of the planting basket has an annular anti-slip gripping ridge, which protrudes radially outward for gripping by a flexible, dexterous hand. The bottom of the planting basket is a permeable root support area, and the outer wall of the basket has reflective positioning marks to assist the 3D binocular camera in identification and positioning. The fingertips of the flexible, dexterous hand have elastic anti-slip cushioning pads, which increase friction during gripping to prevent slippage and avoid rigid contact that could damage the planting basket's ridge.

[0052] II. Autonomous Navigation and Workstation Correction

[0053] like Figure 7 As shown, before the complete workflow of this embodiment begins, the autonomous navigation and workstation calibration steps are performed first.

[0054] The mobile chassis autonomously travels to the target hydroponic workstation using laser SLAM (Simultaneous Localization and Mapping). Upon reaching the vicinity of the workstation, a 3D binocular camera identifies a pre-set ArUco visual calibration code on the end of the hydroponic planting tube, and the mobile chassis completes a secondary docking and locking based on this visual calibration mark. By combining coarse SLAM positioning with fine visual calibration positioning, the chassis docking position accuracy is ensured to meet the requirements of the robotic arm operation.

[0055] After the chassis is locked, as Figure 7 As shown, a 3D binocular camera acquires images of the planting basket at the current workstation. By identifying reflective markers on the basket or segmenting the basket's outline, the horizontal offset and angular deflection of the basket relative to the robotic arm's base coordinate system are calculated. The six-axis robotic arm fine-tunes its end effector pose based on visual feedback, aligning the flexible, dexterous hand directly above the upper grasping protrusion of the planting basket and correcting the basket's posture to a vertical position, preparing it for subsequent grasping operations.

[0056] III. Adaptive Pressure Capture

[0057] like Figure 7 As shown, after the workstation calibration is completed, the grabbing step is executed.

[0058] The robotic arm lowers the flexible, dexterous hand to the upper part of the planting basket to grasp the raised edge area, ensuring that the fingertips of the dexterous hand are positioned precisely on the upper and lower sides of the ring-shaped grasping edge. The flexible, dexterous hand closes its grip, and the elastic, non-slip cushioning pads at the fingertips contact the surface of the grasping edge. A 3D binocular camera visually verifies that the fingertips completely cover the grasping area and do not touch the daffodil bulbs inside the basket.

[0059] In this embodiment, the flexible, dexterous hand integrates a pressure sensing module for real-time feedback of the clamping force value. The edge computing controller adaptively switches the clamping force range according to the operation stage: during the vertical extraction stage of the planting basket (Stage=PULL), a first clamping force range of 0.6–0.8 kgf is used; during the flipping stage (Stage=FLIP), it switches to a second clamping force range of 0.8–1.0 kgf. Through adaptive clamping force adjustment based on different working conditions, a lower clamping force is used during the extraction stage to avoid damaging the protruding edges of the planting basket or squeezing the bulbs inside the basket, while a higher clamping force is used during the flipping stage to balance the weight component of the basket and prevent slippage.

[0060] The specific control logic is as follows:

[0061] Input: Real-time feedback value F~real~ from the pressure sensing module, Stage ∈ {PULL, FLIP} (operation stage flag).

[0062] Output: Target gripping force F~target~ of the flexible, dexterous hand

[0063] Initialization: Set the first clamping force range [F_min1, F_max1] = [0.6 kgf, 0.8 kgf]

[0064] Set the second clamping force range [F_min2, F_max2] = [0.8 kgf, 1.0 kgf].

[0065] Set the switching threshold T_change=0.5s (the delay time after the unplugging is completed).

[0066] Main loop:

[0067] Read Stage value

[0068] IF Stage==PULL:

[0069] F_target = clamp(F_real, F_min1, F_max1) / / Limiting amplitude within the range

[0070] Invoke the PID controller to make F_real approach F_target.

[0071] ELSE IF Stage==FLIP:

[0072] Wait for the T_change time to ensure a smooth transition.

[0073] F_target=clamp(F_real, F_min2, F_max2)

[0074] Invoke the PID controller to make F_real approach F_target.

[0075] END IF

[0076] Real-time monitoring is performed to check if F_real exceeds 1.2 times [F_min, F_max]. If it does, an emergency stop protection is triggered and an anomaly is reported.

[0077] IV. Low-speed extraction and hovering for drainage

[0078] like Figure 7 As shown, after clamping is completed, the removal and draining steps are performed.

[0079] The robotic arm maintains a vertical posture and lifts the planting basket vertically at a constant speed of 0.12 m / s, slowly pulling the basket out of the hydroponic planting tube. During the lifting process, the edge computing controller continuously monitors the load torque M_load at the end of the robotic arm; if M_load is detected to be greater than the preset jamming threshold M_threshold (for example, due to abnormally increased resistance caused by the roots entangled with the tube wall), the lifting is immediately stopped and the basket is lowered by 10 mm before attempting to lift again; if the lifting fails three times in a row, the system alarms and waits for manual intervention to avoid forcibly pulling out the basket and causing root tearing or damage to the basket.

[0080] After the planting basket is completely detached from the PVC hydroponic planting tube (complete detachment is defined as Z_current > Z_tube_top + δ, where δ is a safety margin, taken as 50mm in this embodiment), the robotic arm hovers for T_drain = 2s, allowing the nutrient solution attached to the roots to drain naturally under gravity. By adding a hovering and draining process, the risk of nutrient solution splashing and contaminating the lenses of precision sensors such as the 16-line LiDAR and 3D binocular camera during subsequent turning operations is effectively reduced.

[0081] After draining, the robotic arm continues to lift the planting basket to the preset flipping height Z_flip, and enters the flipping and discharging preparation stage.

[0082] V. Segmented flexible tilting and highly coordinated material discharge

[0083] like Figure 7 As shown, after the planting basket reaches the turning height, the turning and discharging step is executed. This step also involves a segmented flexible turning control program (corresponding to...). Figure 3 The three flip characteristic curves shown) and the highly linked follow control program (corresponding to) Figure 4 (The linkage effect shown).

[0084] The edge computing controller executes segmented angular velocity closed-loop flip logic. For example... Figure 3 As shown, the segmented flexible flipping control program executes three actions sequentially:

[0085] Pre-tilting section: The planting basket is rotated from 0° to 10° at an angular velocity ω1=0.03rad / s. During this stage, the basket is rotated slowly at a low angular velocity to allow the daffodil roots to gradually and gently separate from the part of the basket that is in contact with the wall, thus avoiding the roots being forcibly broken due to a sudden large-angle rotation.

[0086] Discharge sliding section: The planting basket rotates from 10° to the maximum rotation angle θ~max~ (22° in this embodiment), with an angular velocity ω2 = 0.06 rad / s, where ω2 is greater than ω1. During this stage, the main body rotation is completed at a relatively high but still gentle angular velocity, and the bulbs slide smoothly along the bottom of the basket and the surface of the flexible buffer layer under the influence of gravity. It should be noted that the maximum rotation angle θ~max~ is limited to 22°. Figure 3 As shown, this angle is much smaller than the 45° flip angle in the traditional solution. The core consideration is that when the flip angle increases, the component of the planting basket's weight along the clamping surface increases accordingly. If the angle exceeds 22°, even with a higher clamping force (1.0 kgf), there is still a risk that the planting basket will slip out of the dexterous hand. On the other hand, a low flip angle of less than 22° prevents the bulb from being violently thrown when it slides out of the basket. Combined with the low drop of the buffer collection platform below, the risk of damage to the bulb's skin can be greatly reduced.

[0087] Static discharge section: Lock the maximum tilt angle to 22° and let it stand for 3 seconds (which can be adjusted within the range of 2 seconds to 5 seconds). Rely on gravity to allow most of the bulbs to slide out of the planting basket and fall into the buffer collection platform.

[0088] Throughout the flipping process, such as Figure 4 As shown, the highly synchronized follow-up control program synchronously drives the robotic arm to move vertically. Specifically, as... Figure 4 As shown in the right figure, as the rotation angle of the planting basket gradually increases, the height of the basket opening relative to the buffer collection platform changes continuously. The controller reads the current basket opening height Z_basket and buffer platform height Z_platform in real time, calculates the actual drop H_real = Z_basket - Z_platform, and compares it with the desired drop H_desired (the median of the target drop range [20mm, 30mm], which is 25mm). If the actual drop deviates from the desired value, the controller calculates the vertical displacement compensation ΔZ through proportional control (proportional coefficient Kp_z = 0.5), and adjusts the height of the robotic arm end in real time to keep the vertical height difference between the basket opening and the buffer collection platform within the preset small drop range of 20mm to 30mm. Figure 4 As shown in the image on the right. For comparison, Figure 4The left figure illustrates that without height tracking control, the drop difference between the basket opening and the platform gradually increases (usually exceeding 100mm) during the flipping process, making the rim susceptible to impact damage from falling from a height. The linkage tracking control of this invention provides dual cushioning of the rim impact through both the movement trajectory and the drop height, effectively ensuring the integrity of the rim's outer layer.

[0089] The control logic is as follows:

[0090] Input: Maximum target tilt angle θmax = 22°, initial height difference H between the basket opening and the buffer platform.

[0091] Outputs: End-effector rotational angular velocity ω(t), vertical displacement compensation ΔZ(t)

[0092] Define constants:

[0093] The pre-tilt section angle range is [0°, θ1], where θ1 = 10°.

[0094] The angle range of the discharge sliding section is (θ1, θmax), where θmax = 22°.

[0095] The settling time for discharging material is T_still = 3s.

[0096] The first angular velocity ω1 = 0.03 rad / s

[0097] The second angular velocity ω2 = 0.06 rad / s

[0098] The target elevation difference range is [H_min, H_max] = [20mm, 30mm].

[0099] Main loop (executed once per control cycle Δt):

[0100] Read the current flip angle θ_current

[0101] Read the current basket height Z_basket

[0102] Read the buffer platform height Z_platform

[0103] / / Stage determination and angular velocity assignment

[0104] IF θ_current<θ1:

[0105] ω=ω1

[0106] ELSE IF θ_current<θmax:

[0107] ω=ω2

[0108] ELSE:

[0109] ω=0

[0110] Start timer T_count

[0111] IF T_count>=T_still:

[0112] Proceed to the next stage (residual material testing)

[0113] END IF

[0114] END IF

[0115] / / Update flip angle

[0116] θ_new = θ_current + ω × Δt

[0117] If θ_new>θmax, then θ_new=θmax

[0118] / / High-level linkage and follow calculation

[0119] Desired elevation difference H_desired = (H_min + H_max) / 2 = 25mm

[0120] Current actual elevation difference H_real = Z_basket - Z_platform

[0121] Deviation e_H = H_real - H_desired

[0122] / / Adjust the vertical displacement of the robotic arm using proportional control

[0123] Kp_z=0.5 / / Scale factor

[0124] ΔZ=Kp_z×e_H×Δt

[0125] If |e_H| < 2 mm, then it is considered stable, and ΔZ = 0.

[0126] / / Execute motion commands

[0127] Send command: The end effector of the robotic arm rotates Δθ = ω × Δt, and simultaneously translates vertically by ΔZ.

[0128] VI. Buffer Collection Platform Structure

[0129] like Figure 6As shown, the buffer collection platform is composed of an upper flexible buffer layer, a lower liquid collection box and peripheral flexible enclosures. The upper flexible buffer layer is made of flexible sponge material, laid on the surface of the platform, for absorbing the impact energy when the bulbs fall; the flexible buffer layer is preferably arranged in an inclined manner, so as to facilitate the bulbs to roll down and concentrate on one side of the collection area. The lower liquid collection box is used to collect the residual nutrient solution that slides down together with the bulbs, preventing the nutrient solution from accumulating on the platform surface or flowing to other areas of the chassis to cause pollution. The peripheral flexible enclosures are made of flexible rubber material, arranged around the platform, to prevent the bulbs from rolling off the edge of the platform and causing secondary damage.

[0130] As Figure 6 shown, when the bulbs slide out from the opening of the planting basket, they first fall onto the surface of the upper flexible buffer layer, the elastic deformation of the flexible sponge effectively absorbs the impact kinetic energy, and then the bulbs slowly roll down along the surface of the inclined buffer layer to the collection area. The residual nutrient solution flows through the pores of the flexible buffer layer or along the interlayer gaps into the lower liquid collection box for centralized collection. The flexible enclosures form a soft barrier around the platform, so that even if the bulbs roll to the edge of the platform, they will be blocked by the enclosures and will not fall to the ground.

[0131] VII. Visual residual material detection and small-amplitude vibration material cleaning

[0132] As Figure 7 shown, after the static discharging section is completed, the residual material detection and material cleaning steps are executed.

[0133] After standing for 3s, the 3D binocular camera collects the depth image I~depth~ of the inner area of the planting basket. The edge computing controller performs image segmentation processing on the depth image, extracts the foreground area in the basket (that is, the image area of residual bulbs and tangled roots), and calculates the total pixel area S_residual of the foreground area. Compare S_residual with the preset residue determination threshold S_threshold=500pixel²:

[0134] If S_residual<S_threshold, it is determined as "no residue", the vibration material cleaning is skipped, and the process directly enters the empty basket storage step;

[0135] If S_residual≥S_threshold, it is determined as "residue exists", and the process enters the vibration material cleaning subroutine.

[0136] As Figure 5As shown, in the vibration cleaning subroutine, the robotic arm drives the planting basket to perform low-speed reciprocating vibrations at a maximum tilt angle θ~vib_center~=22°, with a center of ±5° (i.e., the planting basket oscillates back and forth along a sinusoidal trajectory within the range of 17° to 27°), with an oscillation frequency f_vib=1Hz (one complete reciprocating cycle per second). Each vibration lasts for one complete cycle (2 seconds), followed by a 0.5-second wait to allow the residual material to stabilize before acquiring a depth image again to update the residual area S_residual. The maximum number of vibrations N_max=5 times. If S_residual is still ≥ S_threshold after reaching the maximum number of vibrations, an error log is recorded and uploaded to the cloud for manual handling.

[0137] It should be noted that, as Figure 5 As shown, the vibratory cleaning method uses a small oscillation of ±5° around a maximum tilt angle of 22°. Its advantage lies in the following: compared to forcibly removing residual material by continuously increasing the tilt angle (e.g., from 22° to over 30°), this method uses small-amplitude high-frequency vibration to disrupt the root entanglement and bulb jamming, allowing the residual material to fall off naturally under the combined action of gravity and inertia. This eliminates the risk of slippage by increasing the tilt angle, avoiding root tearing and bulb compression damage caused by forced discharge at large angles. Simultaneously, the ±5° vibration amplitude is strictly limited to ensure that the instantaneous maximum tilt angle (27°) of the planting basket remains within a safe range throughout the entire vibration process, preventing slippage due to excessive instantaneous tilt angle.

[0138] During the vibration cleaning process, the bulb and root residues fall into the buffer collection platform, where they are received by the upper flexible buffer layer, while the residual nutrient solution flows into the lower collection box.

[0139] 8. Empty basket storage and repositioning

[0140] like Figure 7 As shown, after the residual material detection passes (no residue in the basket), the empty basket storage step is performed.

[0141] The robotic arm returns the planting basket to a vertical position (θ=0°), while the flexible, dexterous hand continues to hold the empty planting basket. The edge computing controller plans a collision-free path from the current position P_empty of the empty basket to the target position P_rack of the on-board empty basket storage rack (using a pre-calibrated robotic arm motion space model for path planning to avoid surrounding obstacles and already stored empty baskets).

[0142] The robotic arm, holding an empty basket, moves along the planned path to above the temporary storage rack. After slowly descending to the placement height, the flexible, dexterous hand releases the basket, allowing it to rest stably on the rack. The robotic arm then returns to its standby position, completing a single harvesting cycle.

[0143] like Figure 1As shown, the vehicle-mounted empty basket storage rack is integrated into the rear of the mobile chassis and can accommodate multiple empty planting baskets. By integrating the empty basket storage station, the step of returning empty baskets to the pipeline process in the traditional solution is eliminated. The entire robot can continuously complete the harvesting operations at multiple stations, storing the empty baskets uniformly on the chassis. This is suitable for multi-batch, one-time batch harvesting scenarios, and the operation process is simple and efficient.

[0144] IX. Adjustable Parameters and Applicable Scope

[0145] In this embodiment, the following key parameters can be adaptively adjusted according to actual operating conditions such as the size of the daffodil bulbs (differences in diameter and weight between different varieties of daffodil bulbs) and the density of the root system (differences in root density and length due to different hydroponic time):

[0146] 1. Maximum tilt angle θmax: The preferred range is 15°~22°. When the bulbs are large or the root system is dense, the maximum tilt angle should be appropriately reduced to avoid jamming; when the bulbs are small or the root system is sparse, the tilt angle can be appropriately increased within a safe range to improve discharge efficiency.

[0147] 2. Tilting angular velocities ω1 and ω2: The preferred range is 0.02 to 0.08 rad / s. A lower angular velocity is used when the root system is dense to protect the root system; a higher angular velocity can be used when the root system is sparse to improve work efficiency.

[0148] 3. Discharge settling time T_still: The preferred range is 2s to 5s. When the bulbs have poor sliding properties, the settling time should be appropriately extended; when the bulbs have good sliding properties, the settling time can be shortened while ensuring the discharge effect.

[0149] 4. Dynamic clamping force (first clamping force range and second clamping force range): The preferred range is 0.7 to 1.0 kgf (it can be adjusted within a wider range according to the actual situation. In principle, the lower limit of the first clamping force range should not be lower than 0.6 kgf and the upper limit of the second clamping force range should not be higher than 1.0 kgf).

[0150] 5. The preferred range for the basket opening drop range [H_min, H_max] is 20mm to 40mm to ensure that the discharge drop of the bulb is low enough to protect the bulb skin, while avoiding collision and interference between the basket opening and the buffer collection platform.

[0151] All of the above parameters can be quantified and stored in the storage unit of the edge computing controller. The parameter combination for each job can be recorded as a log, facilitating subsequent data analysis and job optimization. The parameter adjustment ranges mentioned above are for illustrative purposes only. Any simple modifications, equivalent substitutions, or structural fine-tunings made based on the concept of this invention are within the scope of protection of this invention.

[0152] 10. Full-process collaborative control

[0153] like Figure 7 As shown, the above steps are not executed in isolation, but are coordinated and controlled by the edge computing controller according to a preset timing sequence. Figure 7 The complete workflow of this invention is illustrated: starting with autonomous navigation and station calibration, it sequentially proceeds through adaptive pressure gripping, low-speed extraction and hovering draining, segmented flexible flipping and highly coordinated material discharge, visual residual material detection and small-amplitude vibration cleaning, and finally ends with empty basket storage and reset. The transition conditions, timing parameters, and safety interlock mechanisms between each step are pre-configured in the edge computing controller to ensure the continuity and safety of the entire process.

[0154] The full-process multi-sensor closed-loop control of this invention enables the quantification and storage of motion parameters, clamping force parameters, and timing parameters, making it suitable for data acquisition and scientific research teaching scenarios in smart agriculture experiments at universities.

[0155] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention in any other way. Any person skilled in the art may make changes or modifications to the above-disclosed technical content to create equivalent embodiments. However, any simple modifications, equivalent changes, and modifications made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the protection scope of the present invention.

Claims

1. An autonomous robotic harvesting device for hydroponic daffodil bulbs without damage, characterized in that, include: Mobile chassis; A six-axis robotic arm is mounted on a mobile chassis; A flexible and dexterous hand is attached to the end of a six-axis robotic arm to grip and grasp the protruding edge on the upper part of the planting basket; A 3D binocular vision module is used to acquire depth images of the planting basket and the bulbs inside; A buffer collection platform is used to receive the poured-out daffodil bulbs; The edge computing controller is electrically connected to the six-axis robotic arm, the flexible dexterous hand, and the 3D binocular vision module, respectively. The edge computing controller incorporates a segmented flexible flipping control program, a height-linked follow-up control program, and a residual material detection and vibration cleaning program. The segmented flexible flipping control program is configured to drive a six-axis robotic arm to progressively flip the planting basket in segments with varying speeds. The height-linked follow-up control program is configured to synchronously drive the six-axis robotic arm to move vertically during the flipping operation, maintaining the vertical height difference between the opening of the planting basket and the buffer collection platform within a preset small drop range. The residual material detection and vibration cleaning program is configured to determine whether there are any remaining bulbs or roots in the basket based on the images collected by the 3D binocular vision module after the flipping and settling process is completed. If there are any remaining materials, the planting basket is driven to perform small-angle reciprocating vibrations centered on the maximum flipping angle.

2. The autonomous robot operation device for non-destructive harvesting of hydroponic daffodil bulbs according to claim 1, characterized in that, The flexible and dexterous hand integrates a pressure sensing module; the edge computing controller is configured to use a first clamping force range during the vertical extraction stage of the planting basket, and switch to a second clamping force range greater than the first clamping force range during the flipping operation stage; the flexible and dexterous fingertip is provided with an elastic anti-slip buffer pad.

3. The autonomous robot operation device for non-destructive harvesting of hydroponic daffodil bulbs according to claim 1, characterized in that, The maximum tilt angle is limited to ≤22°; the preset small drop range is 20mm~30mm; the segmented flexible tilt control program executes three actions in sequence: the pre-tilt section tilts from 0° to 10° using the first angular velocity; the discharge sliding section tilts from 10° to the maximum tilt angle using the second angular velocity greater than the first angular velocity; the stationary discharge section maintains the maximum tilt angle and remains stationary for 2s~5s.

4. The autonomous robot operation device for non-destructive harvesting of hydroponic daffodil bulbs according to claim 1, characterized in that, The edge computing controller is also configured to drive a six-axis robotic arm to vertically lift the planting basket at low speed. After the planting basket is detached from the hydroponic planting unit, it is suspended to drain water, and then lifted to the flipping operation height.

5. The autonomous robot operation device for non-destructive harvesting of hydroponic daffodil bulbs according to claim 1, characterized in that, The mobile chassis is equipped with an empty basket storage rack; the edge computing controller is configured to drive a six-axis robotic arm to move the empty planting basket to the empty basket storage rack and release it after harvesting.

6. The autonomous robot operation device for non-destructive harvesting of hydroponic daffodil bulbs according to claim 1, characterized in that, The buffer collection platform is composed of an upper flexible buffer layer, a lower liquid collection box, and a surrounding flexible enclosure.

7. The autonomous robot operation device for non-destructive harvesting of hydroponic daffodil bulbs according to claim 1, characterized in that, The mobile chassis is equipped with a lidar and SLAM positioning module, and the mobile chassis completes secondary docking and locking based on the visual calibration marks of the workstation.

8. A method for harvesting hydroponic daffodil bulbs without damage using an autonomous robot, comprising the autonomous robot device for harvesting hydroponic daffodil bulbs without damage as described in any one of claims 1-7, characterized in that, Includes the following steps: Grasping step: The flexible and dexterous hand grips the upper part of the planting basket and grasps the protruding edge; Removal Step: The six-axis robotic arm lifts the planting basket vertically at low speed, causing the planting basket to detach from the hydroponic planting unit; Draining process: After the planting basket is detached from the hydroponic planting unit, it is suspended to drain water, and then raised to the height for turning over. The flipping and unloading step: The segmented flexible flipping control program drives the six-axis robotic arm to gradually flip the planting basket in segments with variable speed, limiting the maximum flipping angle to ≤22°; at the same time, the height linkage follow control program synchronously drives the six-axis robotic arm to move vertically, so that the vertical height difference between the opening of the planting basket and the buffer collection platform is maintained within the preset small drop range. Residual material detection and cleaning steps: After the plant is turned over and left to stand, the 3D binocular vision module acquires images of the inside of the planting basket. The residual material detection and cleaning program determines whether there are any remaining bulbs or roots in the basket based on the images. If there are any residues, the planting basket is driven to vibrate back and forth at a small angle with the maximum tilt angle as the center. Empty basket storage steps: Move the empty planting basket to the empty basket storage rack and release it.

9. The autonomous robot operation method for non-destructive harvesting of hydroponic daffodil bulbs according to claim 8, characterized in that, In the grasping step, the edge computing controller switches the clamping force according to the operation stage: the first clamping force range is used in the vertical pulling stage of the planting basket, and the second clamping force range, which is greater than the first clamping force range, is switched to in the flipping operation stage.

10. The autonomous robot operation method for non-destructive harvesting of hydroponic daffodil bulbs according to claim 8, characterized in that, In the flipping and discharging step, the segmented flexible flipping control program executes three actions in sequence: the pre-tilting section flips from 0° to 10° using the first angular velocity; the discharging sliding section flips from 10° to the maximum flipping angle using the second angular velocity greater than the first angular velocity; and the stationary discharging section maintains the maximum flipping angle for 2s to 5s.

Images