A full-automatic carrying robot for flower pots of potted flowers

By designing a fully automated flowerpot handling robot, which uses camera and radar information fusion for flowerpot identification and positioning, the problem of high labor intensity and low efficiency caused by relying on manual labor for flowerpot handling has been solved. This has enabled efficient and precise flowerpot handling and placement, and improved the automation and intelligence level of facility horticulture.

CN122253138APending Publication Date: 2026-06-23JIANGSU UNIV
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202610345284.0
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2026-03-20
Publication Date
2026-06-23

AI Technical Summary

Technical Problem

Currently, the handling of potted plants mainly relies on manual labor, which results in high labor intensity, low efficiency, high cost, and poor precision, making it difficult to achieve mechanization and intelligentization of facility horticulture.

Method used

Design a fully automated flowerpot handling robot. The robot uses camera and radar information fusion for flowerpot identification and positioning, and combines computer vision for boundary line detection to achieve autonomous path planning. Through the integrated application of walking and moving mechanisms, flowerpot gripping mechanisms, perception and positioning systems and navigation systems, the robot can complete the autonomous handling of flowerpots.

Benefits of technology

It enables efficient and precise handling and placement of flower pots, reduces labor intensity, improves the automation and intelligence level of facility horticulture, lowers labor costs, adapts to complex outdoor environments, and improves the stability and reliability of robots.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN122253138A_ABST
    Figure CN122253138A_ABST
Patent Text Reader

Abstract

The application discloses a kind of full-automatic carrying robot of potted flower flowerpot, including walking and moving mechanism, flowerpot clamping mechanism, sensing and positioning system, navigation system and flowerpot placing system;Walking and moving mechanism is used to drive robot to move and adapt to topography fluctuation;Flowerpot clamping mechanism includes mechanical arm and clamping device, for performing the grabbing, lifting and releasing operation to flowerpot;Sensing and positioning system includes camera and laser radar, for identifying and positioning target flowerpot;Flowerpot placing system is electrically connected with walking and moving mechanism, flowerpot clamping mechanism and sensing and positioning system respectively, for controlling walking and moving mechanism and flowerpot clamping mechanism according to the information of sensing and positioning system, complete the autonomous grabbing and placing of flowerpot;Navigation system combines boundary line identification to carry out path planning and decision-making, realize the autonomous navigation of robot.The application realizes the full autonomous identification, grabbing, transport and placing of potted flower flowerpot by multi-sensor fusion and visual navigation technology.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] This invention belongs to the field of agricultural robots, specifically relating to a fully automated robot for transporting potted flower pots. Background Technology

[0002] As people's demand for flowers increases, the flower market continues to grow globally. However, the mechanization level of potted flower cultivation is currently low. The handling of pots and the thinning of pots mainly rely on manual labor, which results in high labor intensity, high labor costs, low efficiency, and poor placement accuracy. With the increasing age of practitioners and rising labor costs, the traditional production model of relying on manual labor for potted flower handling is becoming unsustainable.

[0003] How to effectively improve the mechanization, simplification, and intelligence of greenhouse horticulture production, and truly reduce labor intensity, lower labor costs, and increase production efficiency, is an urgent problem to be solved in the industry's development. The development of a potted flowerpot handling robot can improve the handling efficiency in the potted flowerpot production process and significantly reduce labor intensity. This is of great significance for improving the automation and intelligence level of potted flower production, helping flowerpot manufacturers reduce costs and increase efficiency, and promoting the upgrading of my country's greenhouse horticulture industry. Summary of the Invention

[0004] To address the aforementioned technical issues, this invention proposes a fully automated flowerpot handling robot. It utilizes camera and radar information fusion for flowerpot identification and positioning, while simultaneously applying computer vision for boundary line detection, enabling the robot to perform fully autonomous path planning and complete the autonomous handling of flowerpots.

[0005] The present invention achieves the above-mentioned technical objectives through the following technical means.

[0006] A fully automated robot for transporting potted flower pots includes a walking and moving mechanism, a flower pot gripping mechanism, a sensing and positioning system, a navigation system, and a flower pot placement system.

[0007] The walking and moving mechanism is used to drive the robot to move and adapt to the terrain undulations;

[0008] The flowerpot gripping mechanism includes a robotic arm and a gripping device, which is mounted on the robotic arm and is used to perform gripping, lifting and releasing operations on the flowerpot.

[0009] The perception and positioning system includes a camera and a lidar. The camera is used to identify the target flowerpot, and the lidar is used to locate the target flowerpot.

[0010] The navigation system is used to guide the robot to move within the work area;

[0011] The flowerpot placement system is electrically connected to the walking and moving mechanism, the flowerpot gripping mechanism, the sensing and positioning system, and the navigation system, respectively. It is used to control the walking and moving mechanism and the flowerpot gripping mechanism according to the information of the sensing and positioning system and the guidance information of the navigation system, so as to complete the autonomous grasping, transportation and placement of flowerpots.

[0012] In the above scheme, the walking and moving mechanism includes a shock absorption system;

[0013] The shock absorption system includes an upper fixed frame, a vertical plate, an upper connecting rod, a lower connecting rod, and a lower fixed frame; the two ends of the upper connecting rod are respectively hinged to the upper fixed frame and the upper end of the vertical plate, and the two ends of the lower connecting rod are respectively hinged to the lower end of the vertical plate and the lower fixed frame, together forming a parallel four-bar linkage; a shock absorption spring is installed between the lower connecting rod and the upper fixed frame.

[0014] Furthermore, the walking and moving mechanism also includes a DC brushed motor and a drive wheel;

[0015] The output shaft of the DC brushed motor is connected to the axle of the drive wheel via a universal joint to compensate for angular deviations when the shock absorption system is working.

[0016] Furthermore, the walking and moving mechanism also includes a base plate and a lidar mounting plate;

[0017] The lidar mounting plate is fixed to the bottom of the base plate by a support member, and defines an installation space for accommodating the lidar between the lidar and the flowerpot placement plate above, forming a hanging layout.

[0018] Furthermore, the robotic arm is a robotic arm based on a parallel four-bar linkage mechanism, including a front vertical beam, a rear vertical beam, and a lead screw stepper motor mounting plate;

[0019] The lower end of the front vertical beam is connected to the base plate, and the upper end of the front vertical beam is hinged to the front of the lead screw stepper motor mounting plate; the upper end of the rear vertical beam is hinged to the rear of the lead screw stepper motor mounting plate, and the lower end of the rear vertical beam is connected to the output end of the stepper motor; the base plate, the front vertical beam, the rear vertical beam, and the lead screw stepper motor mounting plate together form a parallel four-bar linkage mechanism; a tension spring is connected between the lead screw stepper motor mounting plate and the robot's internal support plate to reduce the load force when the stepper motor lifts the flowerpot.

[0020] Furthermore, the gripping device is a gripping device based on a linkage mechanism, including a lead screw stepper motor, a lead screw slider, a mechanical gripper control lever, a mechanical gripper rotating bar, and a mechanical gripper;

[0021] The lead screw stepper motor drives the lead screw slider to move in a straight line; the lead screw slider is hinged to the mechanical claw control lever through a connecting rod; the mechanical claw control lever is fixedly connected to the mechanical claw rotating bar, and the mechanical claw rotating bar is rotatably mounted on the mechanical claw support; the mechanical claw is fixed to the end of the mechanical claw rotating bar; a flexible pressure sensor is fixed on the inner side of the mechanical claw to sense the gripping force in real time, so as to realize adaptive clamping of flower pots of different sizes.

[0022] In the above scheme, the perception and positioning system is configured to perform a multi-sensor fusion method for flowerpot identification and positioning, the method comprising:

[0023] The YOLO object detection algorithm is used to identify images captured by the camera and obtain the pixel coordinate information of the flowerpot.

[0024] Clustering and circular fitting are performed on the point cloud data collected by lidar to obtain the position information of the flowerpot in the lidar coordinate system;

[0025] The coordinates of the flowerpots obtained by the camera are transformed into the coordinate system of the lidar by using the calibrated homography matrix;

[0026] The Hungarian algorithm is used to perform cross-modal data association and matching on flowerpot targets detected by cameras and lidar.

[0027] Kalman filtering is used to fuse the matched camera observation information and lidar observation information to obtain an accurate positional state estimate of the flowerpot.

[0028] Furthermore, in the step of clustering and circular fitting the lidar point cloud data, the distance-adaptive DBSCAN clustering algorithm is used, and the relationship between its neighborhood radius ε and the Euclidean distance d from the point cloud to the lidar is as follows:

[0029]

[0030] An iterative reweighted least squares method is used for circular fitting, and the sample point weights are dynamically adjusted based on the fitting residuals to reduce the influence of outliers.

[0031] In the above scheme, the navigation system is configured to perform the following navigation implementation method:

[0032] Use red boundary lines to enclose the work area into a rectangle, with one side of the rectangle being the area to be moved and the other side being the area where items have already been placed.

[0033] After the robot picks up the flowerpot in the waiting area, it moves towards the boundary line;

[0034] Upon reaching the boundary line, the vision processing flow is activated to detect the red boundary line in real time and generate tracking control commands, enabling the robot to move along the inside of the boundary line towards the already placed area.

[0035] After the robot arrives at the designated area, it will perform the task of placing the flower pots.

[0036] After placement, the robot returns to the transport area along the boundary line to begin the next transport cycle.

[0037] In the above scheme, the flowerpot placement system is configured to perform the following flowerpot placement and positioning method:

[0038] After the robot arrives at the placement area, it uses a perception and positioning system to detect the positions of the flower pots already placed and generates a set of local flower pot positions.

[0039] Principal component analysis was used to infer the orientation angle of the overall grid layout from the local flowerpot positions. ;

[0040] According to the direction angle Construct a rotation matrix to convert the world coordinates of the placed flower pots into grid coordinates aligned with the grid, and calculate the discrete grid index corresponding to each flower pot;

[0041] Construct and maintain an occupancy status matrix, and update the occupancy status in the matrix based on the detected flower pots and their grid indices;

[0042] Based on the occupancy state matrix, the next vacant grid position is searched as the target placement position according to the row priority strategy;

[0043] The grid index of the target placement position is transformed back to the coordinates in the world coordinate system through an inverse rotation transformation, which is then used as the target position for the robot to place the flower pot.

[0044] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0045] This invention achieves dynamic compensation for motor torque by introducing a tension spring mechanism. The tension spring stores energy under low load and releases energy under high load, thereby significantly reducing the maximum output torque required by the motor. This not only allows for smaller and more economical motor selection but also effectively improves the motor's reliability and lifespan by buffering impact loads.

[0046] After lifting the flowerpot, the robotic arm moves backward to securely place it on the front platform. This action brings the load's center of gravity closer to the machine body, significantly reducing the tipping moment during transport and improving overall operational stability and load capacity.

[0047] The shock absorption system designed in this invention can effectively absorb and buffer the impact and vibration generated by uneven or muddy roads on the robot. While ensuring the stability of the chassis movement, it significantly reduces the transmission of vibration to the upper precision components such as the robotic arm and lidar. This ensures that the robot can maintain high-precision positioning and stable grasping operation in complex outdoor terrain, and improves the overall adaptability and reliability of the system in unstructured environments.

[0048] This invention utilizes advanced deep learning algorithms, laser point cloud processing methods, and multi-sensor fusion technology to provide an automated, high-precision outdoor potted planter handling robot. The robot can accurately locate and grasp pots of different sizes and specifications.

[0049] This invention arranges a rectangular area in the flowerpot cultivation base and combines a multi-sensor fusion perception scheme to achieve autonomous robot navigation, enabling the robot to stably and efficiently complete the entire automated operation within the set area.

[0050] This invention is applied to large-scale flowerpot cultivation bases, assisting in the simple and repetitive tasks of moving and sorting potted flowerpots, significantly reducing labor intensity. It is of great significance for improving the automation and intelligence level of potted flower production, helping flowerpot manufacturers to reduce costs and increase efficiency, and promoting the upgrading of my country's facility horticulture industry. Attached Figure Description

[0051] Figure 1 This is a schematic diagram of the overall structure of the present invention;

[0052] Figure 2 This is a top view of the walking and moving mechanism;

[0053] Figure 3 A front view schematic diagram of the walking and moving mechanism;

[0054] Figure 4 This is a schematic diagram of a robotic arm;

[0055] Figure 5 This is a schematic diagram of a mechanical gripper;

[0056] Figure 6 A schematic diagram of the auxiliary and safety structure;

[0057] Figure 7 This is a schematic diagram of the internal structure of the control box;

[0058] Figure 8 Flowchart for camera-based flowerpot recognition;

[0059] Figure 9 Flowchart for locating flower pots using LiDAR;

[0060] Figure 10 A flowchart illustrating the information fusion process between cameras and LiDAR;

[0061] Figure 11 Flowchart for a robot picking up flower pots;

[0062] Figure 12 This is a diagram of a dual-closed-loop PID control framework.

[0063] Figure 13 A flowchart of the robot's operation process;

[0064] Figure 14 This is a schematic diagram of the robot's operating route.

[0065] In the diagram, 1. Camera; 2. Gripping device; 2-1. Lead screw stepper motor; 2-2. Slider; 2-3. Connecting rod; 2-4. Mechanical gripper control lever; 2-5. Mechanical gripper support; 2-6. Mechanical gripper; 2-7. Limit sensor; 2-8. Mechanical gripper rotating bar; 2-9. Flexible pressure sensor; 3. LiDAR; 4. Walking and moving mechanism; 4-1. Auxiliary wheel; 4-2. Flowerpot placement plate; 4-3. Base plate; 4-4. Shock absorption system; 4-4-1. Upper fixed frame; 4-4-2. Upper connecting rod; 4-4-3. Compression spring; 4-4-4. Vertical plate; 4-4-5. Universal joint; 4-4-6. Lower connecting rod; 4-4-7. Lower fixed frame; 4-5. Drive wheel; 4-6. DC brushed motor 4-7. LiDAR mounting plate; 4-8. Connecting plate; 4-9. DC brushed motor bracket; 5. Auxiliary and safety structure; 5-1. Top plate; 5-2. Left side plate; 5-3. Mudguard; 5-4. Internal support plate; 5-5. Right side plate; 6. Robotic arm; 6-1. Rear vertical beam; 6-2. Stepper motor; 6-3. Stepper motor bracket; 6-4. Bearing seat; 6-5. Front vertical beam; 6-6. Lead screw stepper motor mounting plate; 6-7. Limit sensor; 7. Flower pot; 8. Tension spring; 9. Control box; 9-1. 24V voltage regulator module; 9-2. 12V voltage regulator module; 9-3. Lithium battery; 9-4. Controller; 9-5. IMU sensor; 9-6. Deep learning computing unit. Detailed Implementation

[0066] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. The described embodiments are only some embodiments of the present invention, and not all embodiments.

[0067] In the description of this invention, unless otherwise explicitly specified and limited, the terms "installation," "connection," "linking," "fixing," etc., should be interpreted broadly. For example, they can refer to fixed connections or detachable connections; mechanical connections or electrical connections; direct connections or indirect connections through an intermediate medium. Those skilled in the art can understand the specific meaning of the above terms in this invention according to the specific circumstances.

[0068] This invention relates to a fully automated flowerpot handling robot that can autonomously identify and move flowerpots from one location to another. It is suitable for gardens, nurseries, greenhouses, and other similar locations. The robot integrates advanced sensor technology, navigation algorithms, and a robotic arm control system to ensure efficient and safe completion of handling tasks.

[0069] like Figure 1 The figure shown is a preferred embodiment of the fully automatic outdoor potted planter handling robot of the present invention, including a walking and moving mechanism 4, a flower pot gripping mechanism, a sensing and positioning system, a navigation system and a flower pot placement system;

[0070] The walking and moving mechanism 4 is used to drive the robot to move and adapt to the terrain undulations;

[0071] The flowerpot gripping mechanism includes a robotic arm 6 and a gripping device 2. The gripping device 2 is mounted on the robotic arm 6 and is used to perform gripping, lifting and releasing operations on the flowerpot 7.

[0072] The perception and positioning system includes a camera 1 and a lidar 3. The camera 1 is used to identify the target flowerpot, and the lidar 3 is used to locate the target flowerpot.

[0073] The navigation system is used to guide the robot to move within the work area;

[0074] The flowerpot placement system is electrically connected to the walking and moving mechanism 4, the flowerpot gripping mechanism, the sensing and positioning system, and the navigation system, respectively. It is used to control the walking and moving mechanism 4 and the flowerpot gripping mechanism according to the information of the sensing and positioning system and the guidance information of the navigation system, so as to complete the autonomous grasping, transportation and placement of flowerpots.

[0075] The walking and moving mechanism 4 constitutes the mobile platform and basic load-bearing structure of the entire device, and is located at the bottom. The auxiliary and safety structure 5 is installed above and behind the walking and moving mechanism 4, forming the main frame and shell of the robot, used to enclose internal electrical components and provide external protection. The robotic arm 6 serves as the core operation execution unit; preferably, the base of the robotic arm 6 is rigidly fixed to the upper center of the walking and moving mechanism 4 by bolts. The gripping device 2 is installed at the end of the robotic arm 6, used to directly perform the gripping and releasing operations on the flowerpot 7.

[0076] like Figure 2 As shown, the walking and moving mechanism 4 includes a chassis, a drive unit, and auxiliary wheels 4-1. The robot's base plate 4-3 is clamp-shaped. At the clamp-shaped opening of the base plate 4-3, four copper pillars are used to fix the flowerpot placement plate 4-2 to the bottom of the base plate 4-3, forming a stepped layout. On the left and right sides of the longitudinal center line on the lower surface of the base plate 4-3, two DC brushed motor brackets 4-9 are symmetrically fixed. Each DC brushed motor bracket 4-9 is equipped with a DC brushed motor 4-6 (i.e., the walking motor) with a photoelectric encoder, forming a symmetrical drive layout. The drive wheel 4-5 and the DC brushed motor 4-6 are connected by a shock absorption system 4-4; the output shaft of each DC brushed motor 4-6 is connected to the axle of one drive wheel 4-5 through a universal joint 4-4-5. This universal joint transmission design allows the drive unit to compensate for the angular deviation between the motor shaft and the axle when adapting to terrain undulations, ensuring continuous and stable power transmission.

[0077] To reduce the impact of uneven or muddy roads on the machine body and precision sensors, the DC brushed motor bracket 4-9 is not rigidly fixed to the base plate 4-3. For example... Figure 3 As shown, specifically, the walking and moving mechanism 4 also includes a shock absorption system 4-4; the shock absorption system 4-4 includes an upper fixed frame 4-4-1, a vertical plate 4-4-4, an upper connecting rod 4-4-2, a lower connecting rod 4-4-6, and a lower fixed frame 4-4-7; the two ends of the upper connecting rod 4-4-2 are respectively hinged to the upper fixed frame 4-4-1 and the upper end of the vertical plate 4-4-4, and the two ends of the lower connecting rod 4-4-6 are respectively hinged to the lower end of the vertical plate 4-4-4 and the lower fixed frame 4-4-7, together forming a parallel four-bar linkage; a shock-absorbing spring 4-4-3 is installed between the lower connecting rod 4-4-6 and the upper fixed frame 4-4-1 to provide elastic buffering for the drive unit. When the drive wheel 4-5 encounters an obstacle or pothole, the entire drive unit can swing around the hinge point, and the compression spring 4-4-3 provides buffering and restoring force, forming an effective passive shock absorption system.

[0078] The walking and moving mechanism 4 also includes a DC brushed motor 4-6 and a drive wheel 4-5; the output shaft of the DC brushed motor 4-6 is connected to the axle of the drive wheel 4-5 through a universal joint 4-4-5.

[0079] The walking and moving mechanism 4 also includes a base plate 4-3 and a lidar fixing plate 4-7; the lidar fixing plate 4-7 is fixed to the bottom of the base plate 4-3 by a support member, and defines an installation space for accommodating the lidar 3 between itself and the flowerpot placement plate 4-2 above.

[0080] Two spindle-shaped polyurethane auxiliary wheels 4-1 are symmetrically installed on both sides of the front of the base plate 4-3 via left and right connecting plates 4-8. This spindle-shaped design aims to ensure support stability while minimizing rolling friction with the ground and steering resistance.

[0081] At the front center of the base plate 4-3, there is a downward-curving notch. Four copper pillars extend downwards to connect to a lidar mounting plate 4-7. A single-line lidar is horizontally mounted under this plate, with its scanning plane roughly the height of a typical flowerpot. This "undermounted" layout lowers the overall center of gravity and provides the lidar with a wide-angle forward field of view unobstructed by the vehicle body.

[0082] The drive unit's active wheels 4-5 are pneumatic rubber wheels with a diameter of 213mm; a 24V DC brushed motor is used with a reduction ratio of 1:100 and a rated output speed of 90rpm. According to kinematic principles, the robot's theoretical maximum walking speed can be expressed by the formula:

[0083]

[0084] Calculation, where For wheel diameter, Let the deceleration wheel axle speed be the deceleration wheel speed. Substituting, we get:

[0085]

[0086] Therefore, the robot's maximum walking speed is approximately 1 m / s, which meets the design requirements for movement speed in the task. This speed value is a theoretical value under rated voltage and unloaded flat road conditions. In actual operation, it may vary slightly due to factors such as load, ground adhesion, and power fluctuations, but it can still ensure that the efficiency and stability requirements are met in typical operating scenarios.

[0087] like Figure 4 and Figure 5 As shown, the flowerpot gripping mechanism is the core execution unit for the robot to perform gripping and placing operations, including a robotic arm 6 and a gripping device 2.

[0088] The robotic arm 6 adopts a parallel four-bar linkage configuration. The robotic arm 6 includes a front vertical beam 6-5, a rear vertical beam 6-1, and a lead screw stepper motor mounting plate 6-6. The lower end of the front vertical beam 6-5 is connected to the base plate 4-3, and the upper end of the front vertical beam 6-5 is hinged to the front of the lead screw stepper motor mounting plate 6-6. The upper end of the rear vertical beam 6-1 is hinged to the rear of the lead screw stepper motor mounting plate 6-6, and the lower end of the rear vertical beam 6-1 is connected to the output end of the stepper motor 6-2. The base plate 4-3, the front vertical beam 6-5, the rear vertical beam 6-1, and the lead screw stepper motor mounting plate 6-6 together form a parallel four-bar linkage mechanism. A tension spring 8 connects the lead screw stepper motor mounting plate 6-6 to the robot's internal support plate 5-4. The parallel four-bar linkage configuration is formed by hinged front vertical beam 6-5, rear vertical beam 6-1, and the top lead screw stepper motor mounting plate 6-6. Specifically, the lower end of the front vertical beam 6-5 is connected to the base plate 4-3 via a bearing seat 6-4, and the upper end is hinged to the front end of the lead screw stepper motor mounting plate 6-6 via a pin. The upper end of the rear vertical beam 6-1 is hinged to the rear end of the lead screw stepper motor mounting plate 6-6 via a pin, and the lower end of the rear vertical beam 6-1 is fixed to the output shaft of a double-output-shaft stepper motor 6-2 via a flange. The stepper motor 6-2 is fixed to the base plate 4-3 by a bracket. A tension spring 8 connects the lead screw stepper motor mounting plate 6-6 to the robot's internal support plate 5-4.

[0089] Three high-strength tension springs 8 are connected between the lead screw stepper motor mounting plate 6-6 and the internal support plate 5-4. Their core function is to dynamically compensate for the drive torque: in the initial stage when the robotic arm 6 lifts the load (flowerpot) (peak torque demand point), the tension springs 8 are stretched to store energy, sharing the motor load; in the middle and later stages of lifting, the tension springs 8 release energy to assist in the lifting. This design significantly reduces the required peak torque specification of the motor and effectively buffers impact loads, improving system reliability and lifespan.

[0090] A limit sensor 6-7 is provided at the lowest point of the movement trajectory of the lead screw stepper motor mounting plate 6-6, which is used to calibrate the absolute zero position (standby position) of the robotic arm when the system is powered on.

[0091] The flowerpot gripping mechanism adopts an adaptive parallel gripper configuration. The gripping device 2 includes a lead screw stepper motor 2-1, a lead screw slider 2-2, a mechanical gripper control lever 2-4, a mechanical gripper rotating bar 2-8, and a mechanical gripper 2-6. The lead screw stepper motor 2-1 drives the lead screw slider 2-2 to move linearly. The lead screw slider 2-2 is hinged to the mechanical gripper control lever 2-4 via a connecting rod 2-3. The mechanical gripper control lever 2-4 is fixedly connected to the mechanical gripper rotating bar 2-8, which is rotatably mounted on the mechanical gripper support 2-5. The mechanical gripper 2-6 is fixed to the end of the mechanical gripper rotating bar 2-8. A flexible pressure sensor 2-9 is fixed to the inner side of the mechanical gripper 2-6 for real-time sensing of the gripping force. Specifically, the mechanical gripper support 2-5 is fixed to the front side of the lead screw stepper motor mounting plate 6-6, and the lead screw stepper motor 2-1 is vertically fixed to the lead screw stepper motor mounting plate 6-6. The slider 2-2 on the lead screw is hinged to the rear side of a pair of symmetrically arranged connecting rods 2-3; the front end of each connecting rod 2-3 is hinged to the corresponding mechanical claw operating rod 2-4, and the front end of the operating rod 2-4 is fixedly connected to the mechanical claw rotating bar 2-8. The rotating bar 2-8 is rotatably supported on the inner side of the mechanical claw support 2-5 by bearings on the upper and lower sides, so that the rotating bar 2-8 can rotate around its own axis. The mechanical claw 2-6 is directly welded or fixed to the outer end of the mechanical claw rotating bar 2-8 by a connector. When the lead screw stepper motor drives the slider 2-2 to move horizontally linearly back and forth, the linear motion of the slider 2-2 is converted into the rotation of the mechanical claw rotating bar 2-8 through the planar linkage mechanism formed by the mechanical claw connecting rod 2-3 and the operating rod 2-4, thereby driving the two mechanical claws 2-6 to perform synchronous opening and closing actions, realizing adaptive holding and release of the flowerpot. A flexible pressure sensor 2-9 is attached to the inner side of each mechanical gripper 2-6 to monitor the clamping force in real time, realize closed-loop force control, and prevent damage to the flowerpot or insufficient clamping force. A limit sensor 2-7 is provided at the fully closed position of the gripper for stroke definition and initial positioning.

[0092] like Figure 6As shown, the fully automated potted plant flowerpot handling robot also includes the auxiliary and safety structure 5, which provides protection and a mounting base for internal components. An internal support plate 5-4, made of bent aluminum alloy sheet, is perpendicular to the base plate 4-3 and rigidly connected to it by bolts, forming the main internal load-bearing skeleton of the robot. The outer shell includes a top plate 5-1, a left side plate 5-2, and a right side plate 5-5 made of 304 stainless steel. The left side plate 5-2 and the right side plate 5-5 are respectively connected to the edge of the base plate 4-3 and the side of the internal support plate 5-4 by bolts. The top plate 5-1 is welded to the left side plate 5-2 and the right side plate 5-5 to form an integrated protective structure. Above the two drive wheels 4-5, there are arc-shaped sheet metal mudguards 5-3, which are welded to the left side plate 5-2 and the right side plate 5-5 respectively, effectively preventing mud and water from splashing into the machine body during operation.

[0093] The fully automated flower pot handling robot also includes a control box 9; such as Figure 7 As shown, the control box 9 is located at the rear of the robot and includes a deep learning computing unit 9-6, a controller 9-4, a 24V voltage regulator module 9-1, a 12V voltage regulator module 9-2, an IMU sensor 9-5, and a lithium battery 9-3. The deep learning computing unit 9-6 receives data from the sensor system, performs calculations and information fusion to obtain the robot's walking path and command issuance. The controller 9-4 controls the movement of the robot's chassis motor and flowerpot gripping mechanism to achieve the gripping, lifting, and placement of the flowerpot.

[0094] In a preferred embodiment of the present invention, the fully automated flowerpot handling robot further includes a power and energy system; the power and energy system uses a 24V lithium battery 9-3 with a capacity of 15Ah as the main power source. The main power source outputs a stable 24V voltage through a 24V voltage regulator module 9-1 to power the DC brushed motor 4-6; simultaneously, after being stepped down by a 12V voltage regulator module 9-2, it provides 12V operating power to the deep learning computing unit 9-6 and the controller 9-4.

[0095] In a preferred embodiment of the present invention, the fully automated flowerpot handling robot further includes a human-machine interaction system; the human-machine interaction system includes a 7-inch touchscreen, which is connected to a deep learning computing unit 9-6. The touchscreen displays a host computer interface developed using QT, which can display some status information of the robot in real time; the operator can set flowerpot placement parameters for the robot through the touchscreen, including the flowerpot spacing and arrangement.

[0096] This invention integrates artificial intelligence and embedded technology, achieving a high degree of automation. A touchscreen on the robot's control box 9 allows staff to easily debug the robot and configure its operating modes, followed by one-button start-up. In large-scale flowerpot cultivation bases, after simple debugging and configuration, it can perform automated operations, accurately identifying and transporting flowerpots and placing them in designated locations. Furthermore, the robot can arrange flowerpots into various shapes according to commands set by staff, making it particularly suitable for exhibitions and similar venues.

[0097] In outdoor flowerpot handling scenarios, factors such as complex lighting, diverse targets, and mutual occlusion make it difficult for a single sensor to meet the requirements for high-precision and robust positioning. Specifically: Camera limitations: While vision-based target recognition can provide semantic information, monocular ranging accuracy is limited and easily affected by lighting changes, making it difficult to stably output the precise three-dimensional coordinates required for robotic gripper grasping. LiDAR limitations: LiDAR can provide accurate distance and angle information, but lacks semantic recognition capabilities, easily confusing flowerpots with other curved objects in the environment, and prone to missed detection under occlusion conditions. Therefore, this invention integrates the semantic recognition capabilities of a camera with the precise geometric measurement capabilities of LiDAR. Through coordinate transformation and data association, each detected target is simultaneously assigned accurate "identity" and "location" information, thereby generating more reliable and accurate flowerpot positioning results in complex environments, providing crucial perception support for subsequent precise grasping.

[0098] Preferably, the camera 1 is located at the top of the robot's upper plate and is used to identify the target flowerpot; the lidar 3 is located below the flowerpot support plate and is used to locate the target flowerpot.

[0099] The sensing and positioning system is configured to perform a flowerpot identification and positioning method using multi-sensor fusion, the method comprising:

[0100] The YOLO object detection algorithm is used to identify the image captured by camera 1 and obtain the pixel coordinate information of the flowerpot.

[0101] Clustering and circular fitting were performed on the point cloud data collected by LiDAR 3 to obtain the position information of the flowerpot in the LiDAR coordinate system;

[0102] The coordinates of the flowerpots obtained by the camera are transformed into the coordinate system of the lidar by using the calibrated homography matrix;

[0103] The Hungarian algorithm is used to perform cross-modal data association and matching on flowerpot targets detected by cameras and lidar.

[0104] Kalman filtering is used to fuse the matched camera observation information and lidar observation information to obtain an accurate positional state estimate of the flowerpot.

[0105] In the step of clustering and circular fitting the lidar point cloud data, the distance-adaptive DBSCAN clustering algorithm is used. The relationship between its neighborhood radius ε and the Euclidean distance d from the point cloud to the lidar is as follows:

[0106]

[0107] An iterative reweighted least squares method is used for circular fitting, and the sample point weights are dynamically adjusted based on the fitting residuals to reduce the influence of outliers.

[0108] The navigation system is configured to perform the following navigation implementation method:

[0109] Use red boundary lines to enclose the work area into a rectangle, with one side of the rectangle being the area to be moved and the other side being the area where items have already been placed.

[0110] After the robot picks up the flowerpot in the waiting area, it moves towards the boundary line;

[0111] Upon reaching the boundary line, the vision processing flow is activated to detect the red boundary line in real time and generate tracking control commands, enabling the robot to move along the inside of the boundary line towards the already placed area.

[0112] After the robot arrives at the designated area, it will perform the task of placing the flower pots.

[0113] After placement, the robot returns to the transport area along the boundary line to begin the next transport cycle.

[0114] The flowerpot placement system includes a chassis motion control subsystem and a robotic arm motion control subsystem. The chassis motion control subsystem acquires motion control commands and generates corresponding motor drive signals to control the chassis motor to perform translational or rotational movements. The robotic arm motion control subsystem acquires the operation control commands and generates corresponding joint control signals and end effector control signals to drive the robotic arm 6 to perform spatial pose adjustments and control the robotic gripper to perform grasping or releasing operations.

[0115] The flowerpot placement system is configured to perform the following flowerpot placement and positioning method:

[0116] After the robot arrives at the placement area, it uses a perception and positioning system to detect the positions of the flower pots already placed and generates a set of local flower pot positions.

[0117] Principal component analysis was used to infer the orientation angle of the overall grid layout from the local flowerpot positions. ;

[0118] According to the direction angle Construct a rotation matrix to convert the world coordinates of the placed flower pots into grid coordinates aligned with the grid, and calculate the discrete grid index corresponding to each flower pot;

[0119] Construct and maintain an occupancy status matrix, and update the occupancy status in the matrix based on the detected flower pots and their grid indices;

[0120] Based on the occupancy state matrix, the next vacant grid position is searched as the target placement position according to the row priority strategy;

[0121] The grid index of the target placement position is transformed back to the coordinates in the world coordinate system through an inverse rotation transformation, which is then used as the target position for the robot to place the flower pot.

[0122] The camera 1 identifies and locates the flowerpot as follows: Figure 8 As shown, the specific steps include the following:

[0123] S1.1: Collect RGB images of flowerpots at different times during the use of the flowerpot handling robot, and use Labelimg software to annotate each image to mark all flowerpots in the image;

[0124] S1.2: Use four methods to augment the image: brightness and contrast adjustment, saturation and hue adjustment, Gaussian blur, and random noise, to obtain a more versatile and robust training dataset.

[0125] S1.3: After obtaining enough image samples, the images are randomly divided, with 80% allocated to the training set, 10% to the validation set, and 10% to the test set.

[0126] S1.4: Construct a YOLO11-based object detection model and train it using a pre-defined dataset. In a preferred embodiment, the training parameters of the model are set as follows: the input image size is uniformly 640×640 pixels; the SGD stochastic gradient descent optimizer is used, with an initial learning rate of 0.01, a momentum parameter of 0.937, and a weight decay coefficient of 0.0005; the batch size is set to 8; and the model is trained for a total of 250 epochs. After training, an object detection model capable of recognizing flower pots is obtained. The model is validated using a test set, and the pixel coordinates and confidence scores of the flower pots in the image can be obtained.

[0127] S1.5: Using Zhang Zhengyou's chessboard calibration method, the camera's intrinsic parameter matrix K and distortion coefficients are solved by taking chessboard images from multiple angles, thereby correcting the influence of the camera's own distortion on recognition.

[0128] The laser radar 3 positions the flowerpot as follows: Figure 9 As shown, the specific steps include the following:

[0129] S2.1: In order to focus on the target area in front, the scanning angle of LiDAR 3 is limited, and only the point cloud data within a 120-degree range directly in front of it is retained for flower pot detection;

[0130] S2.2: Collect point cloud data within the range of step S2.1, and filter it to remove outliers;

[0131] S2.3: The DBSCAN clustering algorithm is used to cluster the point cloud. Addressing the characteristics of densely packed greenhouse potted plants and significant differences in point cloud density distribution at different distances, the traditional DBSCAN algorithm is improved by proposing a distance-adaptive parameter adjustment method. The traditional fixed neighborhood radius is prone to missed detections at long distances and false clustering at short distances. The core improvement lies in adjusting the neighborhood radius... Designed as a linear function related to the target distance:

[0132]

[0133] Where d represents the Euclidean distance from the point cloud to the lidar (unit: meters), this formula is applicable at close range ( When smaller, assign a smaller value Values ​​are added to improve the ability to distinguish densely arranged potted plants; at long distances ( Automatically increases when (larger) The value is adjusted to enhance the connectivity of sparse point clouds. Through this adaptive mechanism, the algorithm can maintain stable clustering performance within different distance ranges, effectively improving the segmentation accuracy of greenhouse potted plant point clouds.

[0134] S2.4: Perform circular fitting on each cluster of point clouds to obtain the coordinate information of the flowerpot. Considering the characteristics of noise and outliers in the lidar point cloud, the weighted least squares method is used for circular fitting, and iteratively reweighted least squares (IRLS) optimization is introduced.

[0135] In the initial stage of the iteration, the weights of all sample points Initialize to 1, then calculate the residuals at each point based on the current fitting result. The weights are dynamically updated based on the residual size. This reduces the impact of outliers on the fitting results. The weighting function is defined as follows:

[0136]

[0137] in, This represents the standard deviation of the residual in the current iteration. When the residual exceeds... When the weight of a point decreases exponentially, the influence of outliers on the fitting results will gradually be reduced.

[0138] After each weight update, a weighted least squares circle fit is performed again, and the center and radius parameters are updated. The iterative process terminates when any of the following conditions are met:

[0139] (1) The change in the center position between two consecutive iterations is less than the set threshold. ;

[0140] (2) The change in radius is less than the set threshold;

[0141] (3) Reach the preset maximum number of iterations.

[0142] In this embodiment, the maximum number of iterations was set to 3. After iterative optimization, the radius estimation error was reduced from ±2.1cm to ±0.8cm, and the center positioning accuracy was improved by approximately 62%.

[0143] By fusing the identification and positioning information of the flowerpot from camera 1 with the positioning information from lidar 3, more accurate flowerpot positioning information is obtained, such as... Figure 10 As shown, the specific steps are as follows:

[0144] S3.1: Take four points on the ground directly in front of the robot. Based on the position relationship between the points on the ground in the lidar coordinate system and the position in the camera pixel coordinate system, calculate the homography matrix that includes the transformation between the two coordinate systems, so as to transform the flower pot coordinates obtained by the camera into the lidar coordinate system.

[0145] S3.2: To achieve accurate matching of the flowerpot target in the lidar point cloud coordinates and the camera image coordinates, the Hungarian algorithm is used to perform cross-modal data association on the detection results of the two sensors. By constructing a cost function that fuses spatial consistency and detection reliability, target matching between heterogeneous sensors is achieved.

[0146] Specifically, let lidar 3 detect... A flowerpot target was detected by camera 1. For each flowerpot target, construct a cost matrix.

[0147]

[0148] Among them, matrix elements Indicates the first The lidar target and the first The matching cost between camera targets is calculated using the following formula:

[0149]

[0150] in:

[0151] The weighting coefficients for spatial distance cost. Weighting coefficients for detecting confidence level differences. and satisfy:

[0152]

[0153] For the first The two-dimensional planar coordinates (x, y) of the target detected by a lidar are given by the lidar polar coordinates. The result of the conversion is:

[0154]

[0155] in, The polar radius is obtained from the lidar measurement, representing the straight-line distance from the lidar to the target point; The polar angle (angle value) obtained by the lidar is usually referenced by the lidar's own orientation.

[0156] For the first Each camera detects the target's position vector in the camera coordinate system, and a coordinate transformation function is established based on the extrinsic parameter calibration relationship between the camera and the LiDAR. This is used to map the position of the target detected by the camera to the lidar coordinate system;

[0157] This is the distance normalization factor, which takes the value of the average distance between flowerpot targets in the scene. It is used to eliminate the influence of different units on the cost function.

[0158] The difference in confidence levels between the two modalities:

[0159]

[0160] in:

[0161] The target category probability value output by the camera target detection network;

[0162] The confidence score for lidar detection is calculated based on the point cloud clustering stability and geometric fit quality:

[0163]

[0164] in:

[0165] For the first A point cloud set of flowerpot targets;

[0166] This represents the number of cluster points;

[0167] The point normalization constant;

[0168] The average residual is obtained by performing a circular fit on the point cloud;

[0169] This is the residual normalization factor;

[0170] These are the weighting coefficients.

[0171] Considering that cross-modal calibration errors may lead to spatial position deviations, a distance constraint is introduced to correct the cost matrix:

[0172]

[0173] Among them, threshold Based on the sensor calibration accuracy, during the camera and lidar calibration process, the average reprojection error of multiple sets of calibration points is statistically analyzed. As a calibration error estimate, and set as follows:

[0174]

[0175] in, This is an empirical coefficient.

[0176] Subsequently, the target matching problem is modeled as an assignment problem of minimizing cost, and the following optimization model is solved using the Hungarian algorithm:

[0177]

[0178] Under the following constraints:

[0179]

[0180] in, To match the indicator variable, when the first The lidar target and the first The value is 1 when the target is matched by a camera, and 0 otherwise.

[0181] For the revised first The lidar target and the first The matching cost between camera targets;

[0182] M represents the number of flower pots detected by the lidar, and N represents the number of flower pots detected by the camera.

[0183] Through the above optimization solution, the global optimal matching relationship between the lidar detection results and the camera detection results can be obtained, thereby realizing cross-modal target association and providing a reliable data association basis for subsequent multi-sensor fusion positioning;

[0184] S3.3: A Kalman filter framework is used to fuse the flowerpot observation information from the camera and lidar. The flowerpot state is modeled as... ,in Let be the coordinates of the flowerpot's position in a two-dimensional plane. For flower pots in The speed of movement in a certain direction.

[0185] The state evolution and observation model is as follows:

[0186]

[0187] in, for The system state vector at time t. Let be the state transition matrix, describing the state transition from... Time's up The evolutionary pattern at any given moment is constructed based on a uniform motion model; Let be the process noise vector, representing the uncertainty of the state evolution model. It is assumed to be zero-mean Gaussian white noise, and its covariance matrix is ​​. ; for The observation vector at time t, ,in The coordinates of the flowerpot's planar position obtained by the camera. The coordinates of the flowerpot's planar position obtained from lidar detection; The observation matrix maps the state vector to the observation space, reflecting the linear relationship between state variables and observed variables. : Observation noise vector, representing the uncertainty of the observations, assumed to be zero-mean Gaussian white noise, with its covariance matrix as follows. ;

[0188] State transition matrix Defined as:

[0189]

[0190] in, , is the sensor sampling period.

[0191] The fused observation vector is constructed as follows:

[0192]

[0193] in, The coordinates of the flowerpot's planar position are obtained by camera detection and coordinate transformation. These are the planar position coordinates of the flowerpot obtained by converting LiDAR polar coordinate data.

[0194] Observation matrix Defined as:

[0195]

[0196] Observation noise covariance matrix It is a diagonal array, and the settings are as follows based on the sensor characteristics:

[0197] Camera observation noise standard deviation: ;

[0198] Standard deviation of lidar observation noise: ;

[0199] Process noise covariance matrix The process noise intensity coefficient is obtained by discretizing the constant velocity model. ;

[0200] When the new flowerpot is detected for the first time, initialize the state vector and covariance matrix:

[0201]

[0202] in, Take the fusion result of the camera and lidar observations during the first detection, and initialize the speed to 0.

[0203] Kalman filtering achieves optimal estimation through prediction and update recursive formulas, and integrates position observations from cameras and lidar to achieve accurate tracking of the flowerpot's status.

[0204] In one embodiment of the present invention, preferably, the robot's flowerpot grasping system, such as... Figure 11 As shown, the specific steps include:

[0205] S4.1: Identify and locate the target flowerpots through the perception and positioning system (including camera 1 and lidar 3) to obtain the precise location information of all flowerpots;

[0206] S4.2: Select the flowerpot closest to itself from all merged flowerpot coordinates;

[0207] S4.3: Based on the relative position of this coordinate with the robot itself, a dual closed-loop PID control framework is used, such as... Figure 12 As shown, the coordinates of the flowerpot are used as input to calculate the PWM values ​​of the two chassis motors 4-6; when the robot moves to a suitable position from the flowerpot, the grasping command takes effect.

[0208] S4.4: The host computer and the slave computer achieve bidirectional data exchange through the USART serial communication protocol. The host computer transmits PWM control commands for the dual motors of the chassis and control commands for the robotic arm to the slave computer, and the slave computer transmits real-time encoder feedback data of chassis motors 4-6 to the host computer.

[0209] S4.5: Control the mobile chassis (i.e., the walking and moving mechanism 4) to move to the flowerpot grasping preparation position, and adjust the overall posture of the robot to align it with the flowerpot;

[0210] S4.6: Control the robotic arm 6 to descend to the preset grasping height;

[0211] S4.7: Control the clamping device 2 to perform the clamping action and fix the flower pot;

[0212] S4.8: Control the robotic arm 6 to rise smoothly and move the flower pot to the transport height so that it sits precisely on the flower pot placement plate 4-2.

[0213] In one embodiment of the present invention, preferably taking a large-scale flowerpot cultivation base as an example, a robot navigation method based on multi-sensor fusion suitable for this scenario is selected. The robot operation flowchart and route are as follows: Figure 13 , 14 As shown, the specific steps include:

[0214] S5.1: Use a 2cm wide red cloth strip to enclose the flower pot handling area into a rectangle. One side of the rectangle is defined as the waiting area, where flower pots are piled up randomly; the other side is defined as the already placed area, used to neatly arrange the handled flower pots.

[0215] S5.2: The robot starts in the area to be transported. After picking up the flower pot, it uses the yaw angle information output by IMU sensor 9-5 as the direction feedback and moves initially towards the boundary line area.

[0216] S5.3: After the robot reaches the vicinity of the boundary line, the OpenCV-based vision processing flow is initiated to detect the red boundary line in real time and generate tracking control commands, causing the robot to move along the inside of the boundary line towards the placed area. The vision processing includes the following steps:

[0217] Color space conversion and thresholding: Convert the image to HSV space and perform thresholding based on red features;

[0218] Morphological closing operation: fills in broken areas of boundary lines to enhance continuity;

[0219] Canny edge detection: Extracts boundary line edge features;

[0220] Hough line detection: Identifying line segments in an image;

[0221] Line post-processing and filtering: merge collinear line segments, filter short line segments, calculate the slope of the line, and use Kalman filtering to smooth the detection results in time sequence.

[0222] S5.4: After the robot arrives at the placement area, it executes a differentiated placement strategy based on the scene status;

[0223] S5.5: After placing the flowerpots, the robot returns to the transport area along the boundary line. During the return process, OpenCV vision is used for line following, and the angle information from IMU sensors 9-5 is integrated with the odometer feedback from the wheel encoders (motors 4-6 have built-in encoders) of the walking and moving mechanism 4 to achieve precise posture control. After arriving at the starting point, the robot enters the next transport cycle.

[0224] In one embodiment of the present invention, preferably, the flowerpot placement system dynamically analyzes the spatial distribution of already placed flowerpots based on environmental information acquired in real time by a sensing module mounted on the robot, and then autonomously calculates the precise location where the next flowerpot should be placed. This process does not rely on preset global coordinates or manual calibration. The system integrates local observation data with global layout constraints to complete position planning and output online, achieving adaptive, high-precision placement under limited sensing conditions, effectively ensuring the neatness of the flowerpot array and the continuity of the operation. Specifically, it includes the following steps:

[0225] S6.1: When the robot arrives at the placement area carrying flower pots, it first activates the perception and localization system to detect and identify the positions of the placed flower pots. This process generates a local set of flower pot positions.

[0226] Plocal={(x i ,y i )|i=1,2,…,K}

[0227] in This represents the rectangular coordinates of the flowerpot in the robot's coordinate system;

[0228] This represents the number of flowerpots that can be clearly identified within the current field of view. To ensure the accuracy of subsequent calculations, a minimum detection threshold is set: if fewer than three flowerpots are detected, the robot will adopt one of two strategies: either move slowly within a safe range to expand the perception range and acquire more reference points; or temporarily continue execution using the layout information learned from previous tasks. This threshold setting is based on geometric principles—at least three non-collinear points are required to determine the orientation characteristics of a two-dimensional plane.

[0229] S6.2: Principal Component Analysis (PCA) is used to infer the overall grid orientation from locally cluttered observations. This mainly involves four steps:

[0230] S6.21: Calculate the center position of the point set: First, calculate the average value of all detected flowerpot positions. This point represents the geometric center of the local observation area.

[0231] in:

[0232] Represents all flowerpot targets Average coordinates;

[0233] Represents all flowerpot targets Average coordinates;

[0234] This indicates the total number of flowerpot targets detected;

[0235] Indicates the first The target of the flower pot coordinate;

[0236] Indicates the first The target of the flower pot coordinate;

[0237] An index representing the target flowerpot. ;

[0238] S6.22: Constructing the covariance matrix: The covariance matrix describes the distribution and dispersion of the flowerpot locations in two dimensions.

[0239]

[0240] S6.23: Extracting the principal direction: By performing eigenvalue decomposition on the covariance matrix, two eigenvalues ​​and their corresponding eigenvectors are obtained. The largest eigenvalue... Corresponding feature vector This indicates the main direction of the flowerpot arrangement. Mathematically speaking, this direction is where the data points are most dispersed, corresponding to the row or column direction of the arrangement array.

[0241] S6.24: Calculate the mesh orientation angle:

[0242]

[0243] in:

[0244] Representing the eigenvector exist Component of direction;

[0245] Representing the eigenvector exist Component of direction;

[0246] This angle This refers to the rotation angle of the grid relative to the world coordinate system. If the flowerpots are placed perfectly neatly and the sensor measurements are accurate, this angle should be very close to the actual placement orientation.

[0247] S6.3: To perform discretized position management, the continuous world coordinate system is converted into a discrete grid coordinate system. Based on the angle obtained in the previous step... Construct a two-dimensional rotation matrix to rotate any point. The angle is adjusted so that the rotated coordinate axes are aligned with the direction of the placed grid.

[0248] For each detected flowerpot location First, translate it relative to the global reference point, then perform a rotation transformation:

[0249]

[0250] This is how it was obtained. This refers to the coordinates of the flowerpot in the "aligned" grid coordinate system. Ideally, all flowerpots would be located in this coordinate system. The coordinates should all be Integer multiples of, The coordinates should all be Integer multiples of . Then calculate the discrete grid index:

[0251]

[0252] The rounding function is used here to handle errors in actual measurements. Indicates column index (starting from 0). This represents the row index (starting from 0). For example, if calculated as follows: This indicates that the flowerpot should be located in the 3rd column and 4th row (assuming a 0-based index).

[0253] S6.4: To systematically manage the location information of placed flower pots, an occupancy state matrix is ​​introduced, creating a matrix of size [missing information]. Two-dimensional matrix (in For column numbers, (Number of rows). All elements are initialized to 0, indicating that all grid positions are empty. For each detected flowerpot and its calculated grid index... If the index is within the matrix range (i.e. and If the corresponding matrix element is set to 1, then the matrix element at the corresponding position will be set to 1. In actual operation, there may be two situations that require special handling: First, the same grid position is marked multiple times (which may be due to measurement error or repeated detection). In this case, only one mark is retained. Second, the index of a flower pot exceeds the matrix range. This may be due to improper selection of reference points or serious measurement error. The system will record such out-of-bounds events and form an abnormal data log for subsequent system debugging and parameter calibration.

[0254] S6.5: Based on the occupancy state matrix, a suitable location needs to be found to place the new flowerpot. A "row-first" search strategy is adopted, that is, starting from the first row, searching for empty spaces from left to right; if there are no empty spaces in the current row, then moving to the next row. This strategy simulates the natural habits of humans placing objects and also ensures the neatness of the arrangement.

[0255] S6.6: After finding the target mesh location, it needs to be converted back to world coordinates that the robot can understand, and the precise position in the mesh coordinate system needs to be calculated. Assume the index of the found empty space is... Then the coordinates in the grid coordinate system are:

[0256]

[0257] in:

[0258] Indicates the target flowerpot in the grid coordinate system Direction coordinates;

[0259] Indicates the target flowerpot in the grid coordinate system Direction coordinates;

[0260] The column index representing the detected location of the next free grid cell;

[0261] Indicates the row index of the next detected free grid position;

[0262] Indicates the grid in The unit spacing in the direction, i.e., the physical distance between adjacent grid columns;

[0263] Indicates the grid in The unit spacing in the direction is the physical distance between adjacent grid rows.

[0264] Perform an inverse rotation transformation to convert the mesh coordinates back to the original world coordinate system:

[0265]

[0266] in, It is the inverse of the rotation matrix, which rotates the coordinates back to their original orientation. This is how it is obtained. It refers to the target location that the robot needs to move to.

[0267] This invention uses a fusion of camera 1 and LiDAR 3 to identify the position of flowerpots, employs a flowerpot gripping mechanism to pick up flowerpots of different sizes, and combines visual line following and multi-sensor fusion navigation to achieve autonomous handling and neat placement of flowerpots within a rectangular work area. This invention introduces a tension spring 8 for load reduction, a parallel four-bar shock absorption system 4-4, and a rearward-shifted center of gravity design that places the flowerpot 7 on the flowerpot placement plate 4-2, improving operational stability and motor lifespan. Suitable for outdoor settings such as gardens and nurseries, this invention can significantly reduce labor intensity, improve handling efficiency and placement accuracy, and promote the automation and intelligent development of potted plant production.

[0268] It should be understood that although this specification is described according to various embodiments, not every embodiment contains only one independent technical solution. This way of describing the specification is only for clarity. Those skilled in the art should regard the specification as a whole. The technical solutions in each embodiment can also be appropriately combined to form other implementation methods that can be understood by those skilled in the art.

[0269] The embodiments described above are preferred embodiments of the present invention, but the present invention is not limited to the above embodiments. Any obvious improvements, substitutions or modifications that can be made by those skilled in the art without departing from the essence of the present invention shall fall within the protection scope of the present invention.

Claims

1. A fully automated robot for transporting potted flower pots, characterized in that, Includes a walking and moving mechanism (4), a flowerpot gripping mechanism, a sensing and positioning system, a navigation system, and a flowerpot placement system; The walking and moving mechanism (4) is used to drive the robot to move and adapt to the terrain undulations; The flowerpot gripping mechanism includes a robotic arm (6) and a gripping device (2). The gripping device (2) is mounted on the robotic arm (6) and is used to perform gripping, lifting and releasing operations on the flowerpot (7). The perception and positioning system includes a camera (1) and a lidar (3). The camera (1) is used to identify the target flowerpot, and the lidar (3) is used to locate the target flowerpot. The navigation system is used to guide the robot to move within the work area; The flowerpot placement system is electrically connected to the walking and moving mechanism (4), the flowerpot gripping mechanism, the sensing and positioning system, and the navigation system, respectively. It is used to control the walking and moving mechanism (4) and the flowerpot gripping mechanism according to the information of the sensing and positioning system and the guidance information of the navigation system, so as to complete the autonomous grasping, transportation and placement of the flowerpot.

2. The fully automated flowerpot handling robot according to claim 1, characterized in that, The walking and moving mechanism (4) includes a shock absorption system (4-4). The shock absorption system (4-4) includes an upper fixed frame (4-4-1), a vertical plate (4-4-4), an upper connecting rod (4-4-2), a lower connecting rod (4-4-6), and a lower fixed frame (4-4-7). The two ends of the upper connecting rod (4-4-2) are respectively hinged to the upper fixed frame (4-4-1) and the upper end of the vertical plate (4-4-4), and the two ends of the lower connecting rod (4-4-6) are respectively hinged to the lower end of the vertical plate (4-4-4) and the lower fixed frame (4-4-7), forming a parallel four-bar linkage. A shock-absorbing spring (4-4-3) is installed between the lower connecting rod (4-4-6) and the upper fixed frame (4-4-1).

3. The fully automated flowerpot handling robot according to claim 2, characterized in that, The walking and moving mechanism (4) also includes a DC brushed motor (4-6) and a drive wheel (4-5). The output shaft of the DC brushed motor (4-6) is connected to the axle of the drive wheel (4-5) via a universal joint (4-4-5).

4. The fully automated flowerpot handling robot according to claim 3, characterized in that, The walking and moving mechanism (4) also includes a base plate (4-3) and a lidar fixing plate (4-7). The lidar mounting plate (4-7) is fixed to the bottom plate (4-3) by a support member, and defines an installation space for accommodating the lidar (3) between itself and the flowerpot placement plate (4-2) above.

5. The fully automated flowerpot handling robot according to claim 4, characterized in that, The robotic arm (6) includes a front vertical beam (6-5), a rear vertical beam (6-1), and a lead screw stepper motor mounting plate (6-6). The lower end of the front vertical beam (6-5) is connected to the base plate (4-3), and the upper end of the front vertical beam (6-5) is hinged to the front of the lead screw stepper motor mounting plate (6-6); the upper end of the rear vertical beam (6-1) is hinged to the rear of the lead screw stepper motor mounting plate (6-6), and the lower end of the rear vertical beam (6-1) is connected to the output end of the stepper reduction motor (6-2); the base plate (4-3), the front vertical beam (6-5), the rear vertical beam (6-1), and the lead screw stepper motor mounting plate (6-6) together form a parallel four-bar linkage mechanism; a tension spring (8) connects the lead screw stepper motor mounting plate (6-6) to the robot's internal support plate (5-4).

6. The fully automated flowerpot handling robot according to claim 5, characterized in that, The gripping device (2) includes a lead screw stepper motor (2-1), a lead screw slider (2-2), a mechanical gripper control lever (2-4), a mechanical gripper rotating bar (2-8), and a mechanical gripper (2-6). The lead screw stepper motor (2-1) drives the lead screw slider (2-2) to move in a straight line; the lead screw slider (2-2) is hinged to the mechanical gripper control lever (2-4) through the connecting rod (2-3); the mechanical gripper control lever (2-4) is fixedly connected to the mechanical gripper rotating bar (2-8), and the mechanical gripper rotating bar (2-8) is rotatably mounted on the mechanical gripper support (2-5); the mechanical gripper (2-6) is fixed to the end of the mechanical gripper rotating bar (2-8); a flexible pressure sensor (2-9) is fixed on the inner side of the mechanical gripper (2-6) for real-time sensing of the gripping force.

7. The fully automated flowerpot handling robot according to claim 1, characterized in that, The sensing and positioning system is configured to perform a flowerpot identification and positioning method using multi-sensor fusion, the method comprising: The YOLO target detection algorithm is used to identify the image captured by the camera (1) and obtain the pixel coordinate information of the flower pot; Clustering and circular fitting are performed on the point cloud data collected by the lidar (3) to obtain the position information of the flower pot in the lidar coordinate system; The coordinates of the flowerpots obtained by the camera are transformed into the coordinate system of the lidar by using the calibrated homography matrix; The Hungarian algorithm is used to perform cross-modal data association and matching on flowerpot targets detected by cameras and lidar. Kalman filtering is used to fuse the matched camera observation information and lidar observation information to obtain an accurate positional state estimate of the flowerpot.

8. The fully automated flowerpot handling robot according to claim 7, characterized in that, In the step of clustering and circular fitting the lidar point cloud data, the distance-adaptive DBSCAN clustering algorithm is used. The relationship between its neighborhood radius ε and the Euclidean distance d from the point cloud to the lidar is as follows: An iterative reweighted least squares method is used for circular fitting, and the sample point weights are dynamically adjusted based on the fitting residuals to reduce the influence of outliers.

9. The fully automated flowerpot handling robot according to claim 1, characterized in that, The navigation system is configured to perform the following navigation implementation method: Use red boundary lines to enclose the work area into a rectangle, with one side of the rectangle being the area to be moved and the other side being the area where items have already been placed. After the robot picks up the flowerpot in the waiting area, it moves towards the boundary line; Upon reaching the boundary line, the vision processing flow is activated to detect the red boundary line in real time and generate tracking control commands, enabling the robot to move along the inside of the boundary line towards the already placed area. After the robot arrives at the designated area, it will perform the task of placing the flower pots. After placement, the robot returns to the transport area along the boundary line to begin the next transport cycle.

10. The fully automated flowerpot handling robot according to claim 1, characterized in that, The flowerpot placement system is configured to perform the following flowerpot placement and positioning method: After the robot arrives at the placement area, it uses a perception and positioning system to detect the positions of the flower pots already placed and generates a set of local flower pot positions. Principal component analysis was used to infer the orientation angle of the overall grid layout from the local flowerpot positions. ; According to the direction angle Construct a rotation matrix to convert the world coordinates of the placed flower pots into grid coordinates aligned with the grid, and calculate the discrete grid index corresponding to each flower pot; Construct and maintain an occupancy status matrix, and update the occupancy status in the matrix based on the detected flower pots and their grid indices; Based on the occupancy state matrix, the next vacant grid position is searched as the target placement position according to the row priority strategy; The grid index of the target placement position is transformed back to the coordinates in the world coordinate system through an inverse rotation transformation, which is then used as the target position for the robot to place the flower pot.