As a researcher in agricultural robotics, I have long been fascinated by the challenges of automating delicate harvesting tasks, particularly in tea cultivation. The tea industry, vital to many economies, faces escalating labor shortages and rising costs, driving an urgent need for robotic solutions. Central to any tea-picking robot is the end effector—the component that directly interacts with the crop. An effective end effector must not only sever the tender bud cleanly but also handle it gently to preserve quality and enable continuous operation. In this work, I present the development of a novel cutting-collecting integrated end effector designed specifically for tea harvesting. This end effector embodies simplicity and efficiency, employing a single-degree-of-freedom planar linkage to perform both cutting and collecting actions in a seamless cycle. The design process involved meticulous characterization of tea bud properties, rigorous static and dynamic analysis, and extensive field validation. This article details every step, underscoring how a focused approach to end effector design can address pressing agricultural automation needs.
The impetus for this project stems from the limitations of existing tea harvesting methods. Manual picking is slow and labor-intensive, while mechanical harvesters often lack selectivity, damaging plants and reducing yield quality. A robotic end effector must bridge this gap, combining precision with robustness. My goal was to create an end effector that mimics the swift “pick-and-place” motion of a human harvester but with mechanical reliability. The key innovation lies in integrating the cutting and collection functions into one compact mechanism, eliminating the need for separate actuators or complex material-handling systems. This integration is crucial for minimizing weight and power consumption on a mobile robotic platform. Throughout this article, the term ‘end effector’ will be used repeatedly to emphasize its pivotal role; every design decision revolves around optimizing this critical interface between machine and crop.
Before embarking on the mechanical design, it was essential to understand the physical and mechanical properties of the target: the tea bud. I selected ‘Yinghong No. 9’, a prevalent variety in Southern China, for this study. A sample of 40 fresh buds, each comprising one apical bud and one young leaf (the ‘one-bud-one-leaf’ standard for premium tea), was carefully collected. Their morphological dimensions were measured using digital calipers, with a focus on parameters critical for end effector design: overall size for clearance and stem diameter for cutting force calculation. The results, summarized in Table 1, provided the foundational data for sizing the end effector’s working envelope and its cutting mechanism.
| Parameter | Minimum (mm) | Maximum (mm) | Mean ± Standard Deviation (mm) |
|---|---|---|---|
| Bud Height | 20.0 | 35.0 | 27.5 ± 4.1 |
| Bud Width | 15.0 | 25.0 | 20.0 ± 2.9 |
| Stem Diameter | 1.77 | 2.99 | 2.31 ± 0.32 |
With the dimensional envelope defined, the next step was to quantify the force required to sever the stem—a direct input for the end effector’s actuator sizing. Using a texture analyzer equipped with a flat-blade shear probe, I conducted shear tests on the excised stems. The force-deformation curves exhibited a characteristic pattern: an initial non-linear rise as the blade compressed and sheared the fibrous material, culminating in a peak force at fracture. Analyzing all samples yielded the shear force data in Table 2. The maximum recorded force of 6.4 N became the design benchmark for the end effector’s cutting mechanism. To generalize this relationship, I performed a power-law regression, correlating shear force \( F_s \) with stem diameter \( d \). The derived empirical model is:
$$ F_s = 2.15 \cdot d^{1.48} $$
with a coefficient of determination \( R^2 = 0.89 \). This equation, while specific to our sample set, offers a predictive tool for estimating cutting force requirements across different tea cultivars or growth conditions, ensuring the end effector possesses a sufficient force margin.
| Statistical Measure | Shear Force (N) | Stem Diameter at Shear Point (mm) |
|---|---|---|
| Maximum | 6.40 | 2.99 |
| Minimum | 2.90 | 1.77 |
| Mean | 4.52 | 2.31 |
| Standard Deviation | 0.97 | 0.32 |
Armed with this biological data, I proceeded to the core of the project: the mechanical design of the end effector. The primary objectives were simplicity, reliability, and functional integration. I conceived a mechanism based on a planar four-bar linkage, a classic solution for converting rotary input into a complex output path with just one actuator. The specific configuration is a crank-rocker mechanism. The physical embodiment consists of a servo motor (the prime mover), a driving crank, a connecting rod, a blade plate (the rocker), and a blade holder assembly that incorporates a fixed blade and a collection box. This integrated end effector is designed to execute a sequence of motions: enveloping the stem, shearing it, dynamically ejecting the freed bud, and finally resetting—all within one continuous rotation of the servo motor.
The operational cycle of this novel end effector can be dissected into distinct phases. In the initial ‘approach and position’ phase, the robotic arm maneuvers the end effector so that the target tea stem is located within the V-shaped opening formed by the blade plate and the blade holder. The end effector then moves downward, gently surrounding the stem. The ‘cutting phase’ begins as the servo motor rotates the crank. This rotation is transmitted through the connecting rod, causing the blade plate to swing inward. The lower edge of the blade plate sweeps across the fixed blade mounted on the holder, creating a scissor-like shearing action that cleanly cuts the stem. Immediately following the cut, the ‘throwing and collection phase’ commences. The blade plate, now carrying significant kinetic energy, continues its swing and makes contact with a protrusion on the blade holder. This contact impulsively drives the entire blade holder assembly to rotate about its pivot. This sudden rotational motion flings the freshly cut tea bud off the blade plate and into a waiting collection box attached to the holder. Finally, in the ‘reset phase’, the servo motor reverses direction, returning the blade plate and holder to their initial positions, ready for the next picking cycle. This elegant sequence allows the end effector to harvest and collect without requiring the robot to reposition for disposal, dramatically increasing potential picking rates.
To ensure the end effector can generate the necessary cutting force, a detailed static force analysis of the linkage at the moment of shear is imperative. Consider the linkage as a system of rigid bodies. The servo motor applies a torque \( M \) at the crank axis. Let \( l_1, l_2, l_3, l_4 \) represent the lengths of the crank (AB), connecting rod (BC), blade plate segment from pivot to connection point (CD), and the effective lever arm on the blade plate for cutting (DE), respectively. The fixed frame distance is AD. At the instant of cutting, the angles between the connecting rod and the velocity vectors at B and C are \( \beta \) and \( \alpha \), as determined from the configuration. The force transmission from motor torque to cutting force \( F_c \) at the blade interface follows this logical chain:
The tangential force at point B due to the motor torque is:
$$ F_B = \frac{M}{l_1} $$
The component of this force acting along the axis of the connecting rod (assumed to be a two-force member in static analysis) is:
$$ F_{rod} = F_B \cdot \cos(\beta) = \frac{M \cos(\beta)}{l_1} $$
This force is transmitted undiminished to point C on the blade plate. The component of \( F_{rod} \) that contributes to the torque about the blade plate’s pivot D is the component perpendicular to CD, which is \( F_{rod} \cos(\alpha) \). Therefore, the torque acting on the blade plate about pivot D is:
$$ \tau_{plate} = F_{rod} \cos(\alpha) \cdot l_3 = \frac{M \cos(\beta) \cos(\alpha) l_3}{l_1} $$
This torque is balanced by the moment generated by the cutting force \( F_c \) acting at the blade tip with a lever arm \( l_4 \):
$$ \tau_{plate} = F_c \cdot l_4 $$
Equating the two expressions yields the fundamental design equation for the cutting capability of the end effector:
$$ F_c = \frac{M \cos(\beta) \cos(\alpha) l_3}{l_1 l_4} $$
Substituting the nominal design values—\( M = 0.4 \, \text{N·m} \) (40 N·cm), \( l_1 = 0.018 \, \text{m} \), \( l_3 = 0.024 \, \text{m} \), \( l_4 = 0.079 \, \text{m} \), \( \beta = 12.7^\circ \), \( \alpha = 8.96^\circ \)—we compute:
$$ F_c = \frac{0.4 \cdot \cos(12.7^\circ) \cdot \cos(8.96^\circ) \cdot 0.024}{0.018 \cdot 0.079} $$
Evaluating the trigonometric terms: \( \cos(12.7^\circ) \approx 0.975 \), \( \cos(8.96^\circ) \approx 0.988 \). Thus,
$$ F_c \approx \frac{0.4 \cdot 0.975 \cdot 0.988 \cdot 0.024}{0.001422} \approx \frac{0.00925}{0.001422} \approx 6.51 \, \text{N} $$
This calculated force of 6.51 N exceeds the maximum required shear force of 6.4 N, providing a small but adequate safety margin. This static analysis confirms that the end effector’s mechanical advantage, derived from its specific linkage geometry, is sufficient for its primary task. It is a critical verification step before committing to fabrication.
However, static analysis alone cannot capture the dynamic performance crucial for a fast-cycling end effector. To evaluate speed, inertia effects, and potential vibrations, I conducted a kinematic and dynamic simulation using SolidWorks Motion. The servo motor was modeled with a constant angular velocity profile of 310 °/s over its 155° working rotation. The simulation tracked the angular displacement, velocity, and acceleration of both the blade plate and the blade holder—the two most massive moving components. The results, particularly for angular velocity and acceleration, are highly informative and are consolidated in Table 3.
| Component | Parameter | Minimum Value | Maximum Value | Unit |
|---|---|---|---|---|
| Blade Plate | Angular Velocity | -438 | 487 | °/s |
| Angular Acceleration | -3373 | 3179 | °/s² | |
| Blade Holder | Angular Velocity | -437 | 487 | °/s |
| Angular Acceleration | -4775 | 2382 | °/s² |
The simulation reveals that the complete cutting-throwing cycle can be executed in approximately 0.5 seconds of active motor rotation, leading to a total cycle time near 1 second when including a brief dwell. This promises high operational efficiency for the end effector. However, the extremely high angular accelerations, especially peak magnitudes exceeding 4000 °/s², indicate significant inertial forces. These dynamic loads manifest as shaking and impose high stress on joint bearings and links. This insight is vital; it dictates that the end effector’s design must prioritize rigid connections and possibly incorporate lightweight materials or counterbalance strategies to mitigate vibration, ensuring longevity and consistent accuracy.
The translation from simulation to real-world performance is always the ultimate test. I conducted field trials in a tea garden using a functional prototype of the end effector. The prototype was mounted on a manually positioned test rig to isolate the performance of the end effector itself from the complexities of full robot navigation and vision. The target was consistently the one-bud-one-leaf shoots on Yinghong No. 9 bushes. Three experimental runs were performed, each comprising 20 separate picking attempts. Success was strictly defined as the clean severing of the target bud and its successful deposition into the collection box, without damage to adjacent buds. The time for each successful pick, from the initiation of motor movement to the return to the start position, was recorded. The aggregated results are presented in Table 4.
| Trial Group | Number of Attempts | Successful Picks | Success Rate (%) | Mean Picking Time (s) | Time Standard Deviation (s) |
|---|---|---|---|---|---|
| 1 | 20 | 15 | 75.0 | 1.52 | 0.21 |
| 2 | 20 | 14 | 70.0 | 1.58 | 0.25 |
| 3 | 20 | 16 | 80.0 | 1.45 | 0.18 |
| Overall | 60 | 45 | 75.0 | 1.50 | 0.22 |
The field data demonstrates the practical viability of the end effector. An average success rate of 75% and an average cycle time of 1.5 seconds are promising for an initial prototype. Observing the failures provided invaluable diagnostic information. The primary failure modes were: (1) Clogging: Densely packed buds sometimes caused the end effector’s opening to engage multiple stems, jamming the mechanism. (2) Interference: Protruding screw heads inside the blade holder assembly occasionally snagged the bud during the throwing motion, deflecting it away from the collection box. (3) Incomplete Throws: Some buds, due to their orientation or slight adhesion to the blade, were not projected with enough energy to clear the mechanism and fall into the box. These observations directly inform the next iteration of the end effector’s design, pointing toward a narrower, more streamlined cutting profile and a cleaner internal path for bud ejection.
Beyond the basic performance metrics, a deeper analysis of the end effector’s kinematics can be formalized using loop-closure equations. Defining a coordinate system with origin at the fixed pivot A, the positions of key points can be expressed as functions of the crank angle \( \theta \). For points B and C:
$$ \vec{r_B} = l_1 (\cos\theta \, \hat{i} + \sin\theta \, \hat{j}) $$
$$ \vec{r_C} = \vec{r_D} + l_3 (\cos\phi \, \hat{i} + \sin\phi \, \hat{j}) $$
where \( \vec{r_D} \) is the fixed position of the blade plate pivot, and \( \phi \) is its angle. The constraint imposed by the connecting rod of length \( l_2 \) gives the scalar equation:
$$ (\vec{r_C} – \vec{r_B}) \cdot (\vec{r_C} – \vec{r_B}) = l_2^2 $$
This equation can be solved numerically to find \( \phi(\theta) \), the motion transfer function. The angular velocity \( \omega_{plate} \) of the blade plate is then \( d\phi/dt \), and it relates to the crank angular velocity \( \omega_{crank} = d\theta/dt \) through the derivative:
$$ \omega_{plate} = \frac{d\phi}{d\theta} \cdot \omega_{crank} $$
The derivative \( d\phi/d\theta \), known as the velocity coefficient, is a function of the instantaneous configuration. For the designed linkage, this coefficient varies throughout the cycle, explaining the non-uniform angular velocities observed in simulation. Maximizing this coefficient during the cutting phase can enhance speed, but it also increases inertial loads—a classic engineering trade-off central to optimizing this end effector.
The dynamic behavior can be further modeled using the Lagrangian approach to account for the distributed masses and inertias. Defining the kinetic energy \( T \) of the system as the sum of energies for the crank, connecting rod, blade plate, and blade holder, and assuming gravitational potential energy \( V \) is negligible for this high-speed operation, the equations of motion take the form:
$$ \frac{d}{dt}\left( \frac{\partial T}{\partial \dot{\theta}} \right) – \frac{\partial T}{\partial \theta} = Q_\theta $$
where \( Q_\theta \) is the generalized force (the motor torque \( M \)). Solving this equation, even numerically, would provide precise torque requirements to achieve the desired motion profile, informing motor selection and control strategy. For instance, the peak torque demand likely occurs during the rapid acceleration phase of the blade holder, which aligns with the high angular acceleration peaks from the simulation. This level of analysis is a natural next step for refining the dynamic performance of the end effector.
Material selection plays a subtle yet significant role in the end effector’s performance. The primary structure, including the links and holder, was fabricated from 6061 aluminum alloy, offering an excellent strength-to-weight ratio crucial for minimizing inertia. The cutting blade was made from hardened spring steel (AISI 1095) to retain a sharp edge. The choice of materials directly impacts the dynamic loads; a heavier end effector would require a more powerful motor and generate greater reaction forces on the robotic arm. Future iterations could explore carbon fiber composites for the larger structural elements like the blade holder to reduce mass and the associated dynamic stresses, thereby improving the end effector’s speed and energy efficiency.
The control philosophy for this end effector is intentionally simple: open-loop position control of the servo motor. A pre-programmed sequence of angles defines the complete pick-and-reset cycle. This simplicity reduces cost and computational burden. However, integrating basic sensors could enhance reliability. For example, a current sensor on the servo could detect stall conditions indicative of jamming, triggering an automatic reversal and clear sequence. Similarly, a simple optical sensor at the collection box entrance could confirm successful bud entry, providing feedback for the robotic system to log successful picks or retry failed ones. These are pragmatic upgrades that maintain the end effector’s simplicity while adding a layer of robustness.
To consolidate the technical specifications and performance targets, a comprehensive parameter summary for the end effector is provided in Table 5. This table serves as a concise reference for the key design and operational metrics that define this cutting-collecting integrated end effector.
| Category | Parameter | Symbol | Value/Description | Unit |
|---|---|---|---|---|
| Geometric | Crank Length | \( l_1 \) | 18.0 | mm |
| Connecting Rod Length | \( l_2 \) | 50.0 | mm | |
| Blade Plate Connection Arm | \( l_3 \) | 24.0 | mm | |
| Cutting Force Lever Arm | \( l_4 \) | 79.0 | mm | |
| Fixed Frame Distance | \( l_{AD} \) | 58.0 | mm | |
| Actuation & Force | Servo Motor Rated Torque | \( M \) | 40 | N·cm |
| Servo Motor Operating Speed | \( \omega_{motor} \) | 310 | °/s | |
| Theoretical Cutting Force (Static) | \( F_c \) | 6.51 | N | |
| Required Max. Biological Shear Force | \( F_{s, max} \) | 6.40 | N | |
| Performance | Simulated Cycle Time (Active) | \( t_{cycle} \) | ~1.0 | s |
| Field Average Success Rate | – | 75 | % | |
| Field Average Picking Time | – | 1.50 | s |
Reflecting on the development journey, this end effector project highlights several overarching principles in agricultural robotic design. First, thorough biological characterization is non-negotiable; the end effector must be tailored to its biological counterpart. Second, mechanism elegance often trumps complexity; a single-degree-of-freedom design proved capable of a sophisticated pick-and-place sequence. Third, integration of multiple functions (cutting and collecting) into one mechanism is a powerful strategy for streamlining robotic workcells. While the current prototype shows promise, the path forward is clear. Immediate improvements involve refining the geometry to mitigate clogging, smoothing internal surfaces to ensure reliable bud ejection, and reinforcing joints to withstand the identified dynamic loads. Longer-term, integrating this end effector with a machine vision system for autonomous bud detection and stem localization will transition it from a capable mechanism to a full subsystem of an intelligent harvesting robot.
The implications of this work extend beyond tea. The fundamental design paradigm—a linkage-based, single-actuator, combined cutting and ejecting end effector—is adaptable to other delicate harvesting tasks for crops like strawberries, grapes, or certain flowers. The methodology of combining biological study, static and dynamic analysis, simulation, and iterative field testing provides a robust template for developing effective agricultural end effectors. As robotics permeates agriculture, such focused, practical innovations in end effector technology will be instrumental in overcoming the economic and labor challenges facing the industry, paving the way for sustainable and productive farming futures.