Manual harvesting remains the predominant method for cherry tomatoes in many regions, accounting for 50% to 60% of total production costs. This approach is not only labor-intensive and inefficient but also frequently leads to fruit drop and mechanical damage, which directly impacts post-harvest storage, processing, and overall economic value. The development of mechanical harvesting solutions, particularly robotic systems, is therefore imperative. A critical component of such a system is the end effector, responsible for the delicate task of detaching and handling the fruit. This work focuses on the design and analysis of a gripping-type end effector for cluster harvesting, aiming to achieve non-destructive picking. The process begins with a thorough investigation of the physical and mechanical properties of cherry tomato clusters, proceeds through dynamic simulation to understand cluster behavior during harvesting motions, and culminates in the design and experimental validation of a novel end effector.
The foundational step for designing a non-destructive end effector is the accurate characterization of the cherry tomato cluster. A representative cluster consists of a main stem, secondary branches, peduncles with abscission zones (swollen nodes), and the fruit themselves. Key physical parameters were measured and are summarized below.
| Component | Physical Property | Average Value |
|---|---|---|
| Main Stem | Length | 170.5 mm |
| Base Diameter | 3.3 mm | |
| Tip Diameter | 1.9 mm | |
| Fruit | Mass | 11.7 g |
| Fruit | Radial Diameter | 30.5 mm |
The mechanical properties are crucial for simulating interaction forces. Compression and tensile tests were performed, yielding the following key parameters essential for modeling.
| Component | Mechanical Property | Value / Range |
|---|---|---|
| Main Stem | Longitudinal Compressive Strength | 4.4 MPa |
| Longitudinal Elastic Modulus | 22.48 MPa | |
| Abscission Zone (Swollen Node) | Tensile Failure Force | 5 – 8 N |
| Bending Failure Moment | 7 – 16 N·mm | |
| Fruit | Axial Rupture Force | 15 – 35 N |
| Radial Rupture Force | 10 – 20 N |
To accurately simulate the dynamic response of a tomato cluster during the rapid motions imposed by an end effector, a lumped-parameter mechanical model was developed. The cluster is conceptualized as a “Flexible Rod – Hinge – Rigid Rod – Link Point – Spherical Body” system. The main stem, due to its length and slenderness, is modeled as a flexible beam. The secondary branches are treated as short rigid rods connected to the main stem via hinge joints, which represent the branching nodes. The fruit, modeled as spheres, are connected to the secondary branches via a “link point” that embodies the mechanical properties of the peduncle’s abscission zone. This model allows for the simulation of complex, coupled oscillations.
The governing equation for the small-amplitude bending vibration of the main stem (flexible rod) can be approximated using Euler-Bernoulli beam theory:
$$ EI \frac{\partial^4 w(x,t)}{\partial x^4} + \rho A \frac{\partial^2 w(x,t)}{\partial t^2} = f(x,t) $$
where $w(x,t)$ is the lateral deflection, $E$ is the elastic modulus, $I$ is the area moment of inertia, $\rho$ is density, $A$ is the cross-sectional area, and $f(x,t)$ represents distributed forces from the branches and fruit.
The torque at a branching hinge is modeled with a rotational damper, whose stiffness $k_\theta$ is derived from experimental moment-angle data:
$$ \tau = k_\theta \cdot \theta + c_\theta \cdot \dot{\theta} $$
where $\tau$ is the torque, $\theta$ is the angular displacement, and $c_\theta$ is the damping coefficient.
The primary failure modes for fruit detachment are tensile failure at the abscission zone and fruit-to-fruit collision damage. The condition for fruit detachment can be expressed as:
$$ F_{\text{tension}} \geq F_{\text{failure}} \quad \text{or} \quad M_{\text{bending}} \geq M_{\text{failure}} $$
where $F_{\text{failure}}$ and $M_{\text{failure}}$ are the tensile and bending failure limits of the abscission zone, respectively.
The proposed end effector design comprises two main subsystems: a clamping mechanism and an anti-sway mechanism. The clamping mechanism is responsible for gripping the main stem’s base and severing it. It consists of two opposing gripper arms, each equipped with a rubber lining to distribute pressure and a integrated cutting blade. The anti-sway mechanism features contoured plates that can be rotated to gently cradle the fruit cluster from the sides. These plates are also lined with soft rubber. The operational sequence is as follows: the robotic arm positions the end effector around the target cluster; the anti-sway plates close to stabilize the cluster and dampen oscillations; the gripper arms close simultaneously to clamp the stem and cut it; finally, the cluster is transported and released into a collection bin.
Prior to physical prototyping, extensive multibody dynamics simulations were conducted to analyze the interaction between the end effector and the tomato cluster. The validated “Flexible Rod – Hinge…” model was used. Simulations investigated two critical phases: impact during the initial capture and vibration during accelerated transport.
Collision simulations were performed by moving the clamped cluster towards rigid walls from different directions (front, back, right). The maximum tensile force $F_{\text{max}}$ at the abscission zones was recorded. The results indicated that rearward collisions were most severe, generating forces that exceeded the 5 N failure threshold. A key improvement was simulated by adding a rubber layer (Elastic Modulus = 6 MPa) to the contact surface. This layer significantly reduced peak forces, as shown by the modified contact force model:
$$ F_{\text{contact}} = K \delta^n + D \dot{\delta} $$
where $K$ is the stiffness (much lower for rubber), $\delta$ is penetration depth, $n$ is a nonlinear exponent, and $D$ is a damping coefficient. The rubber’s high damping effectively dissipates impact energy.
The dynamic transport simulation involved applying specific accelerations to the clamped cluster in five directions: right, forward, backward, up, and down. The primary metrics were inter-fruit collision force and abscission zone tensile force. The maximum safe acceleration for each direction was determined as the highest acceleration that did not cause simulated forces to exceed the failure limits.
| Motion Direction | Maximum Safe Acceleration (m/s²) | Critical Limiting Factor |
|---|---|---|
| Right | 8 | Inter-fruit collision |
| Forward | 8 | Inter-fruit collision |
| Backward | 6 | Abscission zone tension |
| Upward | 8 | Inter-fruit collision |
| Downward | 8 | Inter-fruit collision |
The simulation results provided critical guidance for the trajectory planning of the robotic arm manipulating the end effector. By limiting the acceleration, particularly in the backward direction, the mechanical insult to the cluster can be minimized.
A prototype of the end effector was fabricated and tested. The performance metric was the fruit loss rate (damaged or dropped fruit). Tests were conducted for each primary movement direction using the maximum safe accelerations derived from simulation.
| Motion Direction (Acceleration) | Total Fruit | Fruit Loss | Loss Rate (%) |
|---|---|---|---|
| Backward (6 m/s²) | 261 | 22 | 8.43 |
| Right (8 m/s²) | 252 | 17 | 6.75 |
| Forward (8 m/s²) | 263 | 24 | 9.13 |
| Upward (8 m/s²) | 251 | 19 | 7.57 |
| Downward (8 m/s²) | 259 | 25 | 9.65 |
The overall average loss rate was 8.3%, demonstrating a significant improvement over the high damage rates typical of purely rigid grasping mechanisms. The anti-sway mechanism was particularly effective in dampening oscillations during transport, preventing the large pendulum-like swings that lead to high inertial forces and fruit-to-fruit collisions. The rubber linings on both the gripper and anti-sway plates were essential for force distribution and energy absorption.
This study successfully demonstrates a model-driven approach to designing a non-destructive end effector for cherry tomato cluster harvesting. The key conclusions are:
- The proposed “Flexible Rod – Hinge – Rigid Rod – Link Point – Spherical Body” mechanical model, when parameterized with accurate physical data, provides a valid simulation platform for predicting cluster dynamics under manipulator-induced motions.
- Dynamic simulation is a powerful tool for identifying critical failure modes (abscission zone tension and inter-fruit collision) and for optimizing the operation parameters of the end effector, such as maximum safe accelerations.
- The integration of compliant materials (rubber) into the contact surfaces of the end effector is a simple yet highly effective strategy for mitigating impact and clamping damage.
- The final two-fingered gripping end effector design, complemented by an active anti-sway mechanism, achieved a satisfactory average fruit loss rate of 8.3% in prototype testing, confirming the value of the simulation-based design process.
The performance of the end effector is intrinsically linked to the motion profile of the robotic arm that carries it. Future work will involve the co-simulation and co-optimization of the end effector design with the path planning and control algorithms of the complete harvesting robot to further improve efficiency and reliability.