Soft Robotic Gripper Enables Gentle, Adaptive Fruit Harvesting
In an era where agricultural automation is rapidly evolving to meet global food demands, a team of researchers from Nanjing Forestry University has unveiled a breakthrough in soft robotics that could redefine how delicate fruits like apples and tomatoes are harvested. Their innovation—a pneumatically driven, 3D-printed soft gripper—combines material science, mechanical design, and intelligent control to achieve non-destructive, adaptive fruit picking, addressing long-standing limitations of traditional rigid robotic hands.
The new soft picking manipulator, detailed in a recent study published in the Journal of Forestry Engineering, leverages the inherent compliance of silicone-based actuators to safely interact with fragile produce. Unlike conventional end-effectors that rely on rigid links, motors, and precise positioning—often resulting in bruising or crushing—the soft gripper mimics the gentle, conformable grasp of a human hand. This advancement is particularly significant for high-value soft fruits, which are notoriously difficult to automate due to their variable shapes, delicate skins, and susceptibility to mechanical damage.
At the heart of the design are four identical soft fingers, each fabricated using a two-part molding process enabled by 3D-printed templates. The fingers consist of an active “strain layer” embedded with a series of interconnected pneumatic chambers, bonded to a passive “constraint layer” of uniform thickness. When positive pressure is applied, the strain layer expands more than the constraint layer, causing the finger to bend inward—ideal for enveloping and securing an object. Conversely, applying negative pressure (via a vacuum generator) induces outward bending, allowing the gripper to open wider than its neutral state to accommodate larger fruits or adjust its initial posture. This bidirectional actuation capability—rare in many soft grippers—greatly enhances versatility in real-world harvesting scenarios.
The research team, led by Zhu Yinlong and including Hua Chao, Chu Kaimei, and Chen Xin from the College of Electronic and Mechanical Engineering at Nanjing Forestry University, emphasized that the key to performance lay not just in the concept, but in meticulous structural optimization. Using the finite element analysis software ABAQUS, they systematically evaluated how four critical parameters—cavity wall thickness, number of air chambers, inter-chamber spacing, and constraint layer thickness—influenced bending behavior under pressures ranging from 0 to 60 kPa.
Their simulations revealed nuanced trade-offs. For instance, while thinner cavity walls (1.5 mm) yielded greater bending angles, they also led to excessive radial expansion and high stress concentrations at the bond interface, risking structural failure. Thicker walls (3.0 mm), on the other hand, overly restricted deformation. The optimal cavity wall thickness was determined to be 2.5 mm—a balance between flexibility and durability. Similarly, while the number of air chambers (tested at 8, 10, and 12) had minimal impact on overall bending, it offered a practical way to scale finger length for different fruit sizes. An inter-chamber gap of 3 mm proved ideal: smaller gaps caused interference between adjacent chambers at higher pressures, while larger gaps (5 mm) led to uncontrolled bulging and loss of actuation efficiency. The constraint layer also required careful tuning; too thin (1.5 mm), and it couldn’t contain radial expansion; too thick (3.0 mm), and it stifled bending. Again, 2.5 mm emerged as the sweet spot.
These simulation-driven insights directly informed the fabrication of the physical prototype. The team used Dragon Skin 20 silicone—a platinum-cure elastomer known for its high elongation and tear strength—mixed in a 1:1 ratio and cast into custom 3D-printed molds. After curing for five hours, the strain and constraint layers were bonded using silicone adhesive, and quick-connect pneumatic fittings were sealed into the base for reliable air delivery. The four fingers were then mounted onto a central flange, enabling straightforward integration with standard robotic arms.
To bring the gripper to life, the researchers developed a dual-mode pneumatic control system. At its core is an STM32 microcontroller interfacing with a LabVIEW-based graphical user interface on a host computer. This setup allows real-time monitoring and adjustment of air pressure. A proportional valve precisely regulates positive pressure from 0 to 60 kPa, while a vacuum generator provides the negative pressure needed for outward bending. Electromagnetic valves act as switches, toggling between inflation and evacuation modes. This architecture ensures both fine-grained control over grip strength and rapid reconfiguration of the gripper’s opening width.
Experimental validation confirmed the accuracy of the finite element models. Bending angle measurements across the 0–60 kPa range closely matched simulation predictions, with minor discrepancies at very low pressures (e.g., 10 kPa) attributed to initial air flow instability and slight material property variations. More critically, the team quantified the gripper’s load capacity by measuring the normal force exerted at the fingertip across different bending angles (0° to 90°) and pressures. As expected, force output decreased with increasing bend angle—peaking at the straight configuration and dropping to near zero at full curl. At maximum pressure (60 kPa), the fingertip could exert up to 1.36 N when unbent, but this fell to 0.55 N at 90°. When four fingers work in concert during enveloping grasps, the cumulative force enables stable handling of objects weighing up to approximately 590 grams—corresponding to a total grasping force of about 5.8 N.
This capability was demonstrated in practical harvesting trials. In “enveloping grasp” mode—where the fingers wrap fully around the fruit—the gripper successfully picked apples (200.6 g), tomatoes (282.9 g), pears (389.1 g), and even mangoes (580.3 g) without causing visible damage. The required actuation pressure scaled predictably with fruit mass, from 30 kPa for apples to 60 kPa for mangoes. The soft material conformed seamlessly to each fruit’s unique curvature, distributing contact forces evenly and eliminating stress concentrations that cause bruising.
The team also tested a “fingertip grasp” mode for smaller produce, using cherry tomatoes (15.3 g). At just 10 kPa, the gripper could lift the tiny fruit, proving its sensitivity to low-force tasks. However, this mode revealed a limitation: reduced contact area and lower structural rigidity made the grasp less stable, especially under external disturbances. The researchers noted that future iterations could incorporate variable-stiffness mechanisms—such as granular jamming or layer jamming—to enhance robustness during fingertip manipulation without sacrificing compliance.
The implications of this work extend beyond fruit harvesting. The design principles—bidirectional pneumatic actuation, simulation-guided optimization, and modular finger architecture—offer a scalable template for soft grippers in other delicate handling applications, from food processing and biomedical logistics to electronics assembly. Moreover, the use of accessible fabrication methods (3D printing and room-temperature silicone casting) keeps production costs low, a crucial factor for widespread adoption in agriculture.
Critically, this research addresses a core challenge in agricultural robotics: the mismatch between the rigidity of machines and the fragility of biological products. Traditional automation excels in structured, repetitive tasks with uniform objects—but farms are dynamic, unstructured environments filled with irregular, perishable items. Soft robotics bridges this gap by introducing mechanical intelligence through material compliance rather than complex sensing and control algorithms. The gripper doesn’t need to “know” the exact shape of a tomato; it simply adapts to it.
Looking ahead, the Nanjing Forestry University team plans to integrate vision systems for autonomous fruit detection and to explore hybrid actuation schemes that combine pneumatic softness with tendon-driven precision. They also aim to test the gripper in real orchard conditions, evaluating durability against dust, moisture, and temperature fluctuations.
As labor shortages intensify and consumer demand for blemish-free produce grows, the need for gentle, intelligent harvesting solutions has never been greater. This soft robotic gripper represents a significant step toward truly collaborative agricultural robots—machines that don’t just replace human hands, but emulate their dexterity, adaptability, and care.
By Hua Chao, Chu Kaimei, Chen Xin, and Zhu Yinlong, College of Electronic and Mechanical Engineering, Nanjing Forestry University. Published in the Journal of Forestry Engineering, 2021, Article Number: 2096-1359(2021)03-0127-06. DOI: 10.13360/j.issn.2096-1359.202006004.