A Modular Soft Gripper with Embedded Force Sensing and an Iris-Type Cutting Mechanism for Harvesting Medium-Sized Crops

Abstract

Agriculture is facing increasing challenges due to labor shortages, rising productivity demands, and the need to operate in unstructured environments. Robotics, particularly soft robotics, offers promising solutions for automating delicate tasks such as fruit harvesting. While numerous soft grippers have been proposed, most focus on grasping and lack the capability to detach fruits with rigid peduncles, which require cutting. This paper presents a novel modular hexagonal soft gripper that integrates soft pneumatic actuators, embedded mechano-optical force sensors for real-time contact monitoring, and a self-centering iris-type cutting mechanism. The entire system is 3D-printed, enabling low-cost fabrication and rapid customization. Experimental validation demonstrates successful harvesting of bell peppers and identifies cutting limitations in tougher crops such as aubergine, primarily due to material constraints in the actuation system. This dual-capability design contributes to the development of multifunctional robotic harvesters capable of adapting to a wide range of fruit types with minimal requirements for perception and mechanical reconfiguration.

1. Introduction

The application of robotics in the agricultural sector has been steadily increasing over the past decades [1,2]. This domain presents significant challenges for conventional, rigid robotics, including operation in unstructured environments and the need to interact safely with both crops and their surroundings, avoiding physical damage [3]. The incorporation of soft robotics components, which naturally conform to object surfaces due to the compliant materials they are made of, offers a promising alternative [4]. It enables firm yet gentle contact, making it suitable for automating tasks such as harvesting and fruit manipulation.

Soft robotics has been widely applied across diverse domains, including medicine, assistive technologies, and search and rescue operations [5,6,7,8,9,10,11]. In agriculture, soft grippers have demonstrated particular value due to their adaptability to irregular geometries and surface conditions. These systems rely on compliant materials and soft actuators, typically fluidic elastomer actuators (FEAs), to perform controlled deformations such as bending, stretching, and twisting. FEAs are widely used for their high deformability, ease of fabrication, and cost-effectiveness [12,13,14,15,16,17,18,19,20,21]. Other actuation methods include tendon-driven systems [22,23], hybrid FEA–tendon designs [24], and alternatives such as dielectric elastomer actuators and shape-memory materials. These technologies have collectively contributed to the emergence of scalable, versatile, and crop-safe robotic solutions.

Despite these advances, a notable gap remains in the development of soft grippers specifically designed for harvesting fruits that require mechanical detachment from the plant. While the current soft grippers provide gentle grasping capabilities, few integrate cutting mechanisms needed for tasks such as peduncle cutting. In contrast, several rigid robotic systems have incorporated such mechanisms [25,26,27], but often at the cost of increased complexity and the risk of damaging the fruit. Moreover, these solutions typically rely on perception systems for precise localization of the cutting point, which limits their robustness and adaptability in variable conditions. Therefore, there is a clear need for soft grippers that can simultaneously offer secure grasping, reliable cutting, and a reduced dependence on complex perception.

To contextualize this work within the current state of the art,

Table 1. Comparative overview of gripper systems for agricultural harvesting.

System Type	Ref.	Grasping	Soft Actuators	Sensorized Grasping	CuttingMechanism	Self-CenteringCutting Mechanism	Modular	3D-Printed
Soft	[13,30]	Yes	Yes	No	No	-	No	Yes
	[14,31]	Yes	Yes	No	No	-	Yes	No
	[32,33]	Yes	Yes	No	Yes	No	No	No
	This work	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Rigid	[27,34]	Passive	No	No	Yes	No	No	No
	[25,26]	Yes	Yes	No	Yes	No	No	No
	[35]	Yes	No	Yes	No	-	No	Yes

presents a comparative overview of representative soft and rigid gripper systems for agricultural harvesting. The comparison focuses on their actuation type, integration of cutting and sensing mechanisms, self-centering capabilities, modularity, and ease of fabrication via 3D printing. While some designs feature soft actuation or integrated cutting, very few combine all the key functionalities necessary for adaptive and selective fruit harvesting. In particular, similarly to recent developments in 3D-printed sensors, such as the mechano-optic sensor developed by [28] and the magnetic eFlesh sensor developed by [29], our approach emphasizes full functional integration in a low-cost, modular, and application-driven system. The proposed sensor is fabricated entirely using accessible FDM 3D printing with TPE and PLA on a modified Ender 3, offering greater economic and replicability advantages than the approach described in [28], which relies on the use of high-end multi-material printers and proprietary resins. Furthermore, our sensor employs a simplified mechano-optical readout system based on IR LEDs and photodiodes, avoiding the use of complex electronics like the Hall sensors required in [29]. Importantly, the mechanical simplicity and absence of rigid components like neodymium magnets reduce the risk of damaging delicate produce, enhancing safety and hygiene in agricultural environments. Unlike [28,29], which focus primarily on sensor characterization, our system demonstrates real-world applicability by integrating gripping, sensing, and self-centering cutting capabilities into a single 3D-printed end-effector, validated through experiments with bell peppers. This integration also reduces the reliance on computer vision by enabling geometry-independent, self-aligning cutting, thereby simplifying the robotic perception pipeline.

In response to these limitations, this article presents a novel modular hexagonal soft gripper design that enables a circular shell-type grasp. The gripper integrates sensorized soft actuators capable of measuring the contact force applied to the object during manipulation. The entire system is manufactured using 3D printing, allowing for rapid prototyping and structural customization. A key feature of the proposed design is the incorporation of an integrated cutting mechanism based on an iris-like structure. This mechanism enables the harvesting of fruits that require peduncle cutting due to the specific plant morphology. The combination of force-sensing soft actuators and a built-in cutting system provides a multifunctional solution intended for use in delicate agricultural tasks such as selective fruit harvesting.

Therefore, the primary contribution of this work is the integration of three key technologies into a single end-effector for agricultural use: (i) soft pneumatic actuators designed for delicate manipulation, (ii) embedded mechano-optical force sensors allowing for real-time grasp control, and (iii) an iris-type cutting mechanism capable of self-centering on the fruit peduncle. To the best of our knowledge, this is the first soft gripper to combine all three capabilities into a compact, fully 3D-printable design. Unlike prior cutting-enabled harvesters relying on rigid mechanics and perception-dependent slicing, the proposed system reduces the peduncle localization requirements by exploiting the geometry of the iris blades. Furthermore, the entire assembly is modular, scalable, and manufacturable without specialized equipment, providing a low-cost, scalable solution for adaptive harvesting across multiple fruit types.

The remainder of the article is organized as follows. Section 2 presents the design and fabrication of the proposed soft gripper, detailing both the soft actuators and the iris-like cutting mechanism, as well as providing dimensional drawings of the components that constitute the overall system. Section 3 details the process of characterizing the proposed embedded mechano-optical force sensors. Section 4 presents the key outcomes of the system’s experimental evaluation. Finally, Section 5 provides a summary of the main conclusions.

2. Soft Gripper

To determine the essential requirements for a fully operational soft gripper, a review of agricultural processes, particularly harvesting tasks, was conducted [4,36,37]. A key requirement for improving the profitability of harvesting machinery is the ability to handle various crop types. Consequently, we aimed for the design to be both adaptable and modular. The resulting module can be assembled in multiple configurations to produce a wide range of diameters and lengths, enabling the gripper to be reconfigured for different fruit types.

Another important requirement is simplicity, reflected in interchangeable and easy-to-repair components. To address this, the design relies on single, independent modules that are easy to manufacture. Additional requirements relate to fruit quality preservation, such as minimizing damage, using non-hazardous materials, and incorporating hygienic designs that inhibit the spread of pests and diseases. These aspects were often neglected in previous designs, which commonly featured damaging materials or complex geometries that hindered cleaning.

To address these limitations, the integration of soft robotic technologies with hygienic, compliant designs was prioritized to protect both the fruit and the crop. Other essential requirements include environmental sustainability, durability, and robustness for use in both agricultural and industrial applications. Accordingly, the material selection was carefully carried out to satisfy these criteria.

The modular soft gripper is intended to serve as the end-effector of a robotic manipulator [4,38] capable of executing the range of motions required for harvesting, commonly referred to as picking patterns [35,39,40,41]. These patterns typically involve combinations of basic movements such as twisting, pulling, lifting, and bending.

The following sections describe the selected soft material, its modeling for pneumatic actuation, and the integration of embedded mechano-optic sensors. The iris-type cutting mechanism is also detailed, followed by the manufacturing and assembly processes involved in the production of the soft gripper.

2.1. Rigid Structure Design

One of the primary challenges in agricultural automation is the development of a versatile gripper capable of harvesting a wide variety of fruits with minimal reconfiguration. This challenge becomes more pronounced in the context of soft robotics, particularly soft grippers, where predefining the gripper’s shape often constrains its adaptability. Some studies in the literature address the modularity and scalability of grippers. For example, the soft gripper proposed in [14] incorporates a modular design; however, it is specifically optimized for use with small fruits and edible fungi. Moreover, it employs simplified computational models and demonstrates reduced motion under an equivalent inlet pressure to that of the soft actuator proposed here.

This work aimed to further develop modular and scalable solutions for creating a quasi-universal gripper adaptable to different fruit types. To this end, a 3D-printed module was designed using polylactic acid (PLA), a well-known synthetic biodegradable polymer with favorable mechanical properties and low toxicity. Due to its relatively fast degradation, PLA presents a more environmentally sustainable alternative to conventional plastics [42]. Nevertheless, for long-term use where greater toughness is required, various strategies have been proposed to enhance PLA’s mechanical performance, including plasticization, copolymerization, and melt blending with tough polymers, rubbers, and thermoplastic elastomers [43,44,45,46,47,48,49]. Additionally, commercial PLA formulations with improved toughness for applications requiring durability are available, as outlined in [50].

The proposed PLA module offers several advantages. First, it functions independently of the gripper’s main structure, enabling adaptability to a broader range of tasks. Second, its modular and interchangeable design increases the system’s reliability, in contrast to many monolithic soft grippers that require complete replacement even for minor faults. Third, the module is fully replicable; both the rigid and soft parts can be easily fabricated via 3D printing using accessible, low-cost materials, without the need for post-processing.

As shown in Figure 1, the main components of the gripper are (i) a rigid structure composed of modules, (ii) three soft actuators with embedded mechano-optical sensors, and (iii) an iris-type cutting mechanism located at the top. The soft actuators in the gripper provide a range of motion allowing for a maximum opening of 61.4 mm, a minimum opening of 33.4 mm, and a resting opening diameter of 47.4 mm.

2.2. Soft Actuators

The proposed soft actuator belongs to the fluid elastomer actuator (FEA) category. These actuators operate using the pressure of a fluid: in this case, pressurized air. The choice of this type of actuator was due to its high force-to-weight ratio. Regarding its geometry, it features a single-channel diaphragm-type design, which allows for uniaxial movement. Another advantage of this type of soft actuator is its ease of control. This is due to the fact that soft diaphragm actuators are designed to produce motion primarily along a single axis, with the movements along other axes being negligible. As a result, the control of the gripper’s degrees of freedom (DoFs) is significantly simplified.

In addition, the actuator is equipped with an embedded mechano-optical force sensor, which enables real-time measurement of the contact force applied to the fruit, ensuring a firm grip without bruising it. The sensor system, integrated into the soft actuator, consists of a base housing infrared LEDs (emitter and receiver), connected to an Arduino Mega 2560, which provides power and performs the analog readout of the receiver signal. On top of this base, a compression layer with a gyroid-type geometry ensures a linear force response as a function of displacement [28,51]. Finally, a contact layer containing an occluder is placed above the compression stage, partially blocking the infrared beam as force is applied. A schematic of the actuator assembly can be seen in Figure 2.

2.3. Iris Cutting Mechanism

In order to design the cutting mechanism, the requirements needed to be adapted to the collection pattern required by the soft gripper. For the proposed soft gripper, which has a circular shell design, the collection pattern is based on an upward motion, where the fruit is inserted from the top of the gripper and held inside for subsequent manipulation by the soft actuators. Therefore, the cutting mechanism must be designed so that it does not interfere with this motion and ensures that the blades do not interact with the environment during the trajectories executed by the robotic manipulator.

Based on these considerations, an iris-type cutting mechanism was selected, where the blades remain concealed within the gripper, allowing for safe interaction with the environment and nearby operators. The blades are fabricated from 0.5 mm thick 304 stainless steel; see Figure 3. The mechanism includes three cutting blades, each with a machined edge and a weight of 12.6 g per blade.

In addition to the cutting blades, the mechanism is actuated by a motor with a gearbox and encoder, providing the necessary torque for cutting and allowing for real-time monitoring of the blade position. The specifications of these components are detailed in

Table 2. Characteristics of the iris-type cutting mechanism actuation system.

Faulhaber Motor 2342S024CR		Faulhaber Planetary Gearhead 26/1R		Faulhaber Encoder HEDS 5540
Characteristic	Value	Characteristic	Value	Characteristic	Value
Nominal voltage ($U_N$)	24 V	N° of gear stages	4	Resolution	1024/rev
Nominal speed ($n_N$)	5660 rpm	Continuous torque	7	Supply voltage	5 V
Nominal torque ($M_N$)	20.5 mNm	Efficiency, max.	69 %	Max. count frequency	100 kHz
Nominal current ($I_N$)	0.846 A	Reduction ratio	246:1
No-load speed ($n_0$)	93.2 mNm	Mass	162 g
Stall torque ($M_H$)	93.2 mNm
Torque constant ($k_M$)	28.6 mNm/A
Maximum efficiency (${\eta }_{max}$)	79 %
Mass	88 g

. The actuation system also includes two spur gears made of PLA. The first straight gear, with 69 teeth, guides the movement of the blades. The second straight gear, with 18 teeth, is directly coupled to the motor shaft and serves as the drive gear.

2.4. Manufacturing

Several materials, such as Ecoflex [52,53], Dragon Skin [20,54,55,56], Elastosil M4601 [53,57,58,59], and polydimethylsiloxane (PDMS), commercially known as Sylgard 184 [14,52,55,60,61,62], are commonly used for soft robotics. These materials are silicones, usually supplied as two-component systems, which are characterized by their elasticity, low stiffness, room-temperature curing, high availability, and relatively low cost. These properties make them attractive for use in soft robotics. However, the manufacturing method typically used for soft actuators employing these silicones, generally molding, presents a series of issues that hinder their implementation. Problems such as interstitial bubbles and delamination affect the actuator’s performance and reduce its service life [31,63]. Therefore, for the fabrication of the soft actuators that make up the proposed soft gripper, TPE was used. This material can be easily printed using a 3D printer, with minor extruder and temperature adjustments, allowing for the design of complex geometries and significantly reducing manufacturing problems as well as the production time.

The soft gripper was manufactured using a 3D printer, specifically a modified Ender 3. This modification involved the integration of a direct drive system, enabling the processing of flexible TPE filament, which poses challenges due to its hyperelastic behavior. The printing temperature was set to 225 °C to ensure adequate filament extrusion while avoiding hotend clogging or under-extrusion. Furthermore, the extrusion multiplier was fine-tuned through iterative adjustments to maintain a continuous filament flow, resulting in a uniform print quality for both the soft gripper and the integrated sensor.

To assess the reliability and repeatability of the manufacturing process, key parameters related to the printer precision and accuracy were analyzed. The Creality Ender 3 offers a positioning accuracy of ±0.1 mm on the X and Y axes and ±0.05 mm on the Z axis, with a layer resolution adjustable from 0.05 mm to 0.4 mm. The typical dimensional accuracy for printed components is approximately ±0.2 mm. In terms of repeatability, the printer exhibits variations of ±0.1 mm on the X/Y axes and ±0.05 mm on the Z axis, influenced by factors such as the belt tension and mechanical stability. These tolerances support consistent reproduction of printed parts, minimizing the variability in the sensor performance. The manufacturing parameters listed in

Table 3. Printing parameters for flexible TPE 70A and PLA filaments.

Parameter	TPE Value	PLA Value	Unit
Nozzle diameter	0.4	0.4	mm
Layer height	0.2	0.2	mm
Infill	100	20	%
Material temp	230	200	°C
Build plate temp	50	50	°C
Print speed	20	50	mm/s
Extrusion multiplier	115	100	%

were applied to ensure proper layer adhesion and print reproducibility.

A 15% gyroid infill pattern was selected during manufacturing (see Figure 4). This value was determined through iterative testing to develop a sensor capable of operating within a low-force range, suitable for handling delicate items such as small fruits and vegetables without causing damage. A gyroid infill pattern is commonly available in slicer software, including Ultimaker Cura. To ensure proper manufacturing, the external layers of the structure intended to be the interface needed to be removed.

To use the proposed hexagonal soft gripper on commercial cobots, an adapter made of rigid plastic, specifically PLA filament, was designed.

3. Characterization

The mechano-optical force sensor (Figure 5) was characterized using identical dimensions and a 15% gyroid infill pattern, consistent with the final gripper design. The experimental setup comprised a dial indicator to measure the deformation and an Imada SSII-20R dynamometer to apply controlled compression to the sensor’s contact surface (Figure 5a). Following calibration with the dynamometer, multiple loading cycles were conducted to evaluate the repeatability and hysteresis. The results indicated that the deviations remained within acceptable limits, confirming the reliability of the sensor’s output. The measured hysteresis was 6.48%, based on repeated loading and unloading at 15 mm min⁻¹, aligning with values reported by [28]. The maximum sample standard deviation was 0.05 for the voltage deformation measurements (Figure 5c) and 4.9 for the force deformation measurements (Figure 5d). It is worth noting that in Figure 5c, the sensor shows an insensitivity region at below approximately 0.8 mm of deformation. This behavior is attributed to the fabrication precision of the sensor. In the literature, such insensitivity at a low deformation is typically addressed by applying a preload, which helps to overcome insensitivity in this initial range and ensures immediate responsiveness to external forces. In our case, despite the initial insensitivity region, the sensor exhibited an overall sensitivity of −0.01 V/N, indicating its capability to detect small variations in the applied force beyond the initial threshold.

Additionally, due to the internal gyroid infill structure of the sensor, the applied pressure is distributed uniformly across the sensing region. This feature makes the sensor response independent of the shape of the manipulated object, allowing for reliable force measurement across a variety of surface geometries.

4. Experimental Results

Once the soft actuators with embedded mechano-optical sensors had been characterized, they were assembled onto the gripper and unloaded tests were conducted to verify its proper operation. Figure 6 shows the assembled gripper as well as the actuation of the cutting mechanism.

To evaluate the performance of the iris-type cutting mechanism, experiments were conducted with various bell peppers. The peppers had dimensions of approximately 150 mm in height, 42 mm in depth, and 48 mm in width, with an average weight of 220 g. The peduncle diameter ranged from 8 mm to 17 mm. In addition to these tests, and in order to validate the grasping capability of the soft actuators described previously, further experiments were conducted using objects with varying radii and geometries, including cylindrical and ellipsoidal models, specifically selected to cover the entire operational range of the gripper’s motion (opening of 33.4 mm to 61.4 mm). These tests demonstrated the gripper’s adaptability to diverse object sizes and shapes, ensuring reliable contact and grasp stability. Figure 7 shows the cutting experiments performed, along with the resulting condition of the peduncle after cutting.

During the cutting tests, the following observations were made. While cutting the pepper peduncle, even when the peduncle was not centered within the gripper, the design of the cutting mechanism allowed it to self-center, resulting in successful cuts in all the experimental trials. After the tests with the peppers, an attempt was made to cut aubergine peduncles without success (see Figure 8). These peduncles were partially dried, which increased the required cutting force, as freshly harvested aubergines were not available. In addition, the morphology of the aubergine peduncle presents significant mechanical resistance [37]. Since both the gears and the motor coupling were 3D-printed in PLA, the system was unable to complete the cut. As part of future work, these components will be machined in aluminum to enable further testing of the cutting mechanism.

Tests were also conducted to evaluate the harvesting sequence specific to the proposed gripper design (see Figure 9). The sequence involves an upward motion pattern in which the fruit enters through the top of the gripper and is guided through until it reaches the grasping point determined by the robot’s perception system. Once positioned, the iris-type cutting mechanism is activated to mechanically detach the fruit’s peduncle from the main stalk. These experiments were conducted under laboratory conditions.

With the aim of evaluating whether the embedded sensor affects the inherent contact properties of soft robotic systems, two types of soft actuators were tested under identical operating conditions: one equipped with the embedded mechano-optical sensor and one without it. In both cases, no visible damage was observed to the surface or structural integrity of the manipulated fruits. These results indicate that sensor integration does not compromise the gripper’s compliance or its gentle interaction with delicate agricultural products. The mechanical properties of the soft actuator are thus preserved, maintaining its ability to safely grasp produce without causing bruising. Moreover, the inclusion of the mechano-optical sensor provides valuable force estimation capabilities, enhancing the functionality of the system without reducing its core performance.

These results demonstrate the effectiveness of the soft gripper designed with an iris-type cutting mechanism, whose characteristics are summarized in

Table 4. Characterization of the proposed hexagonal soft gripper endowed with three soft actuators and an iris-type cutting mechanism.

Mass of a single module when fully mounted	69 · 10−3 kg
Max. displacement of the soft actuator (150 kPa)	0.010 m
Max. contact force (150 kPa)	54 N
Operating pressure range	0–150 kPa
Mass of a fully assembled hexagonal soft gripper	0.615 kg
Slip payload test (150 kPa)	3 kg
Max. opening	61.4 mm
Min. opening	33.4 mm
Mean grasp response time	$\approx 1$ s
Cutting force	10 N
Cutting time	$\approx 3$ s at 24 V

. This design shows the feasibility of integrating a cutting system into existing soft grippers, enabling the harvesting of various types of fruits, regardless of whether cutting is required. This results in a general-purpose gripper that not only firmly and gently grasps fruits for harvesting using different motion patterns but also has the advantage of incorporating cutting into these movements. Furthermore, due to the design of the cutting mechanism, the blades remain concealed when the gripper is its default or resting state, allowing safe handling both for the operator and during potential interactions with the environment, including different parts of the plant. Finally, thanks to the cutting mechanism design, the peduncle does not need to be positioned precisely for a successful cut, as is the case with other cutting grippers [26,27]; instead, the mechanism self-centers the peduncle for cutting. This reduces both the probability of failure and the complexity of the fruit perception system, requiring it only to detect the position and orientation of the fruit itself, without needing to detect the peduncle’s position.

All the experiments were conducted in a laboratory environment at a temperature of 25 °C and 20% humidity. The surface condition of the fruits was documented both before and after the experiments. No damage was observed following the experimental procedures. Additional observations were made one and two days after the experiments to confirm that no superficial damage had resulted from the manipulation of the various fruits by the gripper.

5. Conclusions

The automation of agricultural tasks currently faces major challenges in both technological and socio-economic terms, due not only to the complexity of unstructured environments but also to labor shortages and the need to improve productivity. The introduction of robotics into this sector, although it presents various challenges, can serve as an accelerator for its automation. This implementation will not start from scratch but rather result from the transfer of knowledge generated by what is known as Industry 4.0 into the agricultural sector, in what is increasingly referred to as Agriculture 4.0 or smart agriculture. Among the areas of knowledge with high potential for transfer between sectors is that of soft robotics, which facilitates the introduction of robotics due to soft robotics components’ adaptability to the environment and ability to interact safely with it.

In the literature, various approaches can be found, mostly aimed at designing soft grippers for harvesting. These end-effectors tend to use pneumatic actuation due to their force-to-weight ratio, although cable-driven and other types of actuation also exist. While these proposals focus on fruit harvesting, they are often tailored to fruits that, due to the morphology of the plant, are collected by pulling them or using a combined twisting and pulling motion. Although these patterns are widely used, there are other types of fruits where, due to a stronger physical attachment to the plant through a peduncle, such patterns cannot be applied without an additional cutting action.

To address this gap, a modular hexagonal soft gripper has been developed for the harvesting of medium-sized produce that does not grow in clusters and is not in contact with the ground, such as bell peppers, aubergines, or similar crops, where harvesting involves cutting the peduncle. The gripper consists of a minimum of three soft actuators, manufactured using TPE, which allow for not only a delicate and firm grasp but also the measurement of the contact force using mechano-optical sensors embedded within the soft actuators. Furthermore, to accommodate crops requiring a cutting motion, an iris-type cutting mechanism has been integrated into the gripper. This mechanism is capable of cutting with a force of 10 N in approximately 3 seconds. Moreover, the cutting mechanism does not require the peduncle to be precisely positioned for a successful cut; instead, the mechanism self-centers the peduncle for cutting. This reduces both the probability of failure and the complexity of the fruit perception system, requiring only the detection of the fruit’s position and orientation, without the need to locate the peduncle.

Overall, the proposed system combines safe, adaptive grasping with effective mechanical cutting in a fully 3D-printed, modular platform. It provides a foundation for developing multifunctional end-effectors that can be rapidly deployed across diverse crops with minimal adjustments. This approach contributes to the development of autonomous agricultural solutions that reduce the dependence on complex perception pipelines.

As part of future work, the soft gripper will be integrated into a robotic arm and tested in real agricultural environments to evaluate its performance on various target fruits, particularly peppers and eggplants, as well as other crops where the fruit does not grow in clusters.