Design and Test of a Force Feedback Seedling Pick-Up Gripper for an Automatic Transplanter

Abstract

Aiming at the problems of seedling injury and planting leakage due to the lack of seeding clamping force detection and real-time control in vegetable transplanting, a force feedback gripper was developed based on the linear Hall element. The mechanical properties of the stem of pepper cavity seedlings were first analyzed to provide a basis for the design of the gripper. A linear Hall sensor, a magnet, an elastic actuator, and an Arduino Uno development board make up the grasping force detecting system. Upon picking up a seedling, the elastic actuator, which is connected to the magnet, bends like a cantilever beam. As a result of the micro-displacement created by the elastic actuator, the Hall sensor’s voltage changes and can be used to determine the clamping force. Detection avoids direct contact between the sensor and the cavity seedlings, reducing the risk of sensor damage. Finite element method (FEM) simulations were used to determine the initial spacing between the magnet and Hall sensor and the effect of the elastic actuator. Control commands are sent to the servo based on the gripping force collected by the Arduino Uno board. Finally, the functions of accurate measurement, display, storage, and control of the clamping force of the cavity tray seedlings are realized, so that the damage rate of the cavity tray seedlings is reduced. In order to explore the influence of the elastic actuators on the clamping force detection system and the performance of the force feedback gripper, a calibration test of the clamping force detection system and a test of the indoor transplantation of pepper seedlings were carried out. Based on the calibration test, the clamping force detection system has a sensitivity of 0.0693 V/N, linearity of 3.21%, an average linear coefficient of determination of 0.986, and a range of 10 N, which fully meet the clamping force detection accuracy requirements during transplantation. Indoor tests showed that the force feedback gripper was stable and adaptable. This study can provide a reference for detecting and controlling clamping forces during transplantation.

1. Introduction

Transplanting plug seedlings can increase vegetable yields and has many benefits [1,2]. Fully automatic mechanized vegetable transplanting has become a widely adopted method of increasing yields and reducing costs [3,4]. It should be noted, however, that different growth conditions, different varieties of cavity seedlings, and the stability of the transplanter can cause the transplanting gripper to miss or damage the seedlings during extraction [5,6,7]. It is necessary to develop a transplanting gripper that can detect and control the magnitude of the gripping force in real time in order to solve the above problems. Research on the detection and control of clamping force in agricultural production is primarily focused on the harvesting of fruits and vegetables [8,9,10]. Force sensors are placed on the gripper surface to detect force and control gripping posture in order to achieve low loss picking [11]. An alternative method is to place the sensor inside the flexible body and measure the clamping force by measuring the deformation of the flexible body [12]. It may be possible to draw inspiration from the gripping force detection and control method used in the fruit and vegetable harvesting process. However, the sizes of those working objects are considerably larger than that of plug seedlings and the planting process is more time-consuming [13,14,15].

Over the years, many studies have been conducted on the clamping process of cavity tray seedling transplants. As early as 1990, Ting et al. incorporated capacitive sensors into the slide pins to detect the clamping force [16,17]. However, the detection accuracy was low, and the efficiency of the transplant was affected by the seedling heap. Various types of transplanting grippers have been developed, and experiments have been conducted by scholars at home and abroad in line with the development of transplanting technology. Using tension and pressure sensors on the end effector, Jiang et al. analyzed the force relationship between vegetable rootstocks and seedling lumps by adjusting the pin depth to suit different cavity trays [18]. The vertical drive mechanism used by Li et al. enabled the simultaneous insertion and retraction of the clamping jaws, which led to increased clamping friction and improved clamping stability [19]. L. Han et al. [20] developed a flexible multi-pin seedling gripping jaw that can effectively grip the roots of potted seedlings, while sensors detect gripping, holding, and the releasing movement to ensure that the grip is stable. Han et al. [21] used adjustable pin end-effectors to deal with a variety of cavity pot seedling extraction problems, with 97.93% successful extraction and 99.19% root system integrity. Liu et al. [22] observed potted seedlings in the clamped state using X-rays and analyzed the causes of seedling pile breakage during transplant clamping using image recognition. In a recent study, Ji et al. [23] integrated a PVDF piezoelectric film on the inner surface of the seedling picking needle to detect the seedling clamping force in real time. Additionally, some scholars have designed and optimized seedling picking mechanisms to improve transplant quality [24,25,26].

Automatic transplanters have greatly benefited from the above research in optimizing the process of picking and dropping seedlings. It is worth noting that, at present, these devices primarily focus on detecting the process of clamping action, and do not involve a load change in clamping force during the process of taking and dropping seedlings. After reading the information on the clamping force, devices that detect seedling clamping force do not take further action. Applications of clamping force feedback control equipment are mainly focused on fruit and vegetable harvesting. In the vegetable transplanting process, clamping force feedback control equipment needs to be studied urgently.

To address the problems caused by the inability to detect and control clamping force during vegetable transplanting, we developed a linear Hall element-based micro-displacement detection device. By measuring the deflection of the elastic actuator, we were able to determine the clamping force magnitude, determine the sensor measurement range using COMSL Multiphysics field simulation software, calibrate the detection device, and then control the clamping force magnitude. A force-feedback seedling gripper was designed for the fully automatic transplanter, and transplanting gripping tests were conducted on cavity tray seedlings to assess stability and adaptability.

2. Materials and Methods

2.1. General Structure and Principle

2.1.1. Analysis of the Work Object

Reference was made to research on and test methods for the mechanical properties of plant stems so that key components of the transplanting gripper could be designed [27]. Using Chiyan No. 2 pepper cavity tray seedlings as the test material, 72-hole cavity trays were used to grow seedlings for mechanical characterization. The cavity tray seedlings were grown on a substrate containing peat, perlite, and vermiculite in a 2:1:1 ratio, and they were 45 days old. This study used 80 randomly selected pepper seedlings as test samples, and the stem diameters of all seedlings were measured. Since the mechanical property test is irreversible, the above-mentioned seedlings were divided equally for the tensile and compressive strength tests. Peppers were tested mechanically using the TA touch texture analyzer (Shanghai Bosin Industrial Development Co., Ltd, Shanghai, China), and their biological parameters were measured using vernier calipers with 0.01 mm accuracy.

Table 1. Biological and mechanical differences between pepper seedlings.

	Parameters	Stem Diameter/mm	Tensile Strength/N	Pulling Force/N	Compression Strength/N
	Max value	2.41	9.45	2.03	12.94
	Min value	2.02	4.69	0.55	9.23
	Average value	2.16	6.76	1.48	11.19
	Standard deviation	0.07	0.84	0.14	0.99
	Coefficient of variation %	3.24	12.35	9.55	8.83

Note: Points a to b are the tensile strength testing area of the pepper seedlings, and point c is the compressive strength test area.

shows the statistical results of the biological and mechanical differences between pepper seedlings.

We selected from point a to b in the figure of Table 1 for the tensile strength testing of pepper seedlings, and point c for the compressive strength testing. Table 1 illustrates some differences in the characteristic mechanical parameters of pepper seedlings with essentially the same growth characteristics. The clamping force results show that the separating force is much smaller than the tensile and compressive strength. During the pick-up process, the seedling picking mechanism does not usually damage or pull off pepper stems while removing the seedings from the cavity trays. Cavity seedlings are mainly damaged during transplantation due to radial compression of the stem by the gripper.

In the process of radial extrusion, the crop stem undergoes a certain degree of elastic deformation. In order to obtain more accurate results, we analyzed the data concerning radial load and displacement changes in pepper cavity seedlings. Figure 1 illustrates the variation curve for the radial load on pepper cavity seedlings with compression. The linear growth fitting formula for this segment is y = 0.2589x − 0.8, and there is an approximately linear relationship between the load and compression when compression increases from 0 to R. After the load is applied and before the R point is released, the pepper seedling stems return to their normal state, and this does not affect the growth of the cavity tray seedlings. There is a point F within the stem rupture limit load point, where the RF section continues to apply a radial load, causing stem plastic deformation. Despite releasing the load at point F, the stem cannot return to its normal state, which reduces the growth of transplanted cavity tray seedlings. This paper presents a theoretical framework for the study of grippers based on the experimental data obtained from the study of the mechanical properties of pepper seedlings.

2.1.2. Overall Structure of the Gripper

Based on the mechanical properties of pepper cavity seedlings, the stalk compression of cavity seedlings should not exceed 50%, and the maximum clamping force of the gripper was set at 10 N. Figure 2a shows the overall structure of the designed seedling picking gripper. The main body of the gripper includes a base, a servo, a mounting plate, a slide rail, and several connectors. The gripper finger I is connected to the rack through the connector, and the slide rail is connected to the gripper base. For transmission, the rack engages with the gear on the servo rudder plate, and the gripper finger II is attached to an elastic actuator attached to the gripper base. To complete the process of detection and grasping, the gripper consists of linear movement of the gripper finger I and micro-movement of the elastic actuator of the gripper finger II.

The gripper was designed based on the 72-hole square vertebral body hole plate size. The length and width of the tray are 540 × 280 mm, the upper hole diameter is 40 × 40 mm, the depth of the hole is 45 mm, and the distance of the hole is 45 mm. For the gripper to be effective, it should be designed in such a way that it does not interfere with other seedlings in the cavity tray. It is shown in Figure 2b that there is a certain distance between the two gripper fingers so that potted seedlings can be picked up accurately when taking the seedlings, and the potted seedlings can fall smoothly when casting. The distance between the two fingers can be adjusted by the servo (0–20 mm). The gear rack drive distance can be calculated based on the formula l = αrπ/2d, when the servo rotates 0.1°, the drive distance is 0.07 mm. In the case of other size trays (maximum 200-hole tray), the gripper fingers need to be replaced and the distance between them needs to be adjusted in order to meet the upper hole diameter requirements of the tray.

2.1.3. Work Process of Gripper

Figure 3a illustrates that when the gripper is not gripping the seedlings, the elastic actuator is not subject to external force, the voltage value detected by the linear Hall element remains constant, and the gripping force is zero.

In Figure 3b, the gripper drives the gripper finger I towards the x-axis through a servo, and the finger continues to move after contacting the stalk. The force causes the gripper finger II, connected to the elastic actuator, to be micro-displaced. This causes the magnet to slide away from the linear Hall element, resulting in a change in the output voltage value. When the detection data reaches the pre-set value, the servo stops running and maintains the output torque; the gripper then moves up along the z-axis to remove the seedling from the cavity tray, and moves to the target area. Then, the servo reverses, the gripper finger is released, and the elastic actuator returns to the original position, completing the seedling feeding operation. Figure 3c shows the operation process of the gripper. To ensure that the gripper is not interfered with during operation by the cavity tray, the tray is bent along the conveyor mechanism. Upon completion of a clamping action, the conveyor mechanism runs a distance, the empty cavity is folded, and the next row of seedlings becomes the “first row”. This action is repeated until all seedlings in the tray are captured.

2.2. Analytical Model

2.2.1. Magnetic Model

The impact resistance of sensors is poor and the sensors are in direct contact with the mechanics of the system and susceptible to damage. Elastic materials are usually used as an actuating element for indirect detection [28,29,30]. The gripper detection system uses an elastic actuator for indirect detection in order to avoid damaging sensitive components. The linear Hall sensor has high linearity when it detects small displacement amounts [31]. The Hall element uses the AH49E integrated module, which can output a low-noise signal without external filtering, with an operating temperature range of −40 to 85 °C, an output voltage sensitivity of 1.6 mV/GS, and a linearity of 0.7%. As shown in Figure 4, the Hall element is mounted perpendicular to the magnet.

The static output voltage, U₀, is half the supply voltage, the linear Hall element produces a coefficient of output of K_H, and output characteristics can be expressed as follows:

$$U=U_0\pm K_{H}B$$

(1)

Considering a magnetic cylinder with height and diameter H and R, respectively, we can derive the magnetic field B on its central axis at a distance from the origin of the coordinates by applying the following equation [32]:

$$B\left(u\right)=\frac{B_r}{2}\left(\frac{u}{\sqrt{u^2-R^2}}-\frac{u-H}{\sqrt{{\left(u-H\right)}^2+R^2}}\right)$$

(2)

The magnetic cylinders H and R have diameters of 1 and 1.5 mm, respectively, and B_r is the remanent magnetization. By setting the distance u between the linear Hall element and the cylindrical magnetic body to 1.5 mm, and given the cylindrical magnetic body’s remanent magnetic strength is 0.2 T, we can derive the magnetic field strength at the Hall sensor from Equation (2). The Arduino Uno development board determines the displacement distance of the gripper finger II based on the read voltage value to derive the deformation of the elastic sheet, and it calculates the gripping force by calculating the deflection of the elastic actuator to detect the dynamic gripping force.

2.2.2. Mechanical Model of Elastic Actuator

There is a vital role for the elastic actuator in the gripper detection system, both in terms of providing the connection between the gripper and the sensor, as well as making sure that a certain amount of rebound strength is maintained to ensure that the whole system functions effectively. In Figure 5, the elastic actuator is L-shaped, with one end attached to the slider (position E) and the other to the base (position D), forming a beam-like cantilever. As the clamping force is transferred from the gripper finger to the elastic actuator, point E is slightly displaced along the x-axis. The elastic actuator was simplified to a cantilever beam model since the displacement is very small. The elastic actuator must meet the force deformation requirements, so there needs to be a certain amount of deformation during clamping. If the deformation is less than the Hall sensor trigger threshold, it will not be able to obtain clamping force values. In contrast, if the elastic strength of the elastic actuator is too small, then plastic deformation will easily occur, and the gripper will lose its clamping ability. Therefore, the elastic actuator must satisfy both clamping and detection requirements.

During the clamping process, the elastic actuator is deformed by the uniform impact load, and the kinetic energy of the gripper finger I is transformed into the bending strain energy of the elastic actuator, in accordance with the formula for calculating the load and deflection of the cantilever beam [33,34], as shown below:

$$E_k=\frac{m{v}^2}{2}$$

(3)

$$\{\begin{array}{c}{V}_{\epsilon }={\displaystyle \int \begin{array}{c}t\\ 0\end{array}\frac{M_d{}^2\left(x\right)}{2EI}dx}\\ M_d\left(x\right)=F_{d}x\end{array}$$

(4)

Based on Equations (3) and (4), the kinetic energy generated by gripper finger I is converted into the bending strain energy of the elastic actuator. After simplification we obtain the following equations:

$$\frac{m{v}^2}{2}=\frac{F_d{}^2{l}^3}{6EI}$$

(5)

$$F_d=v\sqrt{\frac{3mEI}{l^3}}$$

(6)

The elastic actuator’s theoretical critical load P_dmax and maximum deflection w_B can be determined from Equations (6) and (7):

$$P_{d\max }=\frac{M_{d\max }}{W_z}=\frac{F_{d}l}{W_z}=\frac{v}{W_z}\sqrt{\frac{3mEI}{l}}$$

(7)

$$w_B=\frac{F_d{l}^3}{3EI}=v\sqrt{\frac{m{l}^3}{3EI}}$$

(8)

In the equations, v is the loading speed of gripper finger I, W_z is the flexural section coefficient, m is the mass of the parts attached to gripper finger I, E is the modulus of elasticity, I is the moment of inertia of the section, and l is the length of the elastic actuator’s cantilever.

2.3. Design of Gripper Control System

Figure 6 shows the feedback gripper control system for cavity tray seedlings. The cavity tray seedlings are clamped using Arduino technology, computer serial communication technology, and detection and execution units.

The Arduino Uno development board serves as the hardware and software connection part, reading and transmitting voltage signals from the Hall element to the host computer for real-time processing by Matlab2018a (MathWorks, Natick, MA, USA). Bus servo adapter boards provide power to the servo and transmit the servo signals to the Arduino Uno development board. As soon as the Arduino Uno development board determines that the signal value reaches the pre-set numerical value; it will also control the servo to stop rotating and maintain the servo output torque in order to remove the cavity seedlings from the cavity tray in a smooth manner. Once the seedlings have reached the predetermined position, the host computer sends a signal to the Arduino Uno development board, which then releases the gripper, and the seedlings fall into the seedling cup. With the Bluetooth module, mobile devices can receive and send data.

2.4. Modeling and Calibration of Clamping Force Detection System

Based on the elastic actuator’s mechanical model equations, Figure 7 shows the maximum deflection of the elastic actuators at 10 N clamping force. It is obvious that the displacement distance of the slider exceeds the design requirements if the elastic actuator thickness is less than 0.4 mm, and the elastic actuator is likely to fail. The Hall sensor cannot detect micro-displacements in time if the elastic actuator thickness exceeds 0.8 mm. We chose to test three elastic actuators with thicknesses around the critical value and a wide variation in maximum deflection because of actual processing costs. The elastic actuator thicknesses were 0.4, 0.5 and 0.6 mm.

Considering the accuracy of the magnetic parameters and mechanical parameters of the studied system, we used a numerical solver to calculate the magnetic field generated by the permanent magnet. Various parameters of the clamping force detection system were designed based on the relationship between the voltage-distance variation of the Hall element and the deformation of the elastic actuator. The trajectory of the magnet relative to the sensor was assumed to be approximately linear for minor deflections generated during the clamping process. We used the COMSOL Multiphysics finite element method (FEM) to model the clamping force detection system. The detection system consists of a Hall sensor and a magnet. The Hall sensor is arranged parallel to the magnet at the bottom of the base, and the magnet is attached to the elastic actuator connected to the gripper finger II. Using a simplified clamping force detection model, the magnet is placed over the elastic actuator and its displacement is only considered after force deformation, ignoring the influence of the magnetic field on the elastic actuator. This simplifies the calculation process of the simulation software. Figure 8a illustrates the deformation of the elastic actuator and the distribution of magnetic lines of force. Based on the simulation results, the initial magnetic field strength is 0.0161 T at a 1.5 mm distance between the magnet and the Hall sensor.

A multi-physics field module coupling using COMSOL (solid mechanics module and magnetic field without current module) was used to examine the impact of stress input on Hall sensor flux density variation. The elastic actuator sheet has the following material characteristics, according to the manufacturer’s data: density of 7850 kg/m³, Young’s modulus of 2.1 × 10¹¹ Pa, and Poisson’s ratio of 0.282. Figure 8b shows the variation in magnetic flux density due to different forces, which increases with an increase in external forces. In light of the fact that the average force limit for pepper seedling stalks was 11.19 N, an acceptable force range (F ≤ 10 N) was chosen. In the presence of an increased clamping force, elastic actuators of different thicknesses have flux densities of 0.0179, 0.0176, and 0.0172 T, respectively. For the clamping force detection system, the maximum displacement of the elastic actuator at a force of 10 N is 1.431 mm, which is the maximum displacement allowed by the design.

The Hall sensor is designed for a maximum flux density of 0.1 T. If the flux density output reaches saturation, the Hall sensor will fail. With a maximum flux density of 0.0185 T for the Hall sensor, the maximum displacement of the elastic actuator can reach 1.431 mm, which is much smaller than the saturation threshold. A variable distance exists between the Hall sensor and the magnet, ranging from 0 to 10 mm. The Hall sensor detection distance can be adjusted in order to accommodate different clamping objects. The spacing between the Hall sensor and the magnet in the above analysis is 1.5 mm.

The clamping force detection system is calibrated based on the theoretical model derived from the modeling. Figure 9 shows the experimental setup used to characterize the clamping detection system for cavity seedlings. It includes an electro-mechanical universal testing machine, UTM (Changchun Machinery Institute, Changchun City, Jilin Province, China), a force feedback gripper, and a computer. The UTM has a force measurement range of 0–50 kN with an accuracy of ±0.5%. Through experiments and data fitting, the regular function of the Hall sensor’s voltage change is derived when the magnetic field changes. Arduino A0 is used as an analogue input port for capturing voltage values detected by the Hall sensor. The computer reads the magnitude of the force and the voltage value of the sensor. Based on the simulation results and the earlier tests, a known force was applied to the gripper detection system by the UTM. Calibration of the detection system is performed using a value ranging from 0–10 N, and the computer records the voltage value accordingly. We collected 100 sets of data for each thickness of the elastic actuator by increasing the force by 0.1 N each time. The test was conducted three times, 300 sets of data were collected, and an average of each point was calculated. Figure 10 illustrates the calibration curve of the clamping force detection system based on least squares fitting. Load and output voltages were determined with linear coefficients of determination of 0.974, 0.998, and 0.986, respectively, indicating that the load and output voltages of the clamping force detection system are highly linear. The elastic actuator with a thickness of 0.4 mm does not deform plastically when loaded with 10 N and can still return to its original position when unloaded.

2.5. Calibration Results and Anlysis

1.

Range of clamping force detection system

There are two measurement ranges for the clamping force detection, which are the output voltage measurement range and the pressure measurement range. The output voltage measurement range of the gripper is −2.5~2.5 V, whereas the pressure detection range is 0~10 N.

2.

Sensitivity of clamping force detection system

The sensitivity (S) of the clamping force detection system represents the ratio of the change in the output voltage of the detection system, Δy, to the change in the load that causes the change in the output voltage Δx. It is determined by the slope of the output characteristic curve. It is expressed as follows:

$$S=\underset{n\to \infty }{\lim }\frac{\mathrm{\Delta }y}{\mathrm{\Delta }x}=\frac{dy}{dx}$$

(9)

In accordance with Equation (9), the clamping force detection system’s sensitivity is 0.0819, 0.0687, and 0.0573 V/N, and the average is 0.0693 V/N.

3.

Linearity of clamping force detection system

A linearity test indicates whether the output can maintain a linear relationship with the input system, which is also called a nonlinear error. It represents the degree of deviation between the actual input–output relationship curve and the fitted line, usually denoted by δ. The calculation formula is as follows:

$$\delta =\frac{\mathrm{\Delta }{Y}_{\max }}{Y_{FS}}\times 100\%$$

(10)

where Y_max is the maximum deviation between the actual output value of the test system and the fitted line, and Y_FS is the detection system’s measurement range. The maximum deviation values of the three different thicknesses of the elastic actuator detection system are 0.0188, 0.02446, and 0.053 V, respectively, and the clamping force detection system has a range of 5 V. According to Equation (10), the linearity of the detection system was 1.88%, 2.45%, and 5.3%, respectively, with an average linearity of 3.21%.

2.6. Test of Indoor Validation

A pick-and-place test was conducted with pepper cavity tray seedlings in order to determine the stability and applicability of the force feedback gripper. Figure 11 shows the test site.

2.6.1. Test Materials

The test seedling was Chiyan No. 2 with a seedling age of 45 days, and the plug was a 72-hole square vertebral body hole plate. The substrate was made of peat, perlite, and vermiculite at a volume ratio of 2:1:1, with a water content of 40%. A vegetable transplanting test stand, a force feedback gripper, and a portable computer were used to detect the gripping force.

2.6.2. Test Method

Testing was based on the transplanting standard and actual production [35,36,37]. Through the PLC control system of the vegetable transplanting test-bed, the detection systems of 0.4, 0.5 and 0.6 mm thickness elastic actuators were tested at a planting frequency of 60 plants/min. A total of 72 seedlings were clamped in each test (repeated 5 times), and a total of 15 group tests were carried out. To analyze the stability of the clamping force detection system (excluding clamping force detection data of failed transplants), we calculated the mean, standard deviation, and range of the clamping force of each seedling under different thicknesses of elastic actuators. We investigated the transplanting performance of the force feedback gripper under the condition that the seedlings of the cavity tray were taken and dropped smoothly without damaging their stems. The transplanting success rate T_r was defined as the evaluation metric of the transplant performance test, as follows:

$$T_r=\frac{N_{NS}-N_{CF}-N_{SD}}{N_{NS}}\times 100\%$$

(11)

where N_NS is the total number of transplanted seedlings, N_CF is the number of failed seedling pick-ups, including untaken seedlings and seedlings unsuccessfully dropped into the destination pot, and N_SD is the number of injured seedlings.

3. Results

Table 2. Statistical results of cavity tray seedling clamping force test.

Thickness of Elastic Actuator/mm	Test Number	Mean/N	Standard Deviation/N	Range/N	Transplanting Rate/%
0.4	1	7.43	0.179	0.72	100
	2	7.54	0.288	0.86	100
	3	7.43	0.265	0.85	100
	4	7.31	0.223	0.84	98.6
	5	7.41	0.306	0.95	98.6
Average		7.42	0.252	0.84	99.44
0.5	1	7.49	0.127	0.56	100
	2	7.56	0.124	0.61	100
	3	7.54	0.165	0.79	100
	4	7.53	0.139	0.8	98.6
	5	7.51	0.146	0.88	100
Average		7.53	0.140	0.73	99.72
0.6	1	7.53	0.094	0.42	100
	2	7.54	0.095	0.48	100
	3	7.53	0.106	0.46	98.6
	4	7.42	0.102	0.55	100
	5	7.61	0.091	0.68	100
Average		7.53	0.098	0.52	99.72

shows the test results. The table shows that the average clamping force detection between different elastic actuators was between 7.42 and 7.53 N, the standard deviation was between 0.098 and 0.252 N, the range was between 0.52 and 0.84 N, and the clamping success rate was between 98.6% and 100%.

Figure 12 shows the clamping force detection system’s response and the results of the average clamping force and transplantation success rate when elastic actuators of various thicknesses were used. Figure 12a shows that the elastic actuator with a thickness of 0.4 mm detected the greatest range of force variation. Test results for 0.5 and 0.6 mm elastic actuators were similar, with a smaller range of detected force fluctuations, but there were some outliers in the 0.5 mm elastic actuator test results. In each group with an elastic actuator of 0.4, 0.5, and 0.6 mm, the average standard deviation of the clamping force was 0.252, 0.140, and 0.098 N (Figure 12b), while the mean range was 0.84, 0.73, and 0.52 N (Figure 12c), respectively, with a transplant success rate of 99.44%, 99.72%, and 99.82%, respectively (Figure 12d). The results show that the designed clamping force detection system has a small measurement fluctuation range and good stability in the process of detecting the transplantation clamping force.

4. Discussion

All three thicknesses of elastic actuators performed well in the test. Through analysis, the gripper could not correctly read the clamping force after clamping the cotyledons, leading to a failure of transplanting because the cotyledons of individual seedlings were positioned lower. Another reason is that the cavity seedlings entangle with neighboring cavity seedlings during growth, causing the gripper to pull them all out at once. Currently, the majority of transplanting grippers are needle grippers, which insert needles into the substrate in order to grasp the seedlings [18,19]. In spite of these grippers’ high efficiency, their insertion depth needs to be controlled. An insertion that is too shallow will result in insufficient extraction power, and the seedling’s root system will spread out. When inserted too deeply, damage to the seedling tray may occur. A gripper with flexible needles can reduce the amount of damage caused during insertion [20]. Its structural characteristics, however, limit the arrangement of the sensors and may cause them to fall off or be damaged during insertion into the substrate. The integration of sensors into the gripper assembly is also an acceptable solution [23]. However, the manufacturing process in this approach is complex and, if a sensor is damaged, the entire component must be replaced, which undoubtedly increases production costs. By gripping seedling stalks for transplanting, the two-finger gripper avoids contact with the substrate, reducing root damage and giving a higher transplanting success rate than with needle grippers [38].

By using an elastic actuator as a feedback device, the seedling gripper designed in this paper grips the stalks and indirectly detects the amount of gripping force. As compared to existing transplant grippers, the gripper developed in this paper has two main advantages: (1) the stalks are clamped in a way that prevents substrate collapse and root damage due to the clamp being inserted into the substrate (Figure 11), thereby reducing the risk of seedling injury; and (2) an elastic actuator is used to detect clamping force indirectly, reducing the risk of sensor damage. Additionally, the distance between the gripper fingers can be adjusted, and the gripper fingers and elastic actuator can also be replaced to accommodate different cavity trays and types of seedlings.

During the test, we also found the following problems: (1) In the case of the elastic actuator with a thickness of 0.4 mm, no plastic deformation occurred. Therefore, the thickness of plastic deformation may be thinner, and the effect of a thinner elastic actuator on the gripper’s working state can be explored further. (2) In this study, we focused on pepper cavity seedlings. However, the clamping force that the stalks can withstand varies by cavity seedling type. Therefore, the elastic actuator needs to be optimized for that cavity seedling type. (3) The sensitivity of the detection system is high when the thickness of the elastic actuator is small, but the fluctuation around the clamping force value is high when detecting the transplant clamping force. This is due to frame vibrations and gripper disturbances. The use of thicker elastic actuators reduces disturbances, but thicker elastic actuators require a higher activation threshold to detect clamping force, and this should be avoided when transplanting crops with fragile stalks. (4) As a fixed value in this study, the transplanting frequency can be used in the optimization process as an influencing factor in the future.

5. Conclusions

• With the linear Hall sensor and elastic actuator, a force feedback seedling pick-up gripper was designed based on the mechanical characteristics of vegetable cavity seedling stalks. The gripper size is 200 by 60 mm, and gripper fingers are interchangeable to adapt to various working objects. An elastic actuating element with a working part of 20 × 14 mm can effectively prevent damage to the sensor due to vibration or impact during transplanting.
• The gripper control system consists of an Arduino Uno microcontroller, a servo driver board, a servo, a Bluetooth module, and a linear Hall element. The linear Hall element processes the gripping force signal and precisely controls the gripper finger movement distance through the servo, thereby preventing damage to the cavity tray seedlings and increasing transplanting success.
• In this study, three different thicknesses of elastic actuators were used to test the gripping force detection system’s performance. Based on the calibration test, the clamping force detection system has a sensitivity of 0.0693 V/N, a linearity of 3.21%, a linear coefficient of determination of 0.986, and a detection range of 10 N, all of which meet the transplant clamping force detection accuracy requirements. In the indoor transplanting tests, the clamping force detected for elastic actuating elements with thicknesses of 0.4, 0.5, and 0.6 mm ranged from 7.42 to 7.53 N, the standard deviation ranged from 0.098 to 0.252 N, the extreme deviation ranged from 0.56 to 0.84 N, and the transplanting success rate ranged from 99.44% to 99.72%. There was a low fluctuation range between the clamping force detection system and the transplanting apparatus, and the clamping apparatus is stable and adaptable.