*2.1. Device Description and Productive Process*

The robotic equipment for grafting consists of transport devices, the manipulator itself, cutting mechanisms, and devices that facilitate bonding and grip [26]. True to this premise, the study equipment consisted of two anthropomorphic robots equipped with clips adapted for manipulating seedlings, with two seedling bevel cutting devices and a device for the forming, dispensing, and placement of the graft clips (Figure 3).

(a)

**Figure 3.** *(***a**) General view of the robotic cell developed for grafting plants using the splicing technique. (**b**) Work cell sketch. (1) KUKA KR6 900 robots for manipulating the seedlings. (2) Cutting devices for rootstock and scion seedlings. (3) Forming, dispensing, and placement graft clip device. (4) Conveyor belts for seedling trays (rootstock, scion, and grafted plants). Demonstrative video of the system: https://youtu.be/9GvIDyrBBfo (accessed on 10 December 2019).

The seedling trays are loaded at the beginning of the process into conveyor belts suitable for this purpose, so that two seedling starter trays, loaded with rootstock and scion seedlings will be properly positioned in front of the working robots.

Each completed row of work is followed by an advance of the belt, which relocates the next row into the appropriate collection area. This process happens until the rows in the seedling starter trays (rootstock and scion) are finished. Likewise, there is a third conveyor belt with an empty output tray for the resulting seedlings to be placed on once grafted. This output conveyor belt also performs partial advances per row while working on that tray.

Once the trays reach the required locations, the two industrial robots, Kuka model Agilus R6 900 (developed by KUKA Roboter GmbH), work from their home position in a coordinated manner on the rootstock and scion to achieve graft completion. These robots have equivalent commercial equipment from other large manufacturers worldwide, such as the robot model IRB1200 (developed by ABB), the model MH5F (developed by Yaskawa), or the model LR Mate 200iD/7 L (developed by FANUC). All of them have similar load capacities, degrees of freedom, speeds, and working spaces, so their replacement would not lead to significant differences.

Each robot operates independently by handling each of the seedlings, which are obtained from the input trays. The rootstock and scion seedlings are approached in a simultaneous and coordinated manner: <sup>1</sup> approach the input trays (AIT). This displacement is followed by a precision operation that separates the seedlings from their trays: <sup>2</sup> grip and extraction (GE).

The final elements for the seedling manipulation consist of clamps composed of two fingers with an opening and closing parallel model MHZ2-32D from SMC, equipped with a padded extension zone of low-density (150–200 kg/m3) polyurethane foam (neoprene) with high resilience for precise and firm seedling attachment. A photocell is located between the ends of these fingers, and a Sick LL3-TB02 optical fibre sensor detects the precise location of the stems of the seedlings, which can emerge at any position within the alveolus of the tray. The individual seedlings have a unique growth morphology, and the alveoli can even be empty (Figure 4).

**Figure 4.** (**a**) Detailed view of the clamp or terminal element of the robots. (**b**) Robot and clamp sketch. (10) Industrial robots. (11) Clamping base. (12) Robot clamp. (13) Fingers with parallel opening and closing operation. (14) Padded area for holding seedlings. (15) Sensor for seedling detection.

After extracting the seedlings from their alveoli in the trays, the robotic arms carry the seedlings to the point where the cut will be made: <sup>3</sup> approach the cutting zone (AC). The system uses two equal pieces of equipment for cutting the seedlings to guarantee precision in the required cutting angle and the integrity of the dissected seedling. Both pieces of equipment are responsible for acting, one on the rootstock and another on the scion, allowing the cutting angle to be regulated and complementary between both plants. Prior to cutting, the robots insert the seedlings into a slot or channel located in front of the blade, where the stems are embedded to ensure the verticality of the stems during cutting. The cut is executed by activating a double-shank pneumatic cylinder, SMC model CXSM15-15, coupled to a terminal tool of a sharpened, disinfected, and interchangeable blade of stainless steel with a precise cut angle [27]. The cut is performed by a dry shock stroke of the blade against the seedling, which cleanly bisects the stem and ensures a clean cut: <sup>4</sup> cutting process (C). Meanwhile, an external blower separates the non-useful part of the treated seedling (Figure 5).

**Figure 5.** (**a**) Detailed view of the cutting device. (**b**) Cutting device sketch. (21) Pneumatic cylinder to drive the blade. (22) Cutting blade. (23) Slot for seedling lace. (24) Worktable. (25) Hole for waste disposal. (26) Waste accumulation box. (26) Blower.

After the cut, the robotic arms carry the useful parts of the seedlings for the graft towards the bonding area: <sup>5</sup> approach the clip dispensing zone (ACD). The graft equipment that creates the bond is responsible for cutting plastic clips from a continuous roll and then dispensing and placing a clip on the seedlings to be grafted. Thus, this device contains two subsystems: one in charge of preparing the clip and another in charge of placing it.

The first clip preparation subsystem consists of a series of electropneumatic devices that act in a coordinated and sequential manner to perform the clip cutting process from a continuous roll of plastic tube, regulating the advance and thus controlling the length of the cut clips. To obtain a clip with the desired length, an SMC model CJPB10-5 microcylinder presses the plastic tube against the end of a second cylinder, an SMC model CXSM15-15, which advances in the vertical direction the exact length that the clip is desired to have. Finally, an SMC model CXSM15-15 cylinder, equipped with a sharp blade at its end, makes a clean cut with a sharp blow, creating the clip to be used.

The second subsystem contains a rotating cylinder, a Festo model DM-6-90-PA, which grips the cut clip, then tightens its wings and thus clamps and fully opens the clip. Finally, with an SMC model CXSM25-70 cylinder, the clip is brought closer by a precise horizontal movement to the junction point where the two seedlings to be grafted are located. At the point of clip placement, a passive fitting device is adapted to accommodate and locate the seedlings in front of the clip. To ensure a precise bond without unwanted seedling displacement, both parts remain in contact and can thus fasten the bond. Once the clip is placed on the seedlings, the device that kept the graft clip pinched loosens and moves away from the junction point, returning to its resting position (Figure 6).

(a) (b)

**Figure 6.** *(***a**) Detailed view of the device that dispenses and places the graft clip. (**b**) Sketch of the device that dispenses graft clips. (31) Pneumatic cylinders and blade to make the grafting clip. (32) Continuous silicone roll for clip manufacture. (33) Tweezers for the grip and clip opening. (34) Pneumatic cylinder responsible for bringing the grafting clip closer to the junction point. (35) Worktable. (36) Plant placement device and grafting clip insertion point.

The graft is accomplished when both plants are placed in intimate contact with one another and the graft clip is pressed onto them: ➅ clip preparation and placement (CPP). Once the clip is placed, the robot holding the scion releases the graft and withdraws, leaving the bonding area, while the other robot, which holds the completed graft by its lower part, moves the graft by the rootstock to the tray where the grafts will be left once finished: ➆ approach the output trays (AOT). The graft is deposited in an alveolus of the output tray: ➇ insertion on trays and release of plants (ITR).

Once the process is finished, both robots return to their resting positions, either as an end point or as a point of passage from where a new work cycle begins: ➈ return home (RH). The process is repeated until the work trays are completed. The tray with the grafted seedlings is removed, and the grafted seedlings are subjected to a post-grafting process of healing, where their success is examined over 14 days: ➉ post-graft losses (PGL). The entire system and grafting equipment described, as well as other secondary elements and auxiliary equipment, are managed and coordinated in a global manner through a central control unit, consisting of a PLC model CJ2 M, with an Omron CPU32.

During the grafting operation, 10 control points were established as singular intermediate points of reference in the process, which allowed us to record the partial times used and a distribution of failures during grafting (Figure 7).

The flowchart describing the operations and process described above allowed us to evaluate the validity and efficacy of conventional industrial robotics applied to tomato seedling grafting using external low-cost passive devices that facilitate grafting completion. The external devices act as a tool both in cutting the seedlings and in the dispensing and placing of the graft clip (Figure 8).

**Figure 7.** Travel details. Singular points: ➀ approach the input trays (AIT); ➁ grip and extraction (GE); ➂ approach the cutting zone (AC); ➃ cutting process (C); ➄ approach the clip dispensing zone (ACD); ➅ clip preparation and placement (CPP); ➆ approach the output trays (AOT); ➇ insertion on trays and release of plants (ITR); ➈ return home (RH).

**Figure 8.** Flowchart of the developed automated grafting process.

#### *2.2. Definition of Operating Conditions*

The experiment was conducted at the Tecnova Technological Center (Centro Tecnológico Tecnova): Foundation for Agricultural Technologies of Agriculture, in Almería (36◦52 38"N, 2◦19 59"W) between the months of April and June 2017. The environment of Almería is a model of agricultural exploitation. Greenhouse growth of arable fruit crops has a high technical and economic performance, with tomato cultivation being of particular importance.

The rootstock used in the experiment was the interspecific hybrid "Maxifort" from De Ruiter SeedTM, which is recommended for crops with better behavior at low temperatures and under high salinity conditions. The "Ventero" variety from De Ruiter SeedTM was used as an indeterminate hybrid tomato for branch harvest. Both types of seeds are routinely used in seedbeds to perform manual grafts using the "tomato on tomato" (T/T) technique, which demonstrates their prior compatibility with the robotic system.

The growth protocol developed in the nursery attempted to obtain plants that were grown and cared for until reaching a similar growth state between the rootstock and scion, with mature plants and those prepared for the graft having two–four well-defined true compound leaves [28] and stem diameters of at least 1.5 mm for the splicing method [29]. Therefore, stems with some natural variability characteristic of the development of each plant (between 1.5 and 2.5 mm in diameter in the area close to the cut for the scion, and between 2 and 3 mm in diameter for the area close to the cut for the rootstock) were worked with. Usually, in the automatic graft, the requirements in terms of growth and uniformity required for the rootstock and scion seedlings are as critical as in the manual graft [30], demanding an arduous previous task of pre-selection and pairing similar diameters between the linked seedlings. In our experiment, this work was eliminated because the seedlings were cut with a bevel at a 60◦ angle. From a certain cutting angle, between 50◦ and 70◦, and provided that one works within the margins of natural variability between the previously established stems, the success rate of the graft was acceptable and higher than 95%. Therefore, the need to seek equal diameters of the workplaces was of lessened importance [27,31].

Prior to each experiment and for each tray, it was ensured that all the alveoli slots contained seedlings that met the previously established rootstock size criteria. The environmental conditions were regulated during the grafting process, with temperatures between 20 ◦C and 25 ◦C, conditions of relative humidity that were sometimes forced and were guaranteed to be above 75%, and stable, non-direct daylight luminosity conditions.

The data were collected for each test via filming and were then timed; the times until reaching each control point of the process until the graft was completed or the point of generation of each failure were evaluated.

Regarding the post-grafting conditions, the plants began to wilt immediately after cutting and grafting, so once each graft tray was finished, it was immediately introduced into a small healing chamber consisting of a tunnel slightly larger than the dimension of each tray and a low height, covered by a transparent film. This tunnel was placed inside a chamber in which the climatic conditions were controlled throughout the healing process. During the first 48 h, the plants were kept without illumination to reduce transpiration and evaporation. On the following days, the intensity of the light was increased, and a 14 h light photoperiod with a value <sup>≈</sup> 100 <sup>μ</sup>mol·m−2·s−<sup>1</sup> of PAR, (~ 3000 lux) of non-direct and diffuse light was established during the callus formation stage, from LED lights, corresponding to a value slightly above the compensation point because there is evidence that a high intensity of light prevents callus formation [1]. The level of illumination was gradually increased after several days. The temperature was established with a variable set point in the healing chamber between 23 ◦C and 30 ◦C, with an average of approximately 26 ◦C, slightly varying between the diurnal and nocturnal conditions. The relative humidity was initially established between 75% and 95% in an attempt to reduce the transpiration rate of the scions, avoiding high stress and thus preventing the drying of the graft [32]. The humidity level was gradually reduced in successive days to condition the grafts to the outdoors. The vapor pressure deficit (VPD), during the critical graft healing phase was

around 0.8 kPa in the tunnel inside of the healing chamber, with the aim of decreasing transpiration. This value was gradually increased in the following days.

The success or failure of the final graft was evaluated by estimation and visual assessment performed daily across the 14 days after grafting, assessing the natural evolution of the graft, and analyzing other symptoms and external evidence that would determine its classification as either a success or a failure. In making this determination, key intermediate points were considered to mark an inflection of the task or singularity within the process, so that the successful or unsuccessful completion of this phase of the robotic process could be evaluated.
