Development of a precision 3-row synchronised transplanter

1 Development of a Precision 3-row Synchronized 1 Transplanter 2 3 M. Pérez-Ruiz 1,* and D.C. Slaughter 2 4 5 1. Universidad de Sevilla. Área de Ingeniería Agroforestal 6 Dpto. de Ingeniería Aeroespacial y Mecánica de Fluidos. 7 * Ctra. Utrera, km 1 8 Sevilla 41013, Spain 9 Phone: (+34) 954481389 10 Fax: (+34) 954486436 11 E-mail: manuel[email protected] 12 2. University of California, Davis. 13 Department of Biological and Agricultural Engineering 14 15 Abstract. 16 Commercial vegetable crop transplanters currently use several unsynchronized planting units 17 mounted to a common transport frame. The objective of this work was to assess the performance 18 of a new transplanting technology to improve the plant placement accuracy and spatiotemporal 19 planting synchronization across adjacent rows, thus producing a grid-like planting pattern using 20 adjacent vegetable crop transplanters. The feasibility of synchronization of adjacent transplanting 21 units for vegetable crops was demonstrated using tomato as the target crop. A colour, digital, 22 high-speed computer vision analysis of the motion and dynamics of the plant trajectories of 23 transplanted tomatoes was conducted. The high-speed video analysis led to the design and 24 testing of an improved plant support mechanism to enhance the control and precision of the 25 transplanting of vegetable crops. The absolute deviation values of the final location in the soil 26 were reduced by approximately 25% for both the right planter and left planter compared to those 27 in previous years. These results serve as the fundamental basis for a mechatronic system that 28 Con formato: Normal 2 can precisely transplant vegetable crops in a grid-like pattern across rows as a critical first step in 29 a systematic approach to fully automated individual plant care. 30 31 Key words. Precision agriculture; High-speed video; Transplanter; Computer vision 32 1. Introduction 33 Precision agriculture implies accuracy in the distribution of plants and seeds over fields. Optimal 34 distributions are defined by the size and shape of the area that provides nutrients, water and light 35 to plants to obtain the maximum possible yield. These factors are controlled by the spacing 36 between crop rows and the spacing between plants/seeds in a row (Klenin, Popov, Sakun, 1985; 37 Barbieri et al., 2012). The relationship between crop yield and plant spacing has been well 38 documented (Nielsen, 1991; Doerge, Hall, Gardner, 2002). For many crops, the row crop spacing 39 is determined as much by the physical characteristics of agricultural machinery used as by the 40 specific biological spacing requirements of the crop (Blas, Barreireo, Hernández, Dias, 41 Griepentrog, 2013, 201). In addition, a key aspect to the accurate planting of a precision 42 transplanter/seeder for a typical square grid pattern requires equipment calibration checks 43 equipment calibration before planting in both the longitudinal and transverse crop directions 44 (Perez-Ruiz, Slaughter, Gliever, Upadhyaya, 2012). 45 The importance of precision planting is a well-established concept. In addition to its impact on 46 optimizing overall yield, it affects the uniformity of crop emergence and in specialty crops like head 47 lettuce it impacts labour costs associated with the management and harvest of the crop (Harriott, 48 1970). Nielsen et al. (2006) indicated that irregular plant spacing within rows decreased 49 maizecorn grain yield at rates of up to 125.5 kg ha -1 for every 0.25 m increase in the standard 50 deviation of plant-to-plant spacing within a range of plant spacing variability from ~0.05 to ~0.36 51 m. Another reason to improve the uniformity of plant distribution in the field is to ensure a rapid 52 crop emergence and (uniform crop canopy closure) and hence achieving soil cover at the same 53 3 time that crop’s development, thereby avoiding patches of the field that are exposed due to poor 54 plant spacing (Upadhyaya & Blackshaw, 2007; Heege, 2013). If tomato transplants are positioned 55 in a predeterminedregular grid patterns, resource utilization (nutrients, water, light and space) 56 canwill be optimized. The use of a rectangular grid can improve , thereby improving the ability of 57 the crop to suppress weeds and making it easier to utilize automated mechanical or physical 58 weeding treatments. 59 For agronomic reasons (e.g., earlier harvesting, better crop establishment in challenging farming 60 conditions, and better development of root systems), transplants are utilized for some crops; these 61 plants are usually grown in transplant cell trays to reduce plant stress at the time of transplantating 62 and because plants of uniform size are needed for the transplanter (Parish, 2005). In 2017, it was 63 estimated that 70% of the 113000 hectares of canning tomato were transplanted in the USA 64 (Miyao, Aegerter, Sumner, Stewart, 2017); according to technicians and tomato growers today, 65 this percentage may now be 10-20% higher. 66 The plantning of mostMost vegetable crops is currently done using use a mechanical transplanter 67 (carousel or pocket-type transplanter), despite its dependence on athe requirement for significant 68 amount of manual labour to feed and operate the machine. At this point, it is possible to find 69 commercial automatic transplanting machines for greenhouse production (Visser Group, Holland, 70 2014). However, semi-automatic (Zhang, Cao, Wang, Zhang, 2013) and automatic feeding 71 mechanisms are gradually being developed and are beginning to reach the market for vegetable 72 plug seedling (Han, Kumi, Mao, Hu, 2019; Pearson Autoplanter, 2020). A pocket-type planting 73 unit has a number of spring-loaded plant pockets arranged at equal intervals on a drum. The drum 74 is driven from ground-contacting wheels through a gear (Rotty, 1960) or chain drive. When the 75 drum rotates, a plant pocket opens as it approaches the top of the cycle, receives the transplant, 76 closes, carries the transplant downwards, and releases the transplant in the furrow. The current 77 designs of commercial vegetable crop transplanters utilize several unsynchronized (i.e., the units 78 4 operate independently in terms of their drive mechanism) planting units mounted to a common 79 transport frame. These systems use suboptimal open-loop methods that neglect the dynamic and 80 kinematic effects of the mobile transport frame and the plant motion relative to the frame and the 81 soil. The current designs also neglect to employ complete mechanical control of the transplant 82 during the planting process, thus producing an error in the final planting position due to the 83 increased uncertainty of plant location in the soil, which is related to natural variations in the plant 84 size, plant mass, and soil traction and compaction characteristics (Prasanna Kumar & Raheman, 85 2008). Due to plant/field variations and transplanter operation conditions, the desired 86 synchronization can’t possibly occur if there is no linkage between the planting units. The in-field 87 operation is based on the rotation of a shaft which is controlled by the travel speed of the 88 transplanting unitmachine through the field and is often powered by separate ground-driven 89 wheels that are affected by localized slip conditions between each transplanting unit’s wheel and 90 the adjacent soil in each row. While these. These transplanters may seem to be designed to 91 maintain a uniform distance between plants, row-to-row variability in soil conditions causes plant 92 spacing variability and because there is no linkage between adjacent planting modules there is 93 no synchronization in plant placement between rows.according to manufactures, this 94 synchronization does exist 95 With the objective of simultaneously reducing the drudgery of menial labour associated with 96 vegetable production and the need to reduce the negative environmental impact of pesticides, 97 the scientific community is working on the development of new solutions for individualized plant 98 care. One such approach is to create a plant map using real-time kinematic (RTK) global 99 positioning systems (GPSs) by monitoring the seeds (Ehsani, Upadhyaya, Mattson, 2004) or 100 transplants (Sun et al., 2010) while they are being planted. Studies conducted at UC Davis have 101 shown differences between RTK-GPS-based expected seed locations versus actual plant 102 positions. The difference in actual vs. expected position ranged from 3.0 to 3.8 mcm, and for 103 5 seeds, and for tomato transplants, the root mean square error (RMSE) was 2.6.7 mcm in the 104 along-track direction. Taking into account the error due to the accuracy of RTK-GPS and other 105 factors that influence the seed placement and seed dynamics (Ehsani, Upadhyaya, Mattson, 106 2004), these results show good performance.. Nakarmi & Tang (2012) used a platform forof image 107 acquisition after planting operations to estimate the inter-plant distance along crop rows. This 108 system was able to measure inter-plant distance with overall mean RMSE of 1.7 mcm and mean 109 plant misidentification ratio of 2.2%. Similarly, Zong, Liu and Zhao Zong et al. (2020) investigated 110 a real-time plant location using a machine vision system which the purpose of individual plant 111 fertilizationpropose of fertilization for individual plant, under weedy condition on sunny days, has 112 a minimum recognition rate of 88% for maize. 113 Today, one of the challenges associated with agricultural row crop production in industrialized 114 countries is the non-chemical control of intra-row weed plants (i.e., within a crop row). A number 115 of researchers have documented the negative impacts of weeds on crop productivity: Fennimore 116 & Umeda (2003) found that 4 weeks of weed competition reduced lettuce yields by 11 to 23%, 117 Wicks, Johnston, Nuland and Kinbacher (1973) found that onions infested with annual weeds for 118 8 weeks reduced yields by 65%, Ngouajio, McGiffen and Hembree (2001) found that 1.5 velvetleaf 119 weeds per meterfoot of row caused yield losses of up to 80% in tomato. Several researchers 120 (e.g., Bell 1995; Haar & Fennimore, 2003) have documented that weeds in vegetable crops also cause 121 contamination of the harvested crop and can be a host of disease. The complete elimination of weeds 122 within the seedline with mechanical methods is very attractive for organic growers, and such 123 methods involve mechanical knives and rotating hoes that are suited for use with robotic weed 124 control actuators (Perez-Ruiz, Slaughter, Fathallah, Gliever, Miller, 2014; Lati, Siemens, Rachuy, 125 Fennimore, 2015). However, the performance of mechanical weed removal is constrained by 126 plant spacing, the closeness of the weed to the plant, the plant height and the actuation time 127 between plants. 128 6 In the past decade, several research groups have worked diligently on precision intra-row weed 129 control (Jørgensen et al., 2007; Nørremark, Griepentrog, Nielsen, Søgaard, Nørremark et al., 130 2008; UlloaUlloa, Datta, Knezevic et al., 2010; Van Evert et al., 2011; Forcella, 2012; Perez-Ruiz 131 et al., 2012; Raja et al., 2019). These systems have demonstrated the importance of highly 132 accurate plant placement to facilitate the use of mechatronic or robotic technology in weed control 133 tasks. The precise planting of vegetable crop transplants in an orchard-like grid planting pattern 134 is a critical step in a systematic approach to automated individual plant care methods, such as 135 automatic mechanical or grit-abrasion weeding, precision spraying for pest control, fertilization, 136 and irrigation. 137 The goal of this work was to develop and analyse the performance of a precision plant placement 138 technology to improve the plant position accuracy and synchronize the planting pattern between 139 adjacent rows planted in the same planting pass. The specific objectives were as follows: 140 1. To design and evaluate the performance of a three-row synchronized transplanter using 141 tomatoes as the target crop; 142 2. To evaluate the plant kinematics of tomato plants during transplanting; and 143 3. To evaluate the effect of seedling support during transplanting on the precision of 144 transplanting performance. 145 146 147 148 149 150 2. Materials and methods 151 2.1. Synchronized three-row transplanter design 152 The purpose of the 3-row synchronized system is to consistently space tomatoes along a row and 153 align the plants across three rows in a grid-like rectangular planting pattern. Since California 154 processing tomatoes were the target crop, the planters used in this study were spaced 1.5 m 155 7 apart. Three finger-type, positive-placement, vegetable transplanters (model 1600, Holland 156 Transplanter Co., Holland, MI, USA) were mounted on a transplanting sled (SWEMEC Woodland, 157 CA) that was designed for synchronized operation. This modified transplanter design used three 158 precision planting wheels custom manufactured for synchronized operation and allowed a simple 159 telescoping mechanical linkage to connect and synchronize the angular rotation of adjacent 160 planting wheels (Figure 1). Mechanical synchronization was achieved by using splined power 161 take-off (PTO) drive shafts on the planting wheels and two ~1.5 m matching PTO splined linkages 162 to connect the adjacent planting wheels. A torque-limiting hub was mounted in the centre of each 163 planting wheel to allow both overload protection (in the event of an object getting jammed in the 164 planter) and a splined drive shaft. 165 Figure 1. Photograph showing the PTO-style synchronization shafts (black with yellow safety end covers) connecting the three precision planting wheels. 166 In addition to the main frame and 3-planting wheels, this transplanter included soil preparation 167 equipment that opened three planting furrows in the planting beds ahead of the positions where 168 the tomato plants were placed. The transplanter also included three furrow-closing devices that 169 closed the planting furrows and pressed the soil around the plant root balls to provide the 170 necessary growing conditions and support after plants were released by the machine. In a 171 traditional, unsynchronized transplanter, a set of two steel ground wheels (560 mcm diameter and 172 8 7.6 mcm width, with two ground wheels per planter) serve as both the source of mechanical power 173 to rotate the planting wheels and as the furrow-closing device. Due to the location of the PTO 174 synchronization shafts, custom furrow-closing steel wheels (30.5 mcm diameter, 80 mcm width) 175 were required that would fit below the synchronization shaft on both sides of the centre planter 176 and inside the two outside planters. 177 178 2.2. Synchronization system 179 The first prototype of the synchronized 3-row transplanter used custom steel ground-driven 180 wheels (Figure 3) to provide mechanical power to drive the planting wheel rotation (Figure 2). 181 However, planting inconsistencies were observed in the year-1 field trials due to differences in 182 ground wheel slip between the two ground wheels shown in Figure 3 and the elasticity in the long 183 drive shaft that connects the three planting modules. To improve both the accuracy of the plant 184 spacing along a row and the accuracy and precision of the planting pattern synchronization, a 185 hydraulic power system was designed in year-2 to drive the planting wheel. 186 Figure 2. A planting wheel with the splined, PTO-style, index shaft used for synchronization 187 9 Figure 3. View inside the year 1 synchronized transplanter showing the splined PTO-style connecting shafts used to synchronized the rotation of the planting wheels. The red outer ground wheels supplied power. 188 For the second-generation transplanter, the synchronization system was composed of an 189 unpowered gauge wheel equipped with an optical shaft position encoder (model 0622 Grayhill, 190 Inc., IL, USA); a 2-way electrohydraulic, proportional, directional control valve (model 191 16 Figure 6. Photograph showing the position of the root ball relative to the fingers of the transplanter. The root ball (held inside the person's hand in the foreground) is typically cantilevered 7.5 mcm past the end of the finger. 291 Improvements to the 3-row synchronized transplanter for the on-campus field test in year-3 292 included: change to automobile drive shafts, reduced the play in the driveline, added the metal 293 supports and changed the rubber in the fingers. Thirty rows of tomatoes were planted (single crop 294 row/bed with 1.5-m bed spacing), and all rows were planted at a travel speed of 1.6 km h -1 . In this 295 year, seventy-two sets of conditions for the metal support treatment and sixty-nine sets for the 296 control treatment of tomato plants were measured along the 3-row planting sets. These 297 measurements were analysed to determine the accuracy of the synchronization across each 3-298 row planting set as well as the plant spacing consistency along each row in the direction of travel. 299 300 301 17 3. Results and Discussion 302 A synchronized, 3-bed precision transplanter was developed for the precision planting of 303 vegetable seedlings in a three-row, grid-like planting pattern. The field results demonstrate that 304 the system was effective at synchronously placing seedlings plants were precisely placed at the 305 same grid location in the direction of travel along three adjacent rows. This system was specifically 306 designed to achieve a high level of synchronization for the rotational positions of adjacent 5-finger-307 type tomato transplanters. 308 3.1. Three-row transplant position and synchronization 309 A summary of the final plant placement of tomato plants along the row is shown in Table 1 for the 310 trials conducted in three years. In year-1, the planting wheels of the transplanter were powered 311 by a traditional chain drive that linked the shaft of the planting wheel to the 560-mcm-diameter 312 steel (red) soil packing wheels shown in Figure 3. The target plant spacing along the row was 380 313 mcm. The best matching mechanical gear ratio was selected from those provided by the 314 transplanter manufacturer to achieve this target spacing. The mean, as-planted, spacing was 315 fairly close to the goal, with a mean error of ~9 mm and a standard deviation of ~30 mcm. 316 317 In the year-2 trial, the power required to rotate the three planting wheels was supplied by a 318 hydraulic gear motor rather than the soil packing wheels. Thus, the mechanical coupling between 319 the packing wheel and the planting wheel was replaced by a digital control virtualdata coupling 320 between an unpowered ground wheel speed sensor and a feedback control circuit for the 321 hydraulic motor speed controlling the planting wheel rotation. A closed-loop feedback control 322 system was used to control the angular velocity of the planting wheel, where the planting wheel 323 velocity setpoint was determined by sensing the angular velocity of the unpowered ground wheel 324 using an optical shaft rotation encoder. The mean, as-planted, spacing was closer to the target 325 18 (380 mcm) using the hydraulic power system and unpowered ground wheel sensor. The 326 improvements in the design form year 1 to 2 showed that the mean plant spacing error decreased 327 from ~9mm in year-1 to ~3mm in year-2, and the improvement in the plant spacing accuracy was 328 statistically significant at α = 0.01. In contrast to the improvement in the plant spacing accuracy, 329 the plant spacing precision was not significantly (p-values > 0.15) improved for two of three 330 planters (left and centre), while a significant ( α = 0.01) improvement was observed in the precision 331 of the right-side planter. In addition to the planting performance improvements, Because the use 332 of aof the digital interface for the feedback control system in year 2, allowed the operator to easily 333 make, changes to the desired plant spacing could be implemented by keypad entry, rather than 334 by mechanically changing drive sprockets, as was required in year-1, which represents a 335 significant improvement in ease of use. 336 337 Table 1. Summary of the Plant Spacing Accuracy and PrecisionPerformance 338 339 *Treatments with different grouping letters are significant at the alpha = 0.01 level. 340 341 Plant Spacing* (mcm) Year Power Source Obs. Planter Mean Std. Dev. 1 Ground Wheel 288 Left 37 . 1 . 6 0 a 2 . 6 . 4 0 a 2 Hydraulic Motor 90 Left 38.0.50b 2.6.90a 1 Ground Wheel 288 Cent re 37 . 0 . 8 0 a 2 . 8 . 2 0 a 2 Hydraulic Motor 90 Centre 38.2.30b 2.2.09a 1 Ground Wheel 288 Right 37 . 0 . 3 0 a 3 . 0 .0 0a 2 Hydraulic Motor 90 Right 38.3.30b 2.0.30b Tabla con formato Con formato: Izquierda, Posición: Horizontal: Centro, Con relación a: Columna, Vertical: 0 cm, Con relación a: Párrafo, Horizontal: 0,25 cm, Ajuste automático 19 The distribution of the relative plant alignment between the three synchronized planters in the 342 year-1 trial is shown in the upper two histograms in Figure 7, and a summary of the mean and 343 standard deviation values is shown in Table 2. The tomato plants planted by the transplanter on 344 the left lagged behind the centre plants by 1.3.2 mcm on average, with 90% of the tomato plants 345 planted by the left planter being within -5.1 to 2.5 mcm of the corresponding centre plant, where 346 a zero value indicates perfect alignment. Similarly, the tomato plants planted by the transplanter 347 on the right were planted 1.4.7 mcm ahead of the centre plants on average, with 90% of the 348 tomato plants planted by the right planter being within -3.81 to 706.99 mcm of the corresponding 349 centre plant. In general, the plants planted by the centre and left planters were better aligned than 350 those for the centre and right planters. The misalignment between the plants planted by the left 351 and right planters was the most severe, with the left side lagging the right side by 2.7.9 mcm on 352 average. Visual observation during the year-1 trial suggested that a differential in the wheel slip 353 of the packing wheels due to irregularities in planting bed height, soil moisture and compressibility 354 may and the elasticity in the long shaft that connects them may have beenbe responsible for the 355 reduction in lack of synchronization. In the year-1 design, the two packing wheels used to power 356 the rotation of the planning wheel shaft were on the two extreme sides of the 3-row set, making 357 them 2.7.9 mcm apart. This large span may also be a source of the misalignment problem 358 observed in year 1. The year-1 results motivated the design change in year-2 to eliminate the 359 powered ground wheels and mechanically decouple the ground wheel from the planting wheels. 360 In figure 7 the top two histograms are from the on-campus trial of the year-1 prototype. The bottom 361 two histograms are from the year-2 trail in a commercial tomato field using the improved system. 362 Comentado [MPR1]: Reviewer 1: Why? Con formato: Fuente: 11 pto 20 21 22 Figure 7. Histograms showing the distributions of plant alignment, comparing the plant locations of the right and left sides to the centre. The top two histograms are from the on-campus trial of the year-1 prototype. The bottom two histograms are from the year-2 trail in a commercial tomato field using the improved system. The red dotted vertical lines are placed at the origin, as a reference. Ideally all plants would be close to this line. 363 A statistically significant ( α = 0.01) improvement was made observed in the planting pattern 364 synchronization in year-2 using a hydraulic motor to drive the motion of the planting wheels. The 365 distribution of the plant alignment between the three synchronized planters in year-2 is shown in 366 the lower two histograms in Figure 7, and a summary of the mean and standard deviation values 367 are shown in Table 2. The goal is to have most of the data concentrated near the origin of each 368 histogram, as indicated by the red vertical reference line. The mean error was -0.4.6 mcm 369 between the plants planted by the left-side planter and the centre planter in year-2. This value is 370 less than half the -1.3.2 mcm mean error observed in year-1. The mean error was also -0.4.6 371 mcm between the right and centre planters in year-2, which is approximately one-third of the 372 +1.4.7 mcm error observed in year-1. In sharp contrast to year-1, when the error between the two 373 outside planters was high, the two outside planters were, on average, very well synchronized, 374 with an error of only -0.03 mcm. The decrease in the total range of the synchronization error in 375 year-2, represents an improvement of 24% or more. Furthermore, the distributions from year-2 376 were more symmetric about the origin than those in year-1, which were skewed. These results 377 appear to show that the new hydraulic-powered drive system improved the planting 378 synchronizationprecision. We hypothesize that the -0.5 mcm average lag of the right and left 379 planters compared to the centre planter by -0.5 mcm on average, was due to a torque load on the 380 PTO synchronization shafts. Further study is required to investigate whether a design 381 improvement to the planter connecting linkages could further improve the level of synchronization. 382 In Italy, Mazzetto & Calcante (2011) developed a precision vineyard transplanter based on a 383 DGNSS-hydraulic moto that was able to obtain an inter-row average distance of 2.31 m (SD 0.036 384 23 m) and an average distance between plants along the row of 0.9 m (SD 0.041 m). However, this 385 study did not take into account the plant placement between different rows. 386 387 Table 2. Comparison of Plant Location Synchronization for Two Power Sources 388 Synchronization Error (mcm) Year Power Source Obs. Comparison Mean Std. Dev. 1 Ground Wheel 288 Centre vs. Left - 1.3.20a 2.3.90a 2 Hydraulic Motor 90 Centre vs. Left - 0.4.60b 2.4.10a 1 Ground Wheel 288 Centevs. Right 1.4.70a 3.3.30b 2 Hydraulic Motor 90 Centre vs. Right - 0.4.60b 2.2.90a 1 Ground Wheel 288 Right vs. Left - 2.7.90a 3.5.10b 2 Hydraulic Motor 90 Right vs. Left - 0.0.30b 2.6.40a *Treatments with different grouping letters are significant at the alpha = 0.01 level. 389 390 3.2 High-speed plant kinematic results 391 A total of 120 root bulb trajectories were recorded for dynamic motion analysis. Half of the 392 transplants studied corresponded to regularly sized seedlings, of which 30 used the support 393 treatment and 30 were of the unsupported control treatment. The other half of the transplants 394 were larger plants with the same numbers of digital videos analysed for each treatment. Figure 8 395 contains example digital images showing the typical dynamic path of the root ball centroid for the 396 two treatments. Table 3 shows the velocity and acceleration of the root bulb centroid during the 397 period of interest with planting wheel rotational speeds equivalent to forward travel speeds of 0.8 398 and 1.6 km h -1 . For both rotational speeds, the standard deviations of both the velocity and 399 Comentado [MPR2]: Reviewer 1: Surely the inter-row spacing is determined by either the fixed spacing between rows (if there is more t han one row unit to a toolbar) or the accuracy of GNSS guidance of the tractor, if there is only one row per pass. How does the planting mechanism driven by the DGNSS controlled hydraulic motor influence the between row spacing? Tabla con formato Con formato: Centrado Con formato: Izquierda Con formato: Centrado Con formato: Centrado Con formato: Centrado Con formato: Centrado Con formato: Centrado Con formato: Centrado Con formato: Centrado 24 acceleration were higher for the control than for the fully supported treatment. The low variation 400 in the velocity and acceleration for the supported plants may reduce the chance of the accidental 401 release of tomato plants while in transit between the time they are released by the operator and 402 the time they are released by the planter into the soil. In addition, although not quantified in this 403 study, the metal supports facilitate the ease with which the operator places the plant upon the 404 planting arm holder in the radial direction, potentially improving the precision of the depth of the 405 root ball in planting. 406 (a) (b) Figure 8. a) Example of root bulb trajectories from the unsupported control treatment and b) supported treatment 407 Analysis of variance indicated that there was a significant difference in the tangential speed 408 (p<0.01) and acceleration (p<0.01) for large plants at both a nominal tractor speed ofat a nominal 409 tractor speed of 0.8 km h -1 for large plants when comparing the performance using the metal 410 supports and the control system. A nominal tractor speed ofand 1.6 km 1.6 km h -1 when 411 comparing the performance using the metal supports and the control system. also yielded 412 significant differences in the tangential speed (p0.0) and acceleration (p0.0) values for the 413 different treatments (control vs. supported). In addition, the tangential speed (p=0.0794) and 414 25 acceleration (p=0.522) showed no significant differences for regular-size plants under different 415 treatments (control vs. supported) at a nominal speed tractor of 0.8 km h -1 . 416 Table 3. Mean and standard deviation of both the velocity and acceleration for large and 417 regular-size tomato plants. 418 Large Plants Regular size Plants Tangential VelocitySpeed (m s -1 ) Tangential Acceleration (m s -2 ) Tangential Velocity (m s -1 ) Tangential Acceleration (m s -2 ) Mean SD Mean SD Mean SD Mean SD Control (0.8 km h -1 ) 0.069 0.035a 0.970 0.824a 0.058 0.027a 0.755 0.687a Supported (0. 8 km h -1 ) 0.074 0.024b 0.902 0.694b 0.066 0.022a 0.822 0.652a Control (1.6 km h -1 ) 0.136 0.058c 1.442 1.159c 0.125 0.054b 1.149 1.000b Supported (1.6 km h -1 ) 0.150 0.035d 1.319 0.954d 0.149 0.037c 1.269 0.898b *Within each factor level, values followed by different letters are significantly different (p < 0.005 for both factors). 419 Figure 8 shows an example of the root bulb trajectories from the control and the supported 420 treatments obtained from the digital video analysis. In the control treatment (Figure 8a), the root 421 bulb tended to move in a more random path compared to that in the supported treatment (Figure 422 8b). This random motion begins from the point that the operator releases the tomato plant in the 423 holder and normally lasts until the plant is released into the soil. 424 The final portion of the plant trajectory, after the planting arm holder releases the plant, has a 425 crucial influence on the final soil location of plant. To synchronize the planting patterns between 426 adjacent tomato transplanters, it is essential to minimize the variability in the horizontal distance 427 between the release of the plant from the planting arm holder and the final root bulb position in 428 the soil. Figure 9 shows the box plots illustrating the distribution and mean of the distance between 429 the location at which the planting arm holder released the tomato plant and the final plant location 430 in the soil for the two treatments. The distance between the control (6.6.7±2.6.0 mcm) and 431 supported (8.2.8±1.4.5 mcm) treatments werewas significantly different (p=0.0098). The 432 variances were also significantly different between treatments (p=0.0408). An additional 433 laboratory test indicated that plant size did not significantly influence the planting location for the 434 32 In the future, this will allow, among other things, synchronized plant placement in adjacent tomato 505 rows to reduce plant-to-plant competition and facilitate the use of an automatic intra-row weeding 506 co-robot system based on an accurate odometry sensing technique. 507 The study provided new insights into plant dynamics during the transplanting process. The high-508 speed video analysis led to the design and testing of an improved plant support mechanism for 509 the improved control and precision of the transplanting of vegetable crops. The feasibility of 510 synchronization for adjacent transplanting units with vegetable crops was demonstrated. These 511 results serve as the fundamental basis for a mechatronic system that can precisely plant 512 transplanted vegetable crops in a grid-like pattern across rows as a critical first step in a 513 systematic approach to fully automated individual plant care. In the future, this new transplanter 514 design will allow, among other things, synchronized plant placement in adjacent tomato row to 515 reduce plant-to-plant competition and facilitate the use of an automatic intra-row weeding co-robot 516 system based on an accurate odometry sensing technique. 517 Acknowledgements 518 The authors would like to express their gratitude to California Tomato Research Institute 519 (CTRI) to support in part this work and to Scott Park of Park Farming Organics for allowing 520 our team to conduct research on his farm. The authors thank Burt Vannucci, Chris Gliever, 521 Loan-anh Nguyen, Garry Pearson, Jim Jackson, and Mir Shafii of UC Davis, and Claes 522 Jansson and Tord Holmqvist at SWEMEC in Woodland, CA for technical assistance. 523 References 524 Barbieri, P., Echarte, L., Della Maggiora, A., Sadras, V.O., Echeverria, H., & Andrade, F.H. (2012). 525 Maize Evapotranspiration and Water-Use Efficiency in Response to Row Spacing. 526 Agronomy Journal, 104, 939-944. https://doi.org/10.2134/agronj2012.0014 527 Con formato: Fuente: Cursiva 33 Bell, C.E. (1995). Broccoli (Brassica oleracea var. botrytis) yield loss from Italian ryegrass 528 (Lolium perenne) interference. Weed Science , 43, 117-120. 529 Blas, A., Barreireo, P., Hernández, C.G., Dias, G., & Griepentrog, H.W. (2013). Even-sowing 530 pattern strategies for a low-input organic system in forage maize. Agricultural Engineering 531 International: CIGR Journal , 15 (4). 532 Doerge, T., Hall, T., & Gardner, D. (2002). New research confirms benefits of improved plant 533 spacing sensing in corn. Crop Insights, 12 (2) Pioneer Hi-Bred Int´l. 534 Ehsani M. R., Upadhyaya, S.K., & Mattson, M.L. (2004). Seed location mapping using RTK-535 GPS. Transactions of the ASAE , 47, 909–914. 536 Forcella, F. (2012). Air-propelled abrasive grit for postemergence in-row weed control in field corn. 537 Weed Technology , 26, 161–164. doi: 10.1614/WT-D-11-00051.1 538 Fennimore, S. A., & K. Umeda. (2003). Weed control in glyphosate-tolerant lettuce (Lactuca 539 sativa). Weed Technology , 17, 738–746. 540 Gu, X. & Rosenbaum, PR. (1993). Comparison of multivariate matching methods: structures, 541 distances, and algorithms. Journal of Computational and Graphical Statistics , 2 , 405-420. 542 Han, L., Kumi, F., Mao, H., & Hu, J. (2019). Design and test of a multi-pin flexible seedling pick-543 up gripper for automatic transplanting. Applied Engineering in Agriculture, 35(6), 949-957. 544 Heege, H.J. (2013). Precision in crop Farming: Site Specific Concepts and Sensing Methods: 545 Applications and results, Site-Specific Sowing. New York: Springer. 546 Haar, M. J., & S. A. Fennimore. (2003). Evaluation of integrated practices for common purslane 547 management in lettuce. Weed Technology 17, 229-233. 548 Harriott, B.L. (1970). A packaged environment system for precision planting. Transactions of the 549 ASAE ., 13 (5), 550-553. (doi: 10.13031/2013.38660). 550 Con formato: Fuente: Cursiva Con formato: Fuente: Cursiva Con formato: Justificado Con formato: Fuente: Cursiva Con formato: Fuente: Negrita Con formato: Fuente: Cursiva Con formato: Inglés (Estados Unidos) Con formato: Inglés (Estados Unidos) Con formato: Fuente: Cursiva Con formato: Fuente: Cursiva Con formato: Fuente: Cursiva 34 Klenin, N.I., Popov, I.F. & Sakun, V.A. (1985). Agricultural machines. USDA and the National 551 Science Foundation, Washington, D.C., by Amerind Publishing Co. Pvt. Ltd., New Delhi. 552 Lati, R.N., Siemenens, M.C., Rachuy, J.S. & Fennimore, S.A. (2015). Transplanted lettuce with 553 an intelligent cultivator. Weed Technology, 30(3), 655-663 554 Mazzeto, F. & Calcante, A. (2011). Highly automated vine cutting transplanter based on DGNSS-555 RTK technology integrated with hydraulic devices. Computers and Electronics in 556 Agriculture 79 , 20-29.Miyao, G., Aegerter, B., Sumner, D. & Stewart, D. (2017). Sample 557 Costs to Produce Processing Tomatoes. University of California Cooperative Extension. 558 Available at: https://coststudyfiles.ucdavis.edu/uploads/cs_public/d7/b2/d7b2ee5f-8961-559 417d-b9d1-216d41fb47d8/2017processtomssacvalfinaldraft32817.pdf. Accessed 26 560 August 2020. 561 Nakarmi, A.D. & Tang, L. (2012). Automatic inter-plant spacing sensing at early growth stages 562 using a 3D vision sensor. Computers and Electronics in Agriculture 82 , 23-31. 563 Nørremark, M., Griepentrog, H.W., Nielsen, J., & Søgaard. H. T., (2008). The development and 564 assessment of the accuracy of an autonomous GPS-based system for intra-row 565 mechanical weed control in row crops. Biosystems Engineering , 101, 396-410. 566 Nielsen, R.L. (1991). Stand establishment variability in corn. Purdue University, Agronomy 567 Department publication AGRY-91-01. Available online at 568 http://www.agry.purdue.edu/ext/corn/pubs/agry9101.htm (URL verified July 27, 2014) 569 Nielsen, R.L. (2006). Effect of plant spacing variability on corn gfrain yield, Pudue University. 570 Agronomy Department. Available online at 571 http://www.agry.purdue.edu/ext/corn/research/psv/Report2005.pdf (URL verified July 27, 572 2014). 573 Ngouajio. M., McGiffen, M.E., & Hembree KJ. (2001) Tolerance of tomato cultivars to velvetleaf 574 interference. Weed Science, 49, 91-98. 575 Con formato: Español (España) Con formato: Español (España) Con formato: Fuente: Sin Cursiva Con formato: articles, Sangría: Izquierda: 0 cm, Sangría francesa: 1,27 cm, Espacio Antes: 3 pto Con formato: Fuente: 11 pto, Cursiva, Inglés (Estados Unidos) Con formato: Fuente: Cursiva 35 Pearson Autoplanter, (2020). http://www.richardpearson.com/pearson_autoplanter.html (URL 576 verified October 22, 2020) 577 Perez-Ruiz, M., Slaughter, D.C., Gliever, C., & Upadhyaya, S.K. (2012). Tractor-based Real-time 578 Kinematic-Global Positioning System (RTK-GPS) guidance system for geospatial 579 mapping of row crop transplant. Biosystems Engineering, 111(1), 64-71 580 Perez-Ruiz, M., Slaughter, D.C. Fathallah, F.A., Gliever, C.J., & Miller, B.J. (2014). Co-robotic 581 intra-row weed control system. Biosystems Engineering, 126 , 45-55 582 Poll, B. 1956. Plant-Carrying arm for a transplanter, US Patent 2.739.548 A 583 Parish, R.L. (2005). Current developments in seeders and transplanter for vegetable crops. 584 HortTechnology 15 (2), 346-351. 585 Prasanna Kumar, G.V. & Raheman, H. (2008). Vegetable transplanters for use in developing 586 countries- A review. International Journal of Vegetable Science, 14 (3), 232-255. 587 Raja R., Slaughter D.C., Fennimore S.A., Nguyen, T.T., Vuong V.L., Sinha N., Tourte L., Smith 588 R.F. &and Siemens M.C. (2019). Crop signaling: A novel crop recognition technique for 589 robotic weed control. Biosystems Engineering, 187, 278-291. 590 Rotty, R. (1960). Methods and machines used in North American nurseries. Unasylva 14 (2), 22 591 May 2007. <http://www.fao.org/3/x5394e04.htm>. 592 Sun, H, Slaughter, D.C., Perez-Ruiz, M., Gliever, C., Upadhyaya, S.K. & Smith. R.F. (2010). RTK 593 GPS mapping of transplanted row crops. 2010. Computers and Electronics in Agriculture , 594 71, 32-37. 595 Ulloa, S.M., Datta, A. & Knezevic, S.Z. (2010). Tolerance of selected weed species to broadcast 596 flaming at different growth stages. Crop Prot. 29:1381–1388. doi: 597 10.1016/j.cropro.2010.04.009 598 Con formato: Fuente: Sin Cursiva Con formato: Fuente: Cursiva Con formato: Fuente: Cursiva Con formato: Fuente: 11 pto Con formato: Izquierda Con formato: Fuente: 11 pto Con formato: Fuente: 11 pto Con formato: Fuente: 11 pto Con formato: Fuente: 11 pto Con formato: Fuente: 11 pto Con formato: Fuente: 11 pto Con formato: Fuente: 11 pto Con formato: Fuente: 11 pto Con formato: Fuente: 11 pto Con formato: Fuente: 11 pto Con formato: Fuente: 11 pto, Inglés (Estados Unidos) 36 Upadhyaya, M.K., Blackshaw, R.E. (2007). Non-chemical weed management: synopsis, 599 integration and the future. M.K. Upadhyaya, R.E. Blackshaw (Eds.), Non-chemical Weed 600 management, CAB International, Oxfordshire, UK (2007), pp. 201-209 601 602 Van Evert, F.K., Samson, J., Polder, G., Vijn, M., Van Dooren, H., Lamaker, A., Van Der Heijden 603 GWAM, Van der Zalm, T., & Lotz, L.A. (2011). A robot to detect and control broad-leaved 604 dock (Rumex obtusifolius L.) in grassland. Journal of Field Robotics, 28, 264-277. doi: 605 doi:10.1002/rob.20377 606 607 608 609 Visser Group, Holland., (2014). PC-11-2 transplanting machine. Available at 610 https://www.visser.eu/transplanting-made-simple/transplant/ (URL verified October 19, 611 2020) 612 Wicks, G.A., D.N. Johnston, D.S. Nuland & E.J. Kinbacher. (1973). Competition between annual 613 weeds and sweet Spanish onions. Weed Science , 21, 436-439. 614 Zhang, Z., Cao, W., Wang, Q., & Zhang, P. (2013). Development status and prospect of plug 615 seedlings automatic transplanting machine. Journal of Agric Mech Res 5(5), 237-241. 616 Zhao Z. (2004). Using matching to estimate treatment effects: Data requirements, matching 617 metrics, and monte carlo evidence. Review of Economics and Statistics, 86 (1), 91–107. 618 Zong, Z., Liu, G., _& Zhao, S. (, 2020). Real-time localization approach for maize cores at 619 seedling stage based on machine vision. Agronomy , 10 (4), pp 470. 620 https://doi.org/10.3390/agronomy10040470 621 622 Con formato: Justificado Con formato: Fuente: Cursiva Con formato: Espacio Antes: 3 pto Con formato: Sangría: Izquierda: 0 cm, Primera línea: 0 cm Con formato: Sangría: Izquierda: 0 cm, Primera línea: 0 cm, Espacio Antes: 0 pto Con formato: Fuente: Cursiva Con formato: Fuente: Cursiva Con formato: Fuente: Cursiva 37 623 Figure Captions 624 625 Figure 1. Photograph showing the PTO-style synchronization shafts (black with yellow safety 626 end covers) connecting the three precision planting wheels. 627 Figure 2. A planting wheel with the splined, PTO-style, index shaft used for synchronization. 628 Figure 3. View inside the year 1 synchronized transplanter showing the splined PTO-style 629 connecting shafts used to synchronized the rotation of the planting wheels. The red outer ground 630 wheels supplied power. 631 Figure 4. (a) The PTO-style synchronization shafts connecting the three planting wheels and (b) 632 the ground –wheels speed sensor and feedback control system for maintaining the desired plant 633 spacing. 634 Figure 5. (a) Conventional 5-arm planting wheel designed to allow mechanical linkage to connect 635 adjacent planters and (b) Experimental 5-arm planting wheel with the metal spoon supports 636 mounted. 637 Figure 6. Photograph showing the position of the root ball relative to the fingers of the 638 transplanter. The root ball (held inside the person's hand in the foreground) is typically 639 cantilevered 75 mm past the end of the finger. 640 Figure 7. Histograms showing the distributions of plant alignment, comparing the plant locations 641 of the right and left sides to the centre. The dotted vertical lines are placed at the origin, as a 642 reference. Ideally all plants would be close to this line. 643 Figure 8. a) Example of root bulb trajectories from the unsupported control treatment and b) 644 supported treatment. 645 Figure 9 . Box plots showing the mean and spread of the distance between the location at which 646 the planting arm holder released the tomato plant and the final plant location in the soil for the two 647 treatments. 648 Figure 10 . Histograms showing the distributions of plant alignment, comparing the plant locations 649 of the right and left to the centre for unsupported (control) and supported treatments on left and 650 right planter. 651 652