Traction Prediction:

Tillage Studies:

Precision Farming in Vegetable Production:

Figure 4

Figure 5

An Ultra-Precise, GPS Based Planter:

a) Specific Aims: The long-term objective of this research is to develop an ultra-precision planter using a centimeter accuracy GPS device. Such precision in seed placement makes it possible to develop an accurate seed map that can be used to target chemical application and mechanical cultivation thus reducing the cost of production and protecting the environment. Moreover, such seed maps can assist in autonomous vehicle guidance that should reduce operator fatigue. The specific objectives of this research are:
(1) To retrofit a Salvo 650, 4 row, vacuum planter with centimeter accuracy GPS unit and to map the seed delivery.
(2) To compare the actual plant map to the seed delivery map to determine the error in seed placement due to seed dynamics, planter dynamics, seed-planter interaction, and hitch alignment problems.
(3) To model the seed dynamics, planter dynamics, and effect of hitching error on seed placement accuracy.
(4) To develop necessary modifications to the seed delivery mechanism and automatic compensations to hitching inaccuracies to achieve agreement between the seed delivery map and the actual plant map.
(5) To explore the feasibility (accuracy and efficiency) of plant-specific chemical application using seed maps developed by the centimeter accuracy GPS system.
We have addressed objectives # 1 and 2 successfully. We did not face serious problems due to seed dynamics, planter dynamics, and hitching errors with the four row planter we employed in this study. Therefore, objectives # 3 and 4 were not serious concerns. However, detection of seed in the field, effect of dust particles on the sensors which detect seed (optical sensors), inaccuracies of the radar ground speed sensor at low operating speed, and issues related to RTK GPS signal quality and its effect on seed placement accuracy turned out to be serious issues. In essence addressing these issues (seed detection error, effect of dust particles, radar problems, and RTK GPS quality) replaced objective #3 and overcoming these problems replaced objective #4. We feel that we have addressed each one of these problems successfully and achieved a respectable accuracy of about 1.2 to 1.5 in. between plant map and seed map. Because of the amount of time spent on dealing with modified objective #3 and #4, we did not have sufficient time to address objective #5. Now that we have a reliable planter, we plan to continue and address objective #5 as a part of MS thesis of Mr. Mark L. Mattson. Dr. David Slaughter’s work on robotic weeding and Dr. Ken Giles work on smart spraying systems will be a very good starting points for this part of the study. We are also going to explore nonchemical techniques such as zapping the weeds with a high voltage device. Therefore the modified objectives that have been successfully addressed are:
(1) To retrofit a Salvo 650, 4 row, vacuum planter with centimeter accuracy GPS unit and to map the seed delivery.
(2) To compare the actual plant map to the seed delivery map to determine the accuracy of the system.
(3) Determine the sources of errors that influence of the accuracy of the seed map in predicting the plant map.
(4) Develop techniques to minimize the effect of the sources of error determined in objective #3 on the accuracy of seed map in predicting plant map.
b) Project Accomplishments:
Planter instrumentation: A four row Salvo 650 vacuum planter which was loaned to this project by Solex Corporation, Dixon, CA was retrofitted with a complete set of 4700 series centimeter accuracy surveying and mapping system consisting of a base unit, a rover, radio link, and other accessories donated by Trimble Navigation Ltd. Two single board data loggers were used to acquire data in real-time and display it in the tractor cab. The first data logger was interfaced to the centimeter accuracy GPS unit, the optical sensors, an encoder, and an accelerometer. The second data logger was interfaced to a display unit mounted in the cab. The two data loggers and the GPS receiver were mounted in a rugged, weatherproof metal box and secured to the planter frame. The accelerometer and GPS antenna were installed on the top of the toolbar frame. The GPS antenna was directly ahead of planter unit#1. The accelerometer was included to monitor the planter dynamics since it was thought to be a critical issue in determining seed dynamics when we started this project. However, it turned out to be not as important. Four optical sensors (one per planter unit) were mounted directly above planter shoes and detected the seeds as they fell through seed tubes. Originally a radar was included to measure planter travel speed. However, we found that the radar speed sensor was not accurate enough for our purpose especially at low speed. We replaced the radar with a wheel encoder that was capable of producing 256 pulses per revolution of the wheel. The encoder was mounted on the planter frame and was connected to the left planter wheel to measure its rotation. Figure 2 is a schematic diagram of this system. The first data logger obtained the GPS time and UTM co-ordinates every second and stored them with a reference time (time-tag). This data logger also monitored the optical sensors, time-tagged the seed events, and stored the information in the memory. A flow chart of the data acquisition program is given in appendix A. We have also included the actual C program used for a cquiring data in real-time in this appendix. The second data logger monitored the first data logger and reported planter’s performance through the display unit mounted in the tractor cab.
Setting up the Base Station: A permanent base station was installed in the field. It is a punch mark on top of a machined metal rod embedded in concrete located near experimental plots (Fig.3). A surveyed point at the Davis airport was used to determine the exact location of the field base station.
Figure 3. Field base station and punch mark.
Field Experiments: Preliminary tests indicated that all of the sensors and the GPS unit were working. Following the preliminary tests, field tests were conducted over two planting seasons to check the performance of the plant mapping system. In the first year, four rows were planted during each pass through the 165-ft. test plot. Corn seed (hybrid: Pioneer 3162 R24) was planted at a depth of two inches and an operating speed of 2 mph. Figure 1 on page 1 shows the planter during the field trials. In each row 15-ft sections were marked consecutively. Three of the eleven 15-ft sections were selected at random for analysis. The sections were measured by the Trimble RTK surveying equipment (Fig.4). The distance between the plants was also measured with a tape measure in order to check the accuracy of the Trimble surveying equipment. A computer program was written to process the seed and GPS location data and estimate the exact coordinates of each seed and its distance from the adjacent seed. When we examined the RTK GPS position, we found that they did not all lie on a straight line even when the operator drove "straight" through the field. We felt that there were some along track and cross track errors. Our algorithm minimized the cross track error by regressing a line through the RTK GPS points and then dropping a perpendicular from observed RTK GPS points to predict the corrected positions. We felt that this approach improved our seed map accuracy. During the first year trails we achieved a seed map accuracy of 1.8 to 2.1 inches with respect to the actual plant map. We also felt that there was a need to correct along track error. We hypothesized that if we knew the true ground speed of the planter, then we may be able to minimize the along track error using statistical techniques. We felt that the radar gun was not providing an accurate enough speed measurement, especially at low ground speed. We also wondered whether the RTK GPS could provide an accurate and independent measure of ground speed. Our investigation revealed that the velocity value of the "VTG" string outputted from the RTK GPS did not have enough accuracy at low ground speeds. Finally we decided to replace the radar with a wheel encoder as mentioned earlier. However, subsequent trials in the second year indicated that perhaps our "straight" operation of the tractor was not straight enough at the centimeter level and indeed RTK GPS was recording positions very accurately. Figure 5 shows the accuracy obtained by the RTK GPS during one of our tests. The RMS error value of the RTK GPS during this test was 8.7 mm.
Figure 5. Accuracy of the RTK GPS system used in this study.
The important thing we learned was that RTK GPS quality was very important. As long as RTK GPS quality was "3" we were guaranteed centimeter accuracy. We developed a new algorithm that utilized only quality "3" GPS signals and continuously performed instantaneous ground wheel calibration which could be used for prediction purposes if a quality "3" GPS signal was unavailable. A mathematical procedure which utilizes three consecutive quality 3 GPS points to develop an interpolation technique is presented in Appendix B. The interpolation technique provides the antenna location in UTM coordinates. To obtain the exact location of the seeds it was necessary to implement an Eluerian rotation scheme. Appendix B outlines the coordinate transformation technique used to obtain seed location in UTM coordinates. A Visual basic program that implements this technique is also included in appendix B.
This modified scheme was tested during the second year of this study. Some of the early plantings indicated that we were not receiving quality 3 signal most of the time. It turned out that the base station antenna was not high enough. After consulting with Dr. Art Lange, Trimble Navigation Ltd., we constructed a tall tower (about 12 ft tall) to mount the antenna near the base station (Fig.6). Elevating the antenna at the base station virtually eliminated problems with RTK GPS signal quality.
Figure 6. Radio antenna for the base station mounted on an elevated tower to ensure a "good quality" RTK GPS signal.
The subsequent plantings during Summer 2000 showed that the error in seed map compared to plant map was about 1.2 to 1.5 in. This accuracy is very good for the purpose of implementing plant-specific cultivation.
c) Specific Aims Addressed: During the two years of this study we addressed specific objectives #1 and #2 as stated in the original proposal and modified objectives 3 and 4 successfully.
The main modifications done for and during the 2000 planting season included the installation of an encoder for determination of ground speed, new seed sensor amplifiers, adjustments to the data collection program that allowed storage of additional data from the GPS string, and the construction of a tower to elevate the base station radio that sends corrections to the rover.
Following these modifications we found that seed map accuracy in comparison with the plant map was in the range of 1.4 to 1.9 in. The sensitivity of optical sensors was found to be the major cause for higher errors (about 1.9 in.) in rows 3 and 4. Figure 7 shows the orientation of the sensor that detects the seeds as they fall to the ground. The optical passage confirmation sensors are designed to detect any material that passes through the 21 x 21-mm (0.83 in) window of the sensor. The sensor circuit was modified in order to change the range in which seed particles were detected. The modification allowed us to adjust the sensor amplification so the sensors would see objects that were the size of a corn seed but not see objects such as small rocks and dust particles. These modifications helped to improve the seed map accuracy in comparison to plant map to about 1.2 to 1.5 in.
Figure 7. Runner-opener layout showing sensor location.
Results: A summary of the results of the study can be seen in figure 8. In evaluating the performance of the planter two distinct cases are of interest:

i. A seed was seen by the optical sensor, but no plant ever germinated in the vicinity of that site. This could be due to over sensitivity of the optical sensor, seed viability, and birds and/or rodents eating the seed;
ii. A seed was not seen at a location but a plant did germinate there. This is due to the error in detecting seeds.

It is preferable to have a small amount error of the type listed under (i) above. This type of error, if not excessive, provides us a certain degree of protection against mistaking a plant as a weed and killing it. However, we want the second type of error to be negligible. A plant seen at a location near which no seeds were seen would be treated as a weed and eliminated. So we used second criterion as a measure of our planter performance. In other words, we looked for the closest seed to a germinated plant and used this distance as a measure of our accuracy - the smaller this distance the is better the accuracy.
In the first year of testing we were able to get our seed detection accuracy to a range of 1.7 to 2.1 inches. The results of the second year showed that, the accuracy of this planter was very good (1.2 to 1.5 in). This makes it possible to produce plant maps for use in subsequent cultivation.
Figure 8. Comparison of the average error for all rows during plantings on 8/20/99, 9/7/00 and 9/11/00.
d) Significance of the Results: The first major accomplishment of our study was the successful development of a centimeter accuracy planter. We have been able to integrate a four-row Salvo 650 vacuum planter with centimeter accuracy GPS system and use it to collect data for plant map generation. A computer program was developed to convert the raw data from the planter into plant maps. Modifications were made to the planter to deal with problems such as dust, sensor sensitivity, speed measurement, and GPS position quality. By successfully creating plant maps the door is open to continue research on plant specific weed control which can reduce the amount of pesticide applied. It is believed that plant-specific cultivation can reduce pesticide application by 24 to 51% thus saving at least $40 to $50/acre (Ehsani, et al., 2000). This approach not only reduces the chemical cost, it also protects the environment by reducing herbicide load.
e) Interaction with the Private Sponsor: In addition to the numerous e-mail and phone conversations, Dr. Upadhyaya visited Trimble Navigation Limited on 2/11/1998 and received equipment worth over $50,000 (a centimeter accuracy GPS rover, a centimeter accuracy base station, radio link, and RTK software). Dr. James Janky, Vice president Intellectual properties, presented this equipment to Dr. Upadhyaya. Mr. Russ Keller, Products Manager was also present on this occasion.
9/18/98 - 9/91/98: Mr. Russ Keller and Tim Funk, Technical support and training, visited UC Davis to train our research team on the use of centimeter accuracy GPS equipment.
On 5/6/99 Dr. Art Lange, Director of Product Development and Marketing, Mr. Russ Keller, and Mr. Glen Martin, Technical support and training, from Trimble Navigation Limited visited us to look at the GPS based planter developed at U.C. Davis. A travel writer, Michael Baublitz, accompanied them. They were very pleased with the way the planter performed.
On 6/8/99 Dr. Upadhyaya met with Mr. Russ Keller in Davis and discussed the project progress.
Mr. Michael Baulitz, a writer assigned to this project by Trimble Navigation Ltd. to document the major advances in the project has visited us several times since April 1999.
On 9/14/99 Dr. Upadhyaya, Reza Ehsani, and Mark Mattson visited Trimble Navigation Ltd. in Sunnyvale to report on the progress of the project.
Contact from 6/00 to 9/00 with Art Lange about the accuracy of VTG velocity and the factors affecting GPS quality.
On 9/11/00 Drs. Roz Buick, Production Marketing; Art Lange, Precision Agriculture; and Greg May, Product Marketing from Trimble Navigation Ltd., visited U.C. Davis for a demonstration of the planter and to discuss the progress of the project.
Communication has been going on with Trimble Navigation Ltd. about continuing to make progress on the patent procedure.

Ehsani, M. R., M. Mattson, and S. K. Upadhyaya. 2000. An Ultra-Precise, GPS Based Planter for Site-Specific Cultivation and Plant Specific Chemical Application, Paper No. 003065. Presented at the 2000 ASAE Meeting in Milwaukee, Wisconsin.
Ehsani, M. R., M. Mattson, and S. K. Upadhyaya. 2000. An Ultra-Precise, GPS Based Planter for Site-Specific Cultivation and Plant Specific Chemical Application, Fifth International conference on Precision Agriculture. Minneapolis, Minnesota (In Press).

RTK GPS Basic Autoguidance System:

Ivestigators: Shrini K.Upadhyaya, Davis J. Hills, and Davis S. Slaughter
Supporting Personnel: Brian C. Heidman, Zine El Abidine Abdelaziz, and Uriel A. Rosa
Abstract:
This project explores the benefits of RTK GPS based autoguidance systems (centimeter accuracy) in agricultural production. This guidance system allows for automatic steering of the tractor and tillage equipment on specific traffic paths during cultural operations and makes it possible to use subsurface drip irrigation systems in row crops. With autoguidance, the tractor can be steered close to the drip tape and/or plants without damaging either. Subsurface drip irrigation has the potential for improving water use efficiency, reducing weed growth, and improving overall energy efficiency. The ability to steer a tractor automatically and accurately close to the drip-tape or plants allows for high operational ground speeds. An autoguidance system also eliminates guess rows and makes it possible to enhance productivity by increasing the number of beds per unit farm area. This UC Davis project has been designed around a 2k factorial experiment, using four replicates for documenting reduction in energy usage, enhancing water use efficiency, and improving the timeliness of field operations. Project results will be analyzed and made available to farmers interested in using this cutting-edge technology for improving the efficiency of production agriculture.
Acknowledgements:
We are grateful to the California Energy Commission for their financial support of this project. We are also grateful to Trimble, Inc. for donating the autopilot system and 4700 RTK GPS hardware for conducting these tests. Furthermore, we appreciate the help of Button and Turkovich Ranch in loaning transplanting equipment during the course of this project.
Progress:
Processing tomato seedlings have been transplanted in the 5.5 acre plot using a Trimble autopilot equipped John Deere 7800 tractor. The plot consists of four blocks, each of which contains 12 1,000 ft rows of tomatoes. The listing, bed shaping, and transplanting operations were accomplished using the autopilot system. During the transplanting operation, drip tape was installed six inches to one side of each transplant row, at a depth of five inches. The location of each transplant was recorded using the Trimble 4700 RTK GPS system to develop a plant map and to determine how well the implement steers behind the tractor. Several cultural operations will be conducted in these four blocks, using two different forward speeds and two distinct implement locations with respect to plants to determine the benefits of using autopilot systems in terms of increasing productivity, saving energy, reducing plant damage, and minimizing drip tape damage.

Photos of Autoguidance system at work:
Figure 1. Bed shaping

(1a)

(1b)

(1c)
Figure 2. Transplanting with an autopilot system

(2a)

(2b)

(2c)

(2d)
Figure 3. Transplanting Mechanism

Figure 4. Drip tape installation system

(4a)

(4b)
After the completion of the tomato transplanting experiment, a second experiment was conducted in which tomatoes were planted using a vacuum planter mounted on a tractor equipped with the auto-guidance system (Figure 5a). After the seeds germinated, the filed was cultivated at two different speeds and spacing to evaluate the effectiveness of the auto-guidance vehicle in tomato production systems (Figure5b). The field data have been collected for both the transplanting and seed planting experiments and are currently being analyzed.

Figure 5a. RTK GPS based auto-guidance system is being used to plant tomato seeds using a vacuum planter on the same beds in which tomatoes were transplanted previously.

Figure 5b. The experimental plot which was planted with a RTK GPS based auto-guidance system was cultivated using the same auto-guidance system which left behind a very narrow (4 in. wide) untilled strip around the plant line.

Other Research Activities:

Several other projects related to agricultural machinery design, development, and testing have been conducted. Development of a hydro-pneumatic planter, a crust breaker, a device for measuring soil crust strength, etc. are a few of the recently completed projects. More information about these projects can be found in the papers included in the publication list or by directly contacting the author.

---

This page is ©Kishan Web Design 1997.