Computers and Electronics in Agriculture 54 (2006) 36–43

Feasibility of low-cost GPS receivers for ground speed measurement夽 Muharrem Keskin a , Sait M. Say b,∗ a

Department of Agricultural Machinery, Faculty of Agriculture, Mustafa Kemal University, 31040 Antakya, Hatay, Turkey b Department of Agricultural Machinery, Faculty of Agriculture, C ¸ ukurova University, 01330 Balcali, Adana, Turkey Received 11 April 2005; received in revised form 10 March 2006; accepted 15 July 2006

Abstract Ground speed measurement is required in many agricultural machinery operations. There are a number of techniques for the determination of the ground speed; however, each of these methods has some disadvantages. It is known that a GPS receiver is capable of calculating travel speed; however, no study has been reported on the performance of low-cost GPS receivers in measuring the ground speed in agricultural machinery operations. The objective of this work was to study the effectiveness of low-cost GPS receivers for measuring ground speed. Computer programs were written to read, store, and process the GPS data from two different low-cost GPS receivers. Differential correction was not used in the study since there was no correction service in the study area. A very strong relationship between the average GPS speeds and the average calculated speeds was found in the study (R2 > 0.99). Also, the results showed that the GPS receivers followed the speed change quite well. GPS receivers can be used to determine both the position and the speed of the agricultural machinery. In this way, there is no need to use a separate speed sensor leading to simpler and more affordable instrumentation systems. In conclusion, low-cost GPS receivers can be confidently used to measure the ground speed in agricultural machinery operations. © 2006 Elsevier B.V. All rights reserved. Keywords: Ground speed measurement; GPS; Agricultural machinery; Precision agriculture

1. Introduction Measurement of ground speed is a necessary task in some agricultural machinery applications. The true ground speed has to be known to properly change the application rate of the agricultural inputs such as seeds, pesticides, and fertilizers. Ground speed is also used in the evaluation of wheel slip and traction efficiency (Richardson et al., 1982; Khalilian et al., 1989). Beside these applications, the ground speed data are intensively used in precision agriculture technologies. For instance, in yield monitoring and mapping, yield is calculated from the amount of harvested material divided by the harvested area, which is a function of the harvester’s working width and ground speed. Therefore, a ground speed measurement sensor is one of the main components of a yield monitoring system (Keskin et al., 1999). The use of Global Positioning System (GPS) receivers in yield monitoring systems provides the opportunity of geo-referencing the yield data and creating a yield map of the field.

夽 ∗

Mention of specific products is for information only and not for exclusion of the others that may be suitable. Corresponding author. Tel.: +90 322 338 6195. E-mail addresses: [email protected] (M. Keskin), [email protected] (S.M. Say).

0168-1699/$ – see front matter © 2006 Elsevier B.V. All rights reserved. doi:10.1016/j.compag.2006.07.001

M. Keskin, S.M. Say / Computers and Electronics in Agriculture 54 (2006) 36–43

37

There are a number of sensors to measure the ground speed including rotational encoders on front, rear or fifth wheels, magnetic shaft encoders, radar, and ultrasonic sensors (Richardson et al., 1982; Grevis-James and Bloome, 1982; Clark and Adsit, 1985; Summers et al., 1986; Tompkins et al., 1988; Khalilian et al., 1989; Self, 1990; Freeland et al., 1990). Richardson et al. (1982) investigated several ground speed measurement techniques based on wheel rotation and single and dual beam radar. Tompkins et al. (1988) compared the performances of three ground speed sensors; fifth wheel, front wheel, and radar. Similarly, Khalilian et al. (1989) compared four ground speed sensors; fifth wheel, front wheel, ultrasonic, and radar in four levels of ground speed and five surface conditions. Previous studies reveal that these ground speed measurement techniques have some drawbacks. For example, the performances of radar and ultrasonic speed sensors are affected by the surface conditions (tall vegetation) and the motion of the machinery namely yaw, pitch, and roll (Richardson et al., 1982; Khalilian et al., 1989). Also, the signals from the radar and the ultrasonic speed sensors require filtering and averaging for appropriate speed determination (Richardson et al., 1982). In addition, Richardson et al. (1982) reported that the optical and acoustic sensors have problems of obstruction by dust, crop motion, and wind. The performances of front or fifth wheel-based sensors as well as the magnetic pickup-based methods are affected by wheel slip, requiring different calibration equations for different surfaces (Tompkins et al., 1988). It is generally known that a GPS receiver is capable of calculating the travel speed as well as determining the geographic location on the earth. The GPS determines the traveled distance from the instantaneous change in the receiver’s position in a specific time period and then calculates the speed from this data. However, no study has been reported on the performance of GPS receivers in measuring the ground speed in agricultural machinery operations. The GPS receiver is an important device to determine the geographical position of agricultural machinery. If GPS could be used to determine the speed of the machinery, then a separate sensor will not be necessary to measure the speed; since the GPS could provide both instantaneous position and travel speed. The GPS receiver is an important cost factor in precision agriculture applications. Previous studies showed that low-cost GPS receivers can be used in determining the position of the harvester (Shannon et al., 2001, 2002). Therefore, our objective was to study the effectiveness of the low-cost GPS receivers for measuring ground speed. 2. Materials 2.1. Study location The experiments were conducted on the Research and Experiment farm (37.04◦ N, 35.38◦ E) of the Faculty of Agriculture of C¸ukurova University located in Adana, Turkey. 2.2. GPS receivers Two different low-cost GPS receivers were used in the study. The first GPS receiver (GPS receiver #1) was a Magellan SporTrak (Thales Navigation, San Dimas, CA, USA) while the other receiver (GPS receiver #2) was a Garmin eTrex Vista (Garmin International, Inc., Olathe, KS, USA). The receivers were placed about 0.3 m apart from each other on top of an agricultural tractor (Fig. 1). According to the receivers’ specifications, both receivers have an update rate of one per second. Both receivers had 12-channel tracking and differential correction capability. However, the differential correction was not used in the study since there was no differential correction service in the region in which the study was conducted. In addition, both receivers had PC interface cable to allow data transfer to a computer through the serial communication (COM) port. A laptop computer was used to read and store the GPS data at each speed level. As the computer did not have serial communication ports, two COM-to-USB converters were used to create two virtual serial communication ports (COM1 and COM2). 2.3. Agricultural tractor A New Holland TD 95D four-wheel drive agricultural tractor (New Holland, Burlington, Iowa, USA) was used in the study. The tractor had 12 forward speed levels. The arrangement of the two GPS receivers and the laptop computer is presented in Fig. 1.

38

M. Keskin, S.M. Say / Computers and Electronics in Agriculture 54 (2006) 36–43

Fig. 1. The arrangement of the two GPS receivers and the laptop computer in the tractor.

2.4. Software Two computer programs were written in QuickBasic programming language (Version 4.5, Microsoft Corp., Redmond, WA, USA). The first program (GPS-2.bas) was written to simultaneously read and save raw GPS data from two different GPS receivers connected to the virtual COM1 and COM2 ports of the laptop computer. The program shows the raw GPS data from both COM ports on screen, and at the same time, it saves the raw data into two separate data files (one file for each receiver). The program uses the time (hhmm) when the program gets started as the base file name. In this way, the raw GPS data from the GPS receiver #1 is saved to the file named hhmm-1.txt as the data from the GPS receiver #2 is stored in the file named hhmm-2.txt. The aim of the second program (GPS-p.bas) was to process the raw GPS data files for each receiver to extract the speed data and then save the processed data to other data files (hhmm-1p.txt and hhmm-2p.txt). The flow charts of the two programs are given in Fig. 2. 3. Method A distance of 100 m was marked using flags on a straight stabilized farm road (not asphalt-paved or concrete). The GPS receivers were connected to the laptop computer (Fig. 1). At each of the 12 forward speed levels, the tractor was started to travel about 10 m behind the starting point. When the front wheel of the tractor was at the starting point, the computer program GPS-2.bas was started, and when the ending point was reached the program was stopped. At the same time, the travel time between the starting and ending points was measured using a stopwatch. This procedure was applied for all 12 speed levels. The engine RPM was set to 1500 and held constant in each measurement. The measurements were carried out twice during each run. The speed levels varied from approximately 0.3 to 4.4 m/s (1.1–15.8 km/h). The experiment was conducted on three different occasions (20 March 2005, 30 March 2005, and 6 April 2005). In each of the three occasions, the measurements were carried out during different times of the day. Statistical analysis was carried out to study the significance of the differences in the GPS speed data for the three different dates. The Analysis of Variance (ANOVA) was applied by using SPSS statistics software (Version 13, SPSS Inc., Chicago, IL, USA). In each measurement, the GPS receivers were set to send the raw data according to the NMEA (National Marine Electronics Association) 0183 standard with 4800 baud rate (Bennett, 2003). After the data collection procedure was completed, the raw data were processed using the program, GPS-p.bas, to extract the speed data from the raw data files. According to the NMEA 0183 standard, the speed data are available in the data sentence beginning with the $GPRMC data format; therefore, the program extracted the speed data from the data lines starting with $GPRMC (Fig. 2). The speed data are also available in $GPVTG format; however, it was observed that one of the GPS receivers (Magellan SporTrak) does not send data in $GPVTG data sentence while both of the receivers send data in $GPRMC format. Therefore, it was decided to use $GPRMC format to determine the

M. Keskin, S.M. Say / Computers and Electronics in Agriculture 54 (2006) 36–43

39

Fig. 2. Flow charts of the two programs written for the study: the first program collected and saved the raw GPS data from the two GPS receivers (left), and the second program extracted the speed data from the raw data file and saved it to another file (right).

speed data. In this format, the speed is sent in knots (nautical mile per hour) with one decimal point (0.1 knot or 0.05 m/s) resolution. The program, GPS-p.bas, saved the extracted speed data into another data file. Then, the speed data were transferred into spreadsheet software to calculate the average GPS speed. Finally, the two speed values (average calculated speeds and average GPS speeds) were graphed for each speed level. In addition, another experiment was carried out to investigate the capability of the travel speed change response of the GPS receivers. In this experiment, a field road with a length of 200 m was divided into four equal segments (50 m). Different speeds were used in each 50 m segment in an increasing order in a way that the speed was lowest in the first segment and then incremented in each succeeding segment. Eight speed levels, representing the range of speeds commonly used in agricultural operations (0.3–2.0 m/s), were used in two different runs. The four speed levels were between 0.3 and 1.0 m/s in the first run and between 0.6 and 2.0 m/s in the second

40

M. Keskin, S.M. Say / Computers and Electronics in Agriculture 54 (2006) 36–43

run. GPS data were collected in one data file while the travel time was separately recorded for each 50 m segment. After data processing, the instantaneous GPS speed data were plotted against the average calculated speed for each segment. 4. Results The relationship between the average GPS speed and the average calculated speed is presented in Fig. 3 for the data collected on 20 March 2005. Since the results were very similar for the data collected in the second and third experiments, only the results of the first experiment was given. It is seen from Fig. 3 that a very strong relationship exists between the average GPS speeds and the average calculated speeds (R2 > 0.99). The GPS average speed agrees well with the average calculated speed for each 50 m segment for both GPS receivers (Fig. 4), and when the speed of the tractor was increased at the beginning of a new 50 m road segment, the GPS receivers adapted to the new speed instantly. The data also show that there were some instantaneous peaks in the GPS speed data (Fig. 4). Two types of minor peaks and dips appear in the GPS speed data. The first type is the minor peaks and dips along each road segment. We were unable to determine the cause of these minor instant changes; however, we think they might be the result of noise in the GPS signal. The second type of peaks occurred when the tractor speed changed from a lower to a higher speed level (at the entrance to the next road segment). These minor peaks could have been caused by the motion of the tractor as the speed increased; i.e., when the gear was changed to a higher level to increase the speed, the tractor vibrated and wavered causing instantaneous speed changes at some of the transition areas (Fig. 4). The data from all three experiments (Trial I on 20 March 2005, Trial II on 30 March 2005, and Trial III on 6 April 2005) (Table 1) show that the calculated speeds and the GPS speeds were very similar within each trial and among

Fig. 3. The relationship between the average GPS speed and the average calculated speed for the two GPS receivers for the data collected on 20 March 2005.

M. Keskin, S.M. Say / Computers and Electronics in Agriculture 54 (2006) 36–43

41

Fig. 4. The GPS speed change versus the average calculated speed in each 50 m segment.

trials for both GPS receivers. According to the statistical analysis (ANOVA) results, the differences among the GPS speed values for the three experiment dates were not significant for either of the GPS receivers (P > 0.05). 5. Discussion The results of the study show that there is a very strong relationship between the average GPS speeds and the average calculated speeds (R2 > 0.99) (Fig. 3). Thus, low-cost GPS receivers can be used to measure ground speed in agricultural machinery operations. Furthermore, although we were unable to use differential correction signal in the study, the results were very promising. Currently, a separate speed sensor is used to determine the ground speed of agricultural machinery in some precision agriculture applications. Most precision agriculture systems inherently have a GPS receiver to determine the instantaneous position of the machinery. Since the results of this study shows that the ground speed can be confidently measured using a GPS receiver without differential correction, there may not be a need to use a separate speed sensor in the system. Adoption of this technique would reduce the cost of the machinery instrumentation. Although we did not have the opportunity to compare these results with those of other speed sensing techniques, the results of this study suggest that GPS-based speed measurement would be the best choice. Both GPS receivers provided speed data with an accuracy of 0.1 knot (0.05 m/s). Even if this precision is acceptable in agricultural machinery operations, more precision would be more convenient in some applications. The measurements were also repeated on three different dates, and statistical analysis revealed that the differences among the measurements of these three trials were not significant (P > 0.05). Easy usage, wide availability, and low cost of GPS receivers make their use very convenient for ground speed measurement in different applications of agricultural machinery. However, the instantaneous speed data are available after data transfer from the GPS receiver based on the NMEA standard to a computer, and a computer program is needed to extract the speed values from the whole data.

42

M. Keskin, S.M. Say / Computers and Electronics in Agriculture 54 (2006) 36–43

Table 1 Summary of the experimental data Trial number

Gear (speed) level

Calculated speed (m/s)

Receiver #1, average GPS speed (m/s)

Receiver #2, average GPS speed (m/s)

Rep1

Rep2

Rep1

Rep2

Rep1

Rep2

Trial I

1 2 3 4 5 6 7 8 9 10 11 12

0.29 0.43 0.61 0.88 0.67 1.00 1.42 2.03 1.58 2.32 3.28 4.40

0.29 0.42 0.62 0.86 0.69 0.99 1.40 2.02 1.59 2.32 3.23 4.44

0.30 0.42 0.62 0.88 0.67 0.99 1.43 2.00 1.59 2.31 3.21 4.40

0.30 0.41 0.62 0.87 0.68 0.99 1.40 2.02 1.60 2.29 3.17 4.41

0.30 0.42 0.62 0.88 0.67 0.97 1.42 2.02 1.58 2.32 3.29 4.48

0.30 0.41 0.62 0.87 0.67 0.99 1.40 2.02 1.59 2.31 3.23 4.44

Trial II

1 2 3 4 5 6 7 8 9 10 11 12

0.30 0.42 0.61 0.87 0.68 1.00 1.41 2.03 1.61 2.32 3.29 4.57

0.29 0.43 0.61 0.89 0.69 0.99 1.42 2.02 1.61 2.34 3.29 4.65

0.31 0.41 0.62 0.87 0.67 0.99 1.41 2.03 1.60 2.32 3.30 4.53

0.31 0.42 0.61 0.88 0.68 0.99 1.43 2.02 1.61 2.34 3.29 4.58

0.31 0.41 0.62 0.87 0.68 1.00 1.41 2.02 1.61 2.31 3.29 4.58

0.30 0.42 0.63 0.89 0.68 0.99 1.43 2.00 1.59 2.33 3.27 4.61

Trial III

1 2 3 4 5 6 7 8 9 10 11 12

0.29 0.43 0.60 0.88 0.67 0.99 1.41 2.01 1.59 2.32 3.26 4.52

0.29 0.42 0.61 0.88 0.69 1.00 1.45 2.06 1.61 2.31 3.31 4.61

0.30 0.42 0.61 0.88 0.67 0.99 1.43 2.01 1.59 2.33 3.25 4.53

0.31 0.41 0.62 0.88 0.68 1.01 1.45 2.05 1.61 2.32 3.30 4.55

0.27 0.41 0.61 0.88 0.67 0.98 1.43 2.01 1.59 2.33 3.25 4.53

0.30 0.41 0.62 0.88 0.68 1.01 1.45 2.05 1.61 2.31 3.31 4.64

6. Conclusions A very strong relationship between the average GPS speeds and the average calculated speeds was found in the study (R2 > 0.99). Also, results showed that the GPS receivers responded to speed change quite well. The differences among the GPS speed values for the three experiment dates were not significant for either GPS receiver (P > 0.05). In precision agricultural application systems, a GPS receiver can be used for determining both the position and the ground speed of the agricultural machinery. This approach makes a separate speed sensor unnecessary and makes precision agricultural machinery systems simpler and less expensive. In conclusion, low-cost GPS receivers can be confidently used to measure the ground speed in agricultural machinery operations. However, the speed output from a low-cost GPS receiver is only available through NMEA data and needs to be extracted using a computer program.

M. Keskin, S.M. Say / Computers and Electronics in Agriculture 54 (2006) 36–43

43

Acknowledgements ¨ uz, and Mr. Fevzi S¸ahbaz for their support throughout the ¨ u, Dr. Abdullah Oks¨ The authors thank Dr. Mustafa Unl¨ study. References Bennett, P., 2003. The NMEA FAQ, Version 6.4 (http://vancouver-webpages.com/pub/peter/nmeafaq.txt). Clark, R.L., Adsit, A.H., 1985. Microcomputer based instrumentation system to measure tractor field performance. Trans. ASAE 28, 393–396. Freeland, R.S., Tompkins, F.D., Hart, W.E., Wilkerson, J.B., Wilhelm, L.R., 1990. Tractor-mounted instrumentation for monitoring power unit and implement performance parameters. In: Tompkins, F.D., (Ed.), Instrumentation for In-Field Measurement of Agricultural Tractor Operating Parameters and Energy Inputs to Tractor-Powered Implements. University of Tennessee Southern Cooperative Series Bulletin 344, pp. 45–56. Grevis-James, I.W., Bloome, P.D., 1982. A tractor power monitor. Trans. ASAE 25, 595–597. Keskin, M., Han, Y.J., Dodd, R.B., 1999. A review of yield monitoring instrumentation applied to combine harvesters for precision agriculture purposes. In: Proceedings of the Seventh International Congress on Agricultural Mechanization and Energy, 26–27 May, Adana, Turkey, Department of Agricultural Machinery, Faculty of Agriculture, C¸ukurova University, pp. 426–431. Khalilian, A., Hale, S.A., Hood, C.H., Garner, T.H., Dodd, R.B., 1989. Comparison of four ground speed measurement techniques. Paper No. 89-1040, ASAE, St. Joseph, MI. Richardson, N.A., Lanning, R.L., Kopp, K.A., Carnegie, E.J., 1982. True ground speed measurement techniques. SAE Paper No. 821058, SAE, Warrendale, PA. Self, K.P., 1990. Instrumentation for determining tractor and implement performance. In: Tompkins, F.D., (Ed.), Instrumentation for In-Field Measurement of Agricultural Tractor Operating Parameters and Energy Inputs to Tractor-Powered Implements. University of Tennessee Southern Cooperative Series Bulletin 344, pp. 13–22. Shannon, K., Brumett, J., Ellis, C., Hoette, G., 2001. Can a $300 GPS receiver be used for yield mapping? ASAE Paper No. 01-1154, ASAE, St. Joseph, MI. Shannon, K., Ellis, C., Hoette, G., 2002. Performance of low-cost GPS receivers for yield mapping. ASAE Paper No. 02-1151, ASAE, St. Joseph, MI. Summers, J.D., Batchelder, D.G., Lambert, B.W., 1986. Second generation tractor performance monitor. Appl. Eng. Agric. 2, 30–32. Tompkins, F.D., Hart, W.E., Freeland, R.S., Wilkerson, J.B., Wilhelm, L.R., 1988. Comparison of tractor ground speed measurement techniques. Trans. ASAE 31, 369–374.