DGPS-Based Automatic Vehicle Guidance
Nagarajan Ramalingam, Timothy S. Stombaugh, and Jonathan Mirgeaux
Ramalingam, N., T. S. Stombaugh, J. Mirgeaux. 2000. DGPS-Based Automatic Vehicle Guidance. ASAE Paper No. 001068. Annual International Meeting, Midwest Express Center, Milwaukee, Wisconsin, July 9-12.
ABSTRACT
Contemporary automatic guidance research is mainly focused on using expensive real-time kinematic GPS or carrier phase differential GPS. In this paper, an approach for guiding a vehicle using a Differential Global Positioning System (DGPS) based position sensor as the only external posture sensor is highlighted. The guidance system consists of the DGPS-based sensor, a wheel angle sensor (potentiometer), and a pulse-width-modulated electrohydraulic valve. The platform of this research is a John Deere model 6210 agricultural tractor. Results are presented to show system performance.
Navigation of agricultural vehicles is a broad topic, covering a large spectrum of different technologies, as well as some of the most advanced space science and engineering. The navigation of an agricultural vehicle is the answer to the questions “where am I?” and “where am I going?” There are different navigation techniques; each technique has its own advantages and disadvantages, which are limited by the application requirements as well as the environment where the method is going to be used.
There are several benefits of automatic guidance in agriculture. It is the nature of many field operations to include parallel swathing or switchback operations (Grovum and Zoerb, 1970). These repetitive patterns can become somewhat monotonous for the operator, who often must drive continuously for many hours. The increasing size and complexity of modern farm equipment, including variable rate technology, requires more attention from the operator to monitor systems. Automatic guidance technology could guide a machine more accurately than the operator, especially with larger equipment. Improved seed placement when planting and precise chemical placement when spraying would lead to reduced input costs, improved crop productivity, and reduced environmental impact. It could be foreseen that automatic guidance could be developed to a point where the operator is not required at all. This would eliminate some of the dangers of farming, such as tractor overturns and contact with dangerous chemicals.
In parallel swathing field operations, it is imperative that subsequent application passes across a field be separated by the precise treatment width of the machine. If parallel paths are too close, overlapping coverage will cause double application in parts of the field. If parallel paths are too far apart, a part of the field will receive no treatment. Nieminen and Sampo (1993) showed that operators tend to overlap swaths to prevent the more noticeable effects of skips in coverage. Their studies indicated that overlap with larger equipment could be as great as 10% of the effective coverage width of the machine.
The requirement of detecting the position of agricultural vehicles during a field operation is necessary to develop automatically guided agricultural vehicles. The Global Positioning System (GPS) is one of the positioning tools used in precision agriculture. There are numerous parallel applications of GPS in agriculture. One among them is the guidance system using light bar activated by Differential GPS (DGPS). This system was initially built for aerial applications - for guiding an airplane along parallel swaths. Nowadays it is also used for parallel and contour swathing in ground applications. Most DGPS-based guidance aids provide feedback to the operator through a horizontal row of indicator lights or light bar. As the path of the machine deviates from the desired path, the light bar indicates the magnitude of the deviation, and the operator can steer the machine to correct the path.
One drawback to light bar guidance aids is that the operator must still be in full control of the vehicle steering function. Matsuo et al. (1998) showed that an automated vehicle steering system could provide more precise control of the path of agricultural vehicles than is achievable by human operators. In addition, automated guidance systems will improve the working environment for the machine operator.
Steering a vehicle along parallel swaths for long periods of time can become monotonous and contribute to fatigue. An automated guidance system would relieve the operator from the repetitively monotonous steering tasks and allow him/her to devote more attention to other important tasks.
It is anticipated that an automated guidance system could also impact other agricultural operations. For example, controlling the traffic pattern of subsequent field operations can reduce the effects of soil compaction. Also, some field operations could be performed at night with an automatic guidance system that functions adequately in low light conditions.
A vehicle guidance system can be conceptualized as shown in Figure 1. The posture of a vehicle is defined as the position, orientation, and motion of that vehicle relative to a reference frame. A guidance system determines the current position and orientation of the vehicle, compares it to a desired posture, and makes appropriate steering control to direct the vehicle toward the desired posture. When a human operator is controlling the vehicle, he/she senses the posture of the vehicle through visual cues and physical motion sensation, decides what steering actuation is necessary to correct the vehicle path, and then executes the corrective action by turning the steering wheel. An automated guidance system attempts to relieve the operator from many of these tasks. It would have to perform the sensing, decision-making, and actuation functions that the human normally performs.
One of the major hindrances to adoption of automatic guidance technology for agricultural field machinery has been the cost and complexity of the sensors used to measure vehicle position. Most previous GPS-based automatic guidance research has utilized Real-Time Kinematic (RTK) GPS systems (Stombaugh et al., 1999). Many RTK GPS systems will provide dynamic position accuracy of 2 – 20 cm. However, RTK GPS systems are prohibitively expensive and they require the user to set up and maintain a base station within relatively close proximity to the field.
A low cost alternative to RTK GPS would be desirable and would heighten adoption of automatic guidance technology in agricultural practice. This project explored the use of DGPS for autonomous guidance. DGPS receivers are much less expensive, and many producers already own a receiver that they use in other precision agriculture applications. Though inadequate for precise planting or cultivating field operations, the typical 1 m accuracy of DGPS receivers should be adequate for moderately precise field operations such as chemical application.
The goal of this project was to explore the use of DGPS technology for applications in automatic guidance of agricultural equipment. The following specific objectives were accomplished.
1. Develop a DGPS-based navigation system that can be used in an automated guidance system for an agricultural vehicle.
2. Install the guidance controller on a specific agricultural vehicle that will serve as the test bed.
3. Evaluate the performance of DGPS guidance accuracy, convergence and stability.
Satloc, Inc., an industry partner for this project, provided an model SLX[1] DGPS receiver and light bar for this research. The light bar computer was specially equipped with a Digital-to-Analog Converter. A proprietary algorithm developed by Satloc provided an analog signal through the DAC that was proportional to a predicted steering wheel angle. The GPS antenna was mounted on the foremost position on top of the hood of the tractor for early tests. Later tests evaluated system performance with the antenna mounted on top of the tractor cab.
Test Bed Equipment
The test vehicle, in order to be equipped with automated steering, had to have hydraulic power steering, either hydrostatic or metering pump. The vehicle that was used as a test bed for this project was a John Deere model 6210 Tractor (Figure 2). A proportional electrohydraulic valve was installed on the tractor steering system to facilitate automated actuation of the steering wheels. The valve was plumbed in parallel with the existing steering hand pump (Figure 3). Solenoid on-off valves were installed between the electrohydraulic valve and each of the 2 steering cylinders to insure no leakage flow during manual steering. When activated, the electrohydraulic valves would over-ride the manual steering, but would not interfere with manual steering when not activated. A PWM driver for the electrohydraulic valve converted an analog DC voltage to an appropriate pulse-width-modulated signal to activate valve coils. Steering wheel angle was sensed using a 10 K rotary potentiometer arrangement, which was fastened to the kingpin of the tractor left front wheel (Figure 4). The sensor was excited with a 5 V DC signal.
A 133 MHz Pentium-based field computer with data acquisition hardware was installed in the tractor cab to collect sensor information and provide the guidance control signal. All software was developed using a Visual BASICâ compiler.
The researchers in this project initially tried to reproduce previous guidance strategies employed on a different test bed (Stombaugh et al., 1999). That previous system made use of a proportional electrohydraulic valve in feedback control of steering angle. Use of the proportional valve in the current project proved unfruitful because of severe asymmetric and drift characteristics of the valve.
The innovative approach taken by the researchers utilized a type of “bang-bang” control. When a steering correction was initiated, the electrohydraulic valve was turned fully on, held on for a period of time proportional to the steering command (typically less than 200 ms) then turned back off. The flow rate of oil to the valve was restricted to provide stable control. Though a proportional valve was used in these tests because of previous installation, the same control could have been accomplished with a much less expensive solenoid-operated spool valve.
The sampling frequency used in the feedback control algorithm was 3 Hz. Since this sampling rate was low, data were naturally filtered. The digital feedback control equation used was
u(k) = K(ae(k) – be(k-1) + ce(k-2)). (1)
where K is the gain, a, b, and c are feedback coefficients, e(k) is the error at the current step, e(k-1) is the error at the previous step, and e(k-2) is the error at two steps previous. The values for a, b, and c that gave relatively fast convergence, minimum overshoot and maximum stability were 1.0, -0.14, and –0.95, respectively.
Field tests of system performance were conducted on an asphalt surface on the campus of the University of Kentucky. Tests were designed to evaluate the straight-line following ability of the system as well as convergence ability from different initial offsets. The path error (received from light bar), steering wheel angle (received from potentiometer) and steering command (sent to the valve) were recorded at 3 Hz. GPS data were recorded every second. The steering angle was not symmetric. The steering wheel angle output varied from 2720 (for full left), 3136 (for straight-line), and 3471 (for full right). The DAQ board had a 12 bit resolution. The steering command output from the controller varied from –5 V (for full left), 0 V (for straight-line), and +5 V (for full right).
Straight-line following accuracy:
The straight-line tests involved activating the automatic system while the tractor was moving along the intended path in the intended direction. For these tests, the tractor was operated at 5.0 mph (2.22 m/s). The tractor was able to follow the prescribed straight-line, with a maximum deviation of 0.6 m (2 ft) from steady state (Figure 5). The cross-track error is the deviation from the desired path. The steering command is the signal sent to the steering valve. The steering signal is the digital equivalent of the signal received from the proprietary algorithm developed by Satloc through the D/A converter. The reader should note that a positive cross-track error indicates a deviation to the left of the desired path, a positive steering command will steer the vehicle to the right, and an error signal greater than 2048 corresponds to a deviation to the right.
Convergence ability:
The convergence ability of the system was measured by activating the automatic system while the tractor was moving parallel to the desired path but offset by a given distance. The system would then steer the tractor toward the desired path. It was found that that system would converge from any angle and offset. Figure 6 shows the performance of the tractor when it was started at an offset of 5 m (16.67 ft) to the left of the prescribed path. Figure 7 shows the performance of the tractor when it was started at an offset of 5 m on the right of the prescribed path. Figure 8 shows the performance of the tractor when it was started at an offset of 10 m (33 ft) to the left of the prescribed path. Figure 9 shows the performance of the tractor when it was started at an offset of 10 m (33 ft) to the right of the prescribed path.
The system was also tested from an initial orientation perpendicular to the desired path. Figure 10 shows the performance of the tractor, when it was started at an offset of 23 m (77 ft) to the left of and perpendicular to the prescribed path. It is interesting to note that the overshoot is minimal, and the convergence is very fast. The direction in which the tractor approaches the straight line depends on the initial orientation of the tractor front wheels. Tests were made to confirm this, by placing the tractor’s front wheels slightly turned toward the left hand side and the right hand side. Figure 11 shows the performance for the wheels initial orientation toward the right.
It was also found that neither the position of the GPS receiver nor the tractor speed significantly affected the stability and convergence of the system. In the last test reported here, the receiver was placed on top of the cab and the tractor speed was set at 10 mph (4.44 m/s). The tractor was able to guide itself and converge on the straight line in less than 30 seconds (Figure 12). The speed of convergence depended on the speed of movement of the tractor. At higher speeds, the tractor converged faster, but within a similar distance.
In all of these tests, there was a steady state error in the performance and a noticeable slow oscillation. These were caused by non-linearity in the system, namely the steering actuator dead band. The minimum resolution on the software timer used to control the pulse width of the “bang-bang” control signal was 55ms. Consequently, signals that required a control pulse of less than 55 ms received no pulse, and the steering would not respond. Further refinement of timing routines and tuning of the feedback control should eliminate these errors.
An autonomous guidance system was successfully developed for automated steering control of agricultural tractors. This system was comprised of simple and inexpensive components. This guidance system uses DGPS alone as the only position sensor. The control algorithm was based on a “bang-bang” control, where pulses were used to control a solenoid valve. The pulse width was proportional to the steering command. This strategy could eliminate the need for expensive and unreliable proportional valves.
The guidance system was able to converge on to the prescribed path with minimal overshoot from any angle of approach and from any reasonable initial offset. Increases in forward speed and placement of the GPS antenna did not significantly affect system performance.
Grovum, M. A and G. C. Zoerb. 1970. An automatic guidance system for farm tractors. Transactions of the ASAE 13(5):565-573,576.
Matsuo, Y., D. Yukomoto, and N. Noguchi. 1998. Navigation systems and work performance of tillage robot. ASAE Paper No. 98-3192. St. Joseph, MI: ASAE.
Nieminen, T. and M. Sampo.
1993. Unmanned vehicles for agricultural and off-highway applications. SAE
Technical Paper Series No. 932475. Warrendale, PA: SAE.
Stombaugh, T. S., E. R. Benson, J. W. Hummel. 1999. Guidance control of agricultural vehicles at high field speeds. Transactions of the ASAE. 42(2):537-544.
[1] Mention of trade name, proprietary product, or specific equipment does not constitute a guarantee or warranty by The Ohio State University or the University of Kentucky and does not imply the approval of the named product to the exclusion other products that may be suitable.