6.9   Low Cost GPS-Based Sensors for High Speed Agricultural Vehicle Guidance

Principal Investigators

Timothy S. Stombaugh, Assistant Extension Professor, Biosystems and Agricultural Engineering
Larry G. Wells, Professor, Biosystems and Agricultural Engineering
Scott A. Shearer, Associate Professor, Biosystems and Agricultural Engineering

Introduction

Over the past century, many changes have occurred in agricultural field production machinery.  The need for greater productivity with fewer labor inputs has prompted development of larger and faster machines.  This trend has been most evident in chemical application equipment.  Twenty years ago, typical boom widths on liquid chemical application equipment were 6 to 15 m, and maximum operating speeds were 15 km/hr (Kepner et al., 1978).  Today, liquid chemical application machines can carry booms up to 40 m wide and will operate at speeds in excess of 35 km/hr.  The greater operator skill needed to control this larger and faster machinery coupled with related environmental and economic considerations have prompted a heightened interest in automated guidance technology.  The major hindrance to adoption of automated guidance technology has been the complexity and cost of vehicle posture sensors. 

The objectives of this project are:

1)                  To design a low-cost sensor array that can be used in an automated guidance system for high-speed chemical application equipment;

2)                  To design a feedback guidance controller for a chemical application machine, which utilizes the new sensor array and is capable of guiding the machine to within 1 m of a desired path; and

3)                  To test the new sensor array and guidance controller under typical field conditions.

Background

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.  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 would have to perform the sensing, decision making, and actuation functions that the human normally performs.

Posture Sensing - A key element of an automated guidance system is the sensor array that measures the posture of the vehicle relative to some coordinate frame.  Many different sensing technologies have been tested in automatic guidance applications.  One technology that is showing great promise for agricultural applications is the satellite-based global positioning system.

The global positioning system (GPS) is a network of satellites maintained by the Department of Defense that transmit radio signals which can be used to measure position.  GPS receivers are attractive for automatic guidance applications because they give geo-referenced position information that is repeatable from one field operation to the next.

A vast improvement in GPS accuracy is achieved with a Differential GPS or DGPS system (Hurn, 1993), which utilizes error information from a stationary receiver to improve the accuracy of the rover receiver.  There are several commercial and military sources for real-time differential correction information, which causes DGPS receivers to be simple to use.  Some DGPS systems are accurate to within 1 m of the actual position.

 The use of DGPS is becoming more prevalent in agriculture, particularly through Precision Agriculture management strategies (Borgelt et al., 1996).  Producers are utilizing DGPS information in conjunction with yield sensors on harvesting equipment and variable rate application equipment to quantify and manage the spatial variability in farm fields.

Recent automatic guidance research by the principle investigator and other researchers has utilized the most accurate GPS configuration known as Real Time Kinematic GPS (RTK GPS) for position sensing (Stombaugh; 1998; Elkaim et al., 1997).  RTK GPS is a differential GPS system, but the receivers also utilize the transmission carrier wave frequency and phase information to determine more accurate positions (Langley, 1998).  Some RTK GPS systems can achieve real-time kinematic accuracy of 2–20 cm.

Although RTK GPS technology provides the accuracy needed for posture sensing, RTK GPS receivers are expensive and difficult to use because the user is required to setup and maintain a second base station receiver and radio link to the rover receiver.  Therefore, a simple, low cost sensor package is needed to expedite the adoption of automatic guidance technology.  A sensor array utilizing a DGPS receiver that many agriculturists already use on precision agriculture equipment would further increase the marketability of automatic guidance systems.  Obviously DGPS receivers will not deliver adequate accuracy for precise agricultural operations such as planting.  However, given the fact that human operators can adequately guide broadcast chemical application vehicles by utilizing DGPS-based light bar feedback, a properly designed DGPS-based sensor array should permit automated guidance of similar vehicles.

Modeling and Guidance Control - After an appropriate posture sensor is chosen for a guidance system, the next major challenge is to develop steering control strategies to guide the vehicle along a desired path.  The investigators in this study will take a modeled approach to guidance controller design.  In a modeled approach, researchers quantify the dynamics of the vehicle, develop a feedback control strategy to adequately control the model during simulations, implement the controller on the vehicle hardware, and then fine-tune the controller for optimum performance.

Several researchers have developed model-based controllers for agricultural vehicles.  O'Connor et al. (1996) developed a tractor dynamics model using the assumption that the vehicle would not experience any sideslip during cornering maneuvers.  This assumption allowed them to base their model solely on the geometric properties of the vehicle while ignoring its inertial properties.  The consequence of this assumption was that system performance was limited to lower forward velocities.  As vehicle speed increased, the model did not describe the sideslip experienced by the vehicle, and the controller could not compensate adequately.  Later studies by Elkaim et al. (1997) used RTK GPS and Kalman filtering techniques to measure tractor dynamics while pulling implements, but again operating speeds did not exceed 3 m/s.

Since many agricultural operations, such as pesticide application, are conducted at higher forward velocities, the principle investigator has studied the higher speed dynamics of agricultural vehicles (Stombaugh, 1998).  Tests of a RTK GPS-based feedback guidance controller for an agricultural tractor demonstrated successful control of the vehicle at speeds up to 6.8 m/s (Stombaugh et al., 1999).  Those studies also revealed that the dynamics of the steering system of the tractor were critical to guidance controller design.

Procedures

The goal of the project is to develop a DGPS receiver-based sensor array and feedback control for automatic guidance of a commercial chemical application machine.  Satloc, Inc., a major GPS receiver manufacturer, has offered the use of DGPS receivers for this project.  An RTK GPS system will be purchased and used to record the actual motion of the vehicle during tests.

The first task will be to evaluate the steering decisions made by a human operator who is using a DGPS light bar guidance aid.  The goal of the analysis will be to determine what sensory information besides the visual light bar signal that the operator may be using to formulate a steering command decision  (Table 1).  Visual cues beyond the light bar could be used to determine the orientation (heading) of the vehicle as well as the forward velocity.  Motion sensations could be used to determine the angular or linear acceleration.  The operator can process combinations of any or all of this information to determine the best steering input.  The challenge for researchers is to determine which of these guidance parameters the operator considers when formulating a steering decisions.

Three drivers with previous experience using light bar guidance aids will be asked to guide the test vehicle along a predetermined path using the light bar feedback.  The path will be a linear path with an abrupt offset part way through.  This will allow researchers to record the operator’s response to step inputs.  The RTK GPS system will measure the actual path of the vehicle during the tests.  A laptop computer and associated data acquisition equipment will record the actual vehicle path, path indicated by the DGPS receiver, visual steering command observed by the operator, and the steering angle.

Table 1. Vehicle motion parameters with associated human and electronic sensing techniques.

Each of the vehicle motion parameters shown in Table 1 can be calculated from the precise RTK GPS data.  A Kalman filtering technique, similar to that used by Elkaim et al. (1997), will be used to analyze the data.  Essentially, mathematical formulations of various combinations of the sensory inputs will be sought to accurately describe the actual steering command issued by the operator.  The results of the analysis will show which motion parameters the operators considered in their steering decision and the relative importance of parameter.

Once the most critical sensory inputs are identified, appropriate sensors to measure those inputs will be specified and purchased (Table 1).  Electronic compasses can measure vehicle heading.  Radar and sonar sensors can measure vehicle velocity.  Inertial sensors can measure linear and angular acceleration (Will et al., 1998).  These extra sensors will be integrated with the DGPS information using a Kalman filter to form the posture sensor array (Noguchi et al., 1998; Abbott and Powell, 1999).

Before the feedback guidance controller (Figure 1) can be designed, the investigators will need to have a model of vehicle and steering actuator dynamics.  The principle investigator has already performed extensive studies of the high-speed dynamics of an agricultural tractor (Stombaugh, 1998).  His results revealed that, for the limited conditions tested, the guidance dynamics of an agricultural tractor could be quantified by the following transfer function:

        (1)

Equation 1 is similar to the common bicycle models used to describe automobile dynamics (Alleyne and DePoorter, 1997) which is of the form:

         (2)

Stombaugh et al. (1999) concluded that the higher frequency pole-zero pairs that are shown in equation 2 might exist in agricultural tractor dynamics, but because of speed limitations of their posture sensor, they could not quantify these dynamics.  Since the chemical application vehicle that will be used in this project contains an active suspension system similar to an automobile, the investigators expected that the high frequency pole-zero pairs may be at a lower frequency than a fixed suspension tractor, and therefore will be quantifiable.

An electrohydraulic valve will be installed in the steering system of the test vehicle.  Frequency response tests similar to those performed by the principle investigator (Stombaugh, 1998) will be use to quantify the dynamics of the spray vehicle and electrohydraulic steering system.

The feedback guidance controller (Figure 1) will then be designed.  Since the investigators do not expect that the final posture sensor design will measure all components of the vehicle motion, a partial observer will be used in the feedback control.  Matlab (Mathsoft) software will be used to perform digital simulations of the feedback control.  Controller parameters will be tuned to give a step response that is similar to the step response that the human operators achieved.

The new sensor array and automated control will then be installed on the test vehicle.  The Kalman integration of the sensor array as well as the guidance controller will be implemented using LabVIEW (National Instruments) instrumentation software.  The RTK GPS equipment will be used to measure the actual path of the vehicle during tests.  Performance of the guidance system will be evaluated using step responses.  Robustness of the automated guidance system to large initial offset and heading errors will also be evaluated.

Expected Benefits

The guidance system developed in this study will have direct benefits to Kentucky producers.  Since the system will be relatively inexpensive and will utilize existing DGPS equipment, more producers will be able to afford automated guidance technology.  In addition, the studies of human steering performance may lend new insights into ergonomic factors affecting driver performance and fatigue.

Deliverables

Results of this study will be disseminated through, conference presentations, refereed journal articles, and outreach and extension education activities.  Also, Satloc, Inc. is y interested in producing and marketing an automated guidance package that will utilize the posture sensor array developed in this project.