USD

Use Inertial Measurement Units to Enable Precision Agriculture

By Richard A. Quinnell

Contributed By Digi-Key's North American Editors

Modern agriculture increasingly employs sensing and location technology to increase the efficiency of field activity and to maximize crop yields by tracking local growing conditions and applying appropriate resources such as water, insecticide, and fertilizer as needed. Designers of systems for this application understand that satellite positioning has its limitations with respect to precision. However, applying inertial measurement units (IMUs) can fill the gap.

IMUs integrate three-axis accelerometers with three-axis gyroscopes to measure system motion and determine system positioning via dead reckoning. By combining these with Global Positioning System (GPS) information, designers can develop farm machinery control systems with precise, continual knowledge of equipment position relative to fields and crops, while correcting for factors such as terrain tilt, equipment arm movement, and other factors.

This article discusses the importance and role of IMUs in precision agriculture. It discusses potential error sources when using IMUs to perform dead reckoning, the mitigation of those errors, as well as environmental and safety factors developers should consider. Finally, it looks at precision IMUs from Honeywell Sensing and Productivity Solutions and Analog Devices and shows how they can be used to help boost precision beyond what satellite navigation systems alone can achieve.

Why location tracking is critical to agriculture

Traditional farming is a broad-brush process. Plowing, planting, watering, fertilizing, and harvesting are all done more or less uniformly across whole fields, often acres in extent, despite the inevitable variations in soil composition, evaporation, and the like within the field. Manual steering of machinery could result in missed or overlapping areas during these activities, reducing field utilization or wasting resources with redundant application. While a steering error of a foot or two between passes might not seem like much, the losses can accumulate significantly when crisscrossing a large field, adding to the time and fuel required (Figure 1).

Image of traditional farming treats whole fields as uniformFigure 1: Traditional farming treats whole fields as uniform and uses manual steering, both leading to wasted time and resources. Precision agriculture changes that. (Image source: John Deere®)

Accurate knowledge of location carries many benefits. It allows the collection of site-specific information on soil conditions across large areas, and a corresponding site-specific application of water, fertilizer, and pesticides to maximize yield. The greater the precision in location knowledge—ideally down to individual plants—the greater the benefit.

Precision agriculture has changed the way farmers work their land. The advent of satellite navigation technology has allowed farmers to accurately map the variations in growing conditions within fields and provide their farm machinery with real-time information on their location within that space. This combination of detailed mapping and precise location information is allowing farmers to prescribe and apply site-specific treatments of water, fertilizers, and pesticides to increase production, minimize waste, and reduce environmental impact.

Real-time location information also allows farmers to maximize field utilization by avoiding missed or overlapping planting and harvesting efforts while minimizing time and fuel usage through optimized travel. Such systems can also provide semi-autonomous piloting of farm machinery to reduce driver fatigue and allow efficient operation even in low visibility conditions such as dust, fog, rain, and darkness. Currently more than 50% of farmland, great and small in extent, now utilizes precision agriculture methods with adoption continually increasing.

Going beyond GPS

An ideal agricultural positioning system would be precise enough to reliably locate an individual plant or crop row within a field that might extend hundreds of acres—that is, offer precision on the order of a few inches. But there are limits to the positioning precision that satellite navigation systems alone can provide. Basic receivers for the U.S. GPS only provide a few meters of accuracy. Dual-channel GPS receivers or real-time kinematic (RTK) systems, which rebroadcast GPS signals from fixed stations, can achieve accuracies well below a meter (m). Even then, however, they are dependent on the accuracy of the information the satellites broadcast, which has typically yielded an average accuracy of around 0.7 m. Further complications to GPS-based location include the effects of reflections from, or signal blockage by, nearby objects and terrain, satellite constellation geometry, and time of day.

Satellite navigation has other limitations as well. The location that the system provides is simply a point—the phase center of the receiver’s antenna. GPS provides no information on orientation; for example, the facing direction can only be inferred by determining the direction vector between successive point locations. Similarly, GPS is insensitive to pure rotation, so it cannot determine, for instance, any tilt away from vertical GPS.

This antenna-centric location fix and insensitivity to rotation can create position errors in an agricultural application. A GPS-enabled tractor, for instance, might have its antenna on top of the driver’s cab, perhaps 10 feet off the ground, which is where the GPS fix will be centered. It would be reasonable to assume that the position on the ground for the tractor, or any attached equipment, could be reliably determined from the antenna position by simple geometry. The problem is, because the GPS system cannot determine orientation such as the tractor traversing a slope (Figure 2), the actual ground position will be offset from what rigid geometry would predict. Even a tilt as small as five degrees (°) will yield a ground position error of more than 10 inches (in.) in this instance.

Image of GPS cannot determine orientationFigure 2: GPS cannot determine orientation, so slopes could result in errors when determining the equipment’s actual ground position. (Image source: Richard A Quinnell)

One solution to these problems is to supplement the GPS navigation with inertial navigation using dead reckoning from sensors that measure system movement. Inertial dead reckoning can continue to provide accurate position information during times that GPS signals are weak or absent, while also providing a “reality check” on spurious results that might arise from multipath or other signal distortions. Further, inertial navigation sensors can fill in the orientation information that satellite navigation cannot provide. By simply measuring the direction of gravity’s pull, for instance, inertial sensors allow a system to correct for tilt errors in the GPS ground position determination, and increase operator safety by supporting roll-over warnings.

In practice, such inertial measurement units depend on two types of microelectromechanical systems (MEMS) sensors: accelerometers and gyroscopes. The accelerometers measure changes to linear motion along three orthogonal axes, and because gravity’s pull is an acceleration, can also reveal its direction. Gyroscopes measure angular motion (i.e., rotation) about each of the same three linear axes. Combined, the two measure changes in system motion along the six degrees of freedom (Figure 3).

Diagram of inertial navigation uses sensors to measure changes in motion along six degreesFigure 3: Inertial navigation uses sensors to measure changes in motion along six degrees of freedom—three linear and three angular—to support dead reckoning of position. (Image source: Honeywell Sensing and Productivity Solutions)

These inertial sensors do not directly reveal position, however. The accelerometers measure only the system’s surge, heave, and sway. These values must be integrated with respect to time to obtain system velocity and integrated again to obtain position. Similarly, gyroscopes measure roll, pitch, and yaw, which must be integrated with respect to time to obtain angular orientation.

These integrations may help to reduce the effects of random motion noise in the sensor measurements as such signals often tend to average out. But integration can compound the effects of some key systemic error sources inherent in the inertial sensors. Left uncorrected, these errors can accumulate to destroy the precision of the dead reckoning position, limiting the approach’s effectiveness in replacing lost GPS information. In general, the less error in the sensor measurements, the longer dead reckoning can provide position to the required accuracy.

Error sources in IMUs

Bias error: One of the key error sources in MEMS inertial sensors, for both accelerometers and gyroscopes, is the bias error. The bias error is the residual signal a sensor produces in the absence of rotation or linear acceleration. This error tends to be deterministic, unique to each individual device, and is often also a function of temperature. Integrating this signal over time can quickly build to unacceptable levels, but with proper calibration testing the bias errors of sensors can be determined and factored out of calculations.

Bias instability: Related to bias error, bias instability is the random change in a device’s bias error that occurs over time. This error source cannot be calibrated out, so developers must assess how great a change their design can tolerate and look for a sensor with a bias stability specification low enough to meet their needs.

Scale factor error: This is another of the deterministic errors found in inertial sensors. Scale factor, also called sensitivity, is the best-fit linear relationship mapping sensor input to output. The sensor’s scale factor error is its output’s deviation from that straight-line relationship, typically expressed as a percentage or in parts per million. This can also be temperature dependent and can be compensated for with proper calibration.

g sensitivity: An error source unique to gyroscopes is their sensitivity to linear acceleration, also known as g sensitivity (the g is from the abbreviation for gravitational acceleration, typically 9.8 meters per second squared (m/sec2)). This linear acceleration error can arise in MEMS gyroscopes as a result of asymmetry in their proof masses.

A MEMS gyroscope works by vibrating a test mass in one direction while sensing any motion in an orthogonal direction. While the sensor is rotating around an axis orthogonal to these other two directions, the Coriolis effect results in detectable sideways movement of the test mass.

Linear acceleration of the sensor orthogonal to the test mass vibration can also produce such sideways motion due to test mass inertia. The gyroscope’s sensitivity to this acceleration is a function of its design and fabrication accuracy. Using data from an independent accelerometer, however, allows a system to compensate for the error.

Vibration rectification error (VRE): This is another unique gyroscope error source and is also called g-squared error. It is the response of an accelerometer to ac vibrations that get rectified to dc, manifesting as an anomalous shift in the offset of the accelerometer. VRE can occur via several mechanisms and is not something that can be compensated for in real time as it is highly dependent on application specifics. Developers should determine if their sensor’s VRE is within acceptable limits. The use of vibration dampening sensor mounting techniques can help mitigate some vibration problems.

Cross-axis sensitivity: At the system level, mechanical misalignment of the sensors can also introduce errors. One such error is cross-axis sensitivity. This occurs when the actual sensing axis deviates from the intended direction, resulting in a signal from orthogonal motions that the sensor should not have been detecting. For example, a sensor that is intended to be horizontal may still detect the pull of gravity if it is misaligned. Misalignment between accelerometer and gyroscope axes can compromise system efforts to compensate for gyroscope g-sensitivity errors.

Off-axis errors: Mechanics also plays a part in generating off-axis errors in accelerometers. If the point of impact for a shock to the sensor is not centered on the accelerometer’s proof mass, the sensor can see an additional acceleration due to the slight rotation the proof mass makes around the line of impact.

Integrated IMUs ease sensor error issues

This multitude of error sources creates significant challenges for developers seeking to create an IMU from discrete sensors. Fortunately, pre-integrated IMUs with six degrees of freedom are widely available that simplify things considerably. Some of these are available in module form, such as the ADIS16465-3BMLZ precision IMU module from Analog Devices and Honeywell’s 6DF-1N6-C2-HWL (Figure 4). These allow developers to simply bolt them onto a chassis in order to include them in a system design.

Image of Honeywell 6DF-1N6-C2-HWL integrated IMUFigure 4: Integrated IMUs like the Honeywell 6DF-1N6-C2-HWL help simplify system design by eliminating alignment issues along with many other error sources. Board-mountable BGA IMUs are also available. (Image source: Honeywell Sensing and Productivity Solutions)

Precision IMUs are also available as chip-like, board-mountable devices, such as the ADIS16500/05/07 family from Analog Devices. These are suitable for incorporation with other sensors and GPS receivers into a unified assembly.

Both types of IMUs help ease development effort by eliminating or mitigating many of the potential errors in IMU development. The Analog Devices ADIS16500/05/07 family, for example, integrates a three-axis accelerometer with a three-axis gyroscope and a temperature sensor in a single BGA package. These devices have built-in calibration and filtering that combine with other features to help mitigate many IMU error sources (Figure 5).

Diagram of Analog Devices ADIS1650 integrated IMUFigure 5: An integrated IMU, like the Analog Devices ADIS16505 shown here, can help simplify system design by mitigating many potential error sources through on-board calibration, filtering, and alignment. (Image source: Analog Devices)

Errors such as cross-axis sensitivity are addressed in device fabrication. The ADIS16505, for instance, limits axis-to-axis alignment errors to less than 0.25°. This careful alignment, along with common clocking of the sensor readings, simplifies the designer’s use of accelerometer readings to correct for linear acceleration errors in the gyroscopes. The built-in temperature sensor supports efforts to mitigate the temperature dependency of many error sources.

The internal signal chain of these integrated IMUs provides additional error mitigation (Figure 6). Raw sensor information first passes through a digital filter to remove noise, it then passes through a user-configurable Bartlett Window Filter. The Bartlett Window is a finite impulse response (FIR) averaging filter using two cascaded stages.

Diagram of Analog Devices factory-determined calibration parametersFigure 6: Integrated IMU devices can offer built-in filtering and compensate for many systemic sensor errors by applying factory-determined calibration parameters. (Image source: Analog Devices)

Signals next pass through a calibration stage that applies device-specific corrections based on factory calibration tests run at multiple temperatures that span the device’s full operating temperature range. Using matrix multiplications on all six sensor samples simultaneously, this stage is able to compensate for bias, scale factor, and alignment errors in both the accelerometers and gyroscopes. It also corrects for linear acceleration errors in the gyroscopes and axis offset errors in the accelerometers.

A user-selectable point of percussion alignment correction is also available to adjust the accelerometer outputs to behave as though they were all located at the same reference point in the package. All other factory calibration features are inaccessible, but the devices do provide users the ability to adjust the factory’s sensor bias compensation with additional values of their own choosing.

Following the calibration corrections, the signals pass through a second digital filter. This decimation filter averages multiple samples together to produce the final output, providing additional noise reduction. The number of samples averaged together depends on the user’s choice of sampling and register updating frequencies.

System considerations

One of the few error sources that the integrated IMU cannot correct for is VRE. With agricultural machinery strong vibration is inevitable, so designers must carefully evaluate their system’s requirements on this issue. Many low-cost IMUs have very poor VRE; some with values so poor that vendors don’t bother specifying. To be fair, in the intended applications of these low-cost IMUs, VRE is not a significant issue. Devices intended for high-vibration environments such as precision agriculture, however, need to have VRE as low as possible. The ADIS16500 family, for instance, has a VRE on the order of 4 x 10-6 (°/sec)/(m/sec2)2. Thus, a sustained 1 g vibration (strong enough to bounce the driver off the seat) would only result in a rotational error of about one degree per hour.

Being free of mounting, alignment, and calibration issues is a major step toward obtaining a working system, but it’s only a beginning. Developers must still turn the inertial measurements into location tracking, resolve differences between the dead reckoning and the GPS location determinations, and understand and mitigate application-specific factors such as the amount and frequency of system shocks and vibrations during routine use.

If the location system is being used to provide autonomous or even semi-autonomous control of moving machinery, there are safety factors to consider as well. MEMS sensors can be overwhelmed by shocks of too high a magnitude. While devices are often able to survive large shocks without damage, a shock that drives a sensor past its limits could result in a temporary sensor shutdown or its output staying pinned to its maximum as it recovers. The system needs to be designed so that such momentary shocks do not inadvertently lead to dangerous or annoying system behaviors such as suddenly changing directions or falsely triggering a system safety shutdown.

A good way to start is with an evaluation board like the Analog Devices EVAL-ADIS2Z (Figure 7). This board gives developers PC-based access to device registers and data and is small enough to easily mount on representative target machinery to gather vibration and motion statistics.

Image of Analog Devices EVAL-ADIS2Z evaluation boardFigure 7: Boards such as the EVAL-ADIS2Z simplify the experimentation stage and are small enough to be mounted to the side of machinery for data gathering purposes. (Image source: Analog Devices)

The board supports application software for basic demonstration, individual register access, and high-speed data capture.

Conclusion

Precision agriculture based on satellite navigation is already providing farmers with enhanced productivity while lowering resource usage. By adding inertial positioning, designers can greatly improve the precision of locationing and help farmers achieve plant-level precision in the management of fields. To get there, however, developers will need to address sensor and systemic error sources in their designs. The availability of integrated, six-degrees-of-freedom precision inertial measurement units goes a long way toward easing that development burden by providing careful alignment, filtering, and built-in, calibrated error correction.

Disclaimer: The opinions, beliefs, and viewpoints expressed by the various authors and/or forum participants on this website do not necessarily reflect the opinions, beliefs, and viewpoints of Digi-Key Electronics or official policies of Digi-Key Electronics.

About this author

Richard A. Quinnell

Richard Quinnell has been an engineer and writer for 45 years, covering topics such as microcontrollers, embedded systems, and communications for a variety of publications. Prior to becoming a technical journalist he spent more than a decade as an embedded systems designer and engineering project manager for companies such as the Johns Hopkins University’s Applied Physics Laboratory (JHU/APL). He has degrees in electrical engineering and applied physics, with additional graduate work in communications, computer design, and quantum electronics.

About this publisher

Digi-Key's North American Editors