Note: Descriptions are shown in the official language in which they were submitted.
CA 02848150 2014-03-07
WO 2013/033753
PCT/AU2012/001007
1
METHOD OF RECALIBRATING INERTIAL SENSORS
FIELD OF THE INVENTION
The invention relates to a method of recalibrating sensors. In particular
the invention relates, but is not limited, to recalibrating inertial sensors
by
taking sensor readings when the sensor is stationary.
= BACKGROUND TO THE INVENTION
Reference to background art herein is not to be construed as an
admission that such art constitutes common general knowledge in Australia or
elsewhere.
Inertial sensors are used in many applications to measure movement
of objects. For example, vehicles, such aeroplanes and automated vehicles,
. and many electronic devices, such as smart phones, have inertial
sensors to
determine orientation, movement, and/or other relevant variables.
Inertial sensors typically include gyroscopes, which measure the rate of
change of angle, and accelerometers, which measure linear acceleration.
Often such sensors are collectively packaged into an inertial measurement
unit (IMU). A typical IMU will contain at least a three-axis accelerometer,
and
often includes one or more gyroscopes. IMUs sometimes also contain a 2 or
3 axis magnetometer for sensing the Earth's magnetic field (although not
actually an inertial sensor).
Inertial sensing is often used to determine an 'attitude' of an object or a
vehicle (i.e. the rotation of object or vehicle with respect to a reference
frame,
CA 02848150 2014-03-07
WO 2013/033753 =
PCT/AU2012/001007
2
usually a theoretical perfectly level ground surface). In many applications,
accurate inertial sensing is critical. For example, in precision agriculture,
knowledge of 'attitude' of a vehicle is required to compensate for movements
of a Global Navigation Satellite Systems (GNSS) antenna through terrain level
changes and undulation.
In machine control applications, such as autonomous vehicles, sensor
precision is often high enough that an offset induced by the tilting of a GNSS
antenna mounted on a vehicle can produce a measurable positioning error
(e.g. of at least the same order of magnitude as the GNSS system itself). As
a result, tilt angle is sometimes compensated with the use of angular
estimates derived from sensor measurements produced by an IMU mounted
in the vehicle.
For many inertial sensors, notably industrial grade inertial sensors often
used in machine control applications, there are error characteristics, notably
sensor bias, which change with temperature and age. These errors affect
system accuracy and typically require the sensors to be sent back to the
manufacturer for recalibration periodically (e.g. once per year). Such
recalibration is costly and time consuming as it not only requires the device
to
be removed, but also requires the device to be returned to the manufacturer
for a period of time, resulting in significant down-time.
Furthermore, even a yearly calibration can be insufficient in minimising
bias as ambient temperature fluctuates over a year and, accordingly,
temperature errors arise when the sensor is used in a different temperature
range to what it was calibrated for. For example, if the sensor is calibrated
in
CA 02848150 2014-03-07
WO 2013/033753
PCT/AU2012/001007
3
summer, the temperature errors will likely become prevalent in winter when
the ambient temperature is lower.
If the user does not send the device back to the manufacturer for
factory calibration in an effort to avoid the costs and downtime then, in
addition to the temperature error, age induced errors will also arise meaning
that the device will lose accuracy over time.
One approach to assisting with keeping the sensors calibrated,
particularly for temperature induced bias, is to add temperature sensing
components and a sensor bias model to estimate the sensor bias at
measured temperatures. However, this increases the cost and complexity of
devices that use the sensors. Furthermore, calibration using such models
often only includes temperature variation of the inertial senor over a limited
temperature range. The model must also be updated as the inertial sensor
ages to account for age induced bias. Updating the model is commonly done
by yearly factory calibration or by calibration using additional sensors.
These
strategies add further cost and complexity to recalibrating the sensors.
Recalibration may be conducted by taking measurements at different
attitudes and determining the bias from the measurements, but this requires a
user to move the sensors to different attitudes between each measurement,
which is an inconvenience to the user. = =
CA 02848150 2014-03-07
WO 2013/033753
PCT/AU2012/001007
4
OBJECT OF THE INVENTION
It is an aim of this invention to provide a method of recalibrating
sensors which overcomes or ameliorates one or more of the disadvantages or
problems described above, or which at least provides a useful alternative.
Other preferred objects of the present invention will become apparent
from the following description.
SUMMARY OF INVENTION
According to a first aspect of the invention, there is provided a method
of determining an inertial sensor bias, the method including the steps of:
using the inertial sensor to determine when the inertial sensor is
stationary;
automatically obtaining a first inertial sensor measurement when the
sensor is determined to be stationary; =
automatically obtaining a second inertial sensor measurement when
the sensor is determined to be stationary and at a different attitude to the
first
inertial sensor measurement;
automatically obtaining a third inertial sensor measurement when the
sensor is determined to be stationary and at a different attitude to the first
and
second inertial sensor measurements; and
determining the inertial sensor bias using the first, second, and third
. inertial sensor measurements.
Preferably the method further comprises storing a plurality of obtained
inertial sensor measurements. Preferably the step of determining the inertial
CA 02848150 2014-03-07
WO 2013/033753
PCT/AU2012/001007
sensor bias uses at least a portion of the plurality of stored inertial sensor
measurements. Preferably only inertial sensor measurements within a
predetermined period of time are used to determine the inertial sensor bias.
Sensor measurements may be weighted during determination of the inertial
5 sensor bias according to how recently they were obtained. Preferably
older
measurements are given less weight than more recent measurements.
Preferably, a weighted least squares model may be utilised to determine the
inertial sensor bias using weighted inertial sensor measurements.
Preferably the step of determining when a sensor is stationary includes
determining a period of no movement of the sensor and determining whether
the sensor is under, or at least close to, normal operating conditions.
Preferably the temperature of the sensor is considered when obtaining the
inertial sensor measurement. The temperature of the sensor may be stored
with the obtained inertial sensor measurement. Where the temperature of the
sensor is stored with the inertial sensor measurement, a temperature bias of
the sensor may be estimated to provide a more accurate inertial sensor
measurement and/or a more accurate determination of the inertial sensor
bias.
Preferably the step of determining when the sensor is stationary
includes determining a period of no movement for a predetermined length of
time. Preferably the step of automatically obtaining an inertial sensor
measurement includes processing data received from the inertial sensor.
Preferably the processing includes using signal _processing to account for
external factors such as, for example, removal of vibration. Preferably the
=
CA 02848150 2014-03-07
WO 2013/033753
PCT/AU2012/001007
6
step of obtaining the inertial sensor measurement, including processing data
received from the inertial sensor, takes between approximately 10 seconds
and 2 minutes, even more preferably between approximately 20 seconds and
90 seconds.
Preferably the method further comprises determining when movement
of the sensor resumes. Preferably any inertial sensor measurement that is
interrupted by resumption of movement of the sensor is discarded.
'
Preferably the inertial sensor is an accelerometer or a gyroscope.
Preferably the inertial sensor measurements are obtained from an inertial
= measurement unit (IMU) containing the accelerometer and/or gyroscope.
Preferably the accelerometer is at least a three-axis accelerometer.
Preferably the inertial =sensor measurements are accelerometer
measurements that consist of a measurement of gravity only. In an
alternative form, the inertial sensor measurements are gyroscope
measurements that consist of a measurement of the rotation rate of the Earth
only.
The step of determining the inertial sensor bias using obtained inertial
sensor measurements may use more than three inertial= sensor
measurements. Preferably each of the inertial sensor measurements are at
different attitudes to each other. The method may include measuring the
change in inertial sensors between the measurements.
Preferably the step of determining the sensor bias using the first,
second, and third inertial sensor measurements includes the steps of
CA 02848150 2014-03-07
WO 2013/033753
PCT/AU2012/001007
7
considering the three inertial sensor measurements, determining possible bias
values for each measurement, and/or determining an intersection of the bias
values.
Preferably the step of determining possible bias values includes
determining a 'sphere' of possible bias values and preferably the step of
determining the intersection of the bias values includes determining an
intersection of the three spheres. Preferably determining an intersection of
the three spheres includes utilising linear algebra. Preferably, the step of
determining possible bias values includes determining an estimate of the bias
values. =The estimate of the bias values is preferably determined using least
squares.
= According to another aspect of the invention there is provided a
method of calibrating an inertial sensor, the method including the steps of:
determining a sensor bias according to the aforementioned method;
and
calibrating the inertial. sensor using the determined sensor bias.
According to another aspect of the invention there is provided a
method of determining a location of a chassis, the method including the steps
of:
determining a sensor bias according to the aforementioned method;
and =
determining the location of the chassis using a global navigation
satellite system (GNSS) component, the inertial sensor, and the determined
sensor bias.
CA 02848150 2014-03-07
WO 2013/033753
PCT/AU2012/001007
8
According to another aspect of the invention, there is provided a
system configured to determine an inertial sensor bias, the system including;
an inertial measurement unit (IMU) containing the sensor; and
a computing resource in communication with the IMU and including a
processor and memory;
wherein the memory of the computing resource is programmed to
instruct the processor to:
using the sensor to determine when the sensor is stationary;
automatically obtain a first inertial sensor measurement from the
IMU when the sensor is determined to the stationary;
automatically obtain a second inertial sensor measurement
when the sensor is determined to be stationary and at a different attitude to
the first inertial sensor measurement;
automatically obtain a third inertial sensor measurement when the
= sensor is determined to be stationary and at a different attitude to the
first and
second inertial sensor measurements; and
determine the inertial sensor bias using the first, second, and
third inertial sensor measurements.
According to another aspect of the invention there is provided a system
of calibrating an inertial measurement unit (IMU), the system including:
an IMU; and
a computing resource in communication with the IMU and including a
processor and memory; wherein the IMU:
CA 02848150 2014-03-07
WO 2013/033753
PCT/AU2012/001007
9
obtains a first inertial sensor measurement when the processor,
using output from the IMU, determines the IMU is stationary;
obtains a second inertial sensor measurement when the
processor, using output from the IMU, determines that the IMU is stationary
and at a different attitude to the first inertial sensor measurement; and
obtains a third inertial sensor measurement when the processor,
using output from the IMU, determines that the IMU is stationary and at a
different attitude to the first and second inertial sensor measurements;
and wherein the processor of the computing resource:
receives the first inertial sensor measurement, the second
inertial sensor measurement, and third inertial sensor measurement from the
IMU;
determines a sensor bias from the first, second, and third inertial
sensor measurements; and
calibrates the IMU using the determined sensor bias.
Preferably the computing resource is an embedded system. The
system may include a graphical display that may advise a user when the
sensor is determined to be stationary and/or when an inertial sensor
measurement is being obtained.
The IMU preferably includes a three-axis accelerometer. The IMU may
further include one or more angular rate sensors and/or a 2 or 3 axis
magnetometer. Preferably the system includes a chassis that contains the
IMU and computing resource.
CA 02848150 2014-03-07
WO 2013/033753
PCT/AU2012/001007
The inertial sensor bias may be determined according to any suitable
method once sufficient information, including sufficient inertial sensor
measurements, is obtained.
Further features and advantages of the present invention will become
5 apparent from the following detailed description.
BRIEF DESCRIPTION OF THE DRAWINGS
By way of example only, preferred embodiments of the invention will be
described more fully hereinafter with reference to the accompanying figures,
wherein:
10 Figure
1 is a flow chart illustrating steps of a method according to the
invention;
Figure 2 is a diagrammatic view illustrating an example application of
the method according to the invention; and
Figure 3 is a flow chart illustrating sub-steps of step 30 of the flow chart
in figure 1.
DETAILED DESCRIPTION OF THE DRAWINGS
The invention generally relates to determining sensor bias for an
inertial sensor, particularly an accelerometer. Inertial sensors have a bias
that
changes with temperature and time. Such inertial-sensors are used in many
applications including vehicles. Although the invention is primarily described
with reference to vehicles, and even more particularly with reference to land
vehicles, no limitation is meant thereby and the invention could be applied to
CA 02848150 2014-03-0.7
WO 2013/033753
PCT/AU2012/001007
=11
other embodiments including, for example, in electronic devices such as
electronic and electromechanical tools, mobile phones, consoles, game
controllers, remote controls, etc.
Figure 1 illustrates a flow chart that has steps (10 to 40) that outline a
method according to an embodiment of the invention. Typically, the inertial
sensor will be mounted in a chassis such as a vehicle chassis such that as the
chassis moves, the sensor moves with it. The invention relates to in field
recalibration of an inertial sensor which first starts with determining when
the
sensor is stationary (step 10). Typically the sensor will be part of an
inertial
measurement unit (IMU) and the sensor will be determined to be stationary
using all the IMU sensors combined to provide an accurate determination of
when the IMU is stationery. Once a sensor is determined tb be stationary
(step .10) an inertial sensor measurement is automatically obtained (step 20)
for that position and attitude. As the sensor is not moving, the measurement
should be a measurement of gravity only.
Once an inertial sensor measurement is obtained (step 20) it is stored
(step 30). Then, once sufficient inertial sensor measurements are available
the inertial sensor bias can be determined (step 40). According to a preferred
method of determining the sensor bias (discussed hereinafter), at least three
sensor measurements at different attitudes are required.
Figure 2 illustrates a diagrammatic view of an example application,
namely an agricultural application with a vehicle that traverses a route 50.
The vehicle has a chassis with an inertial measurement unit (IMU) mounted
CA 02848150 2014-03-07
WO 2013/033753
PCT/AU2012/001007
12
thereon. Age and temperature induced biases affect the accuracy of the IMU
and, hence, it needs to be calibrated (or recalibrated).
As the vehicle traverses route 50 stationary periods of no movement
are determined. For example, after starting the vehicle at a barn 52, a driver
of the vehicle may stop for a period of time to perform other actions (e.g. to
get equipment or engage a towed apparatus etc.). Once a period of no
movement is determined (step 10) an inertial sensor measurement may
= automatically be obtained (step 20) and stored (step 30). As the vehicle
continues there may be other stationary periods such as, for example, when
opening a gate 54 and/or when stopped in the field 56 (e.g. for a break or
when checking something). Inertial sensor measurements are automatically
obtained (step 20) and stored (step 30) at each period of no movement that is
sufficient to obtain a satisfactory measurement. Once sufficient inertial
sensor measurements have been obtained, the sensor bias may be
/5 determined (step 40). The sensor bias may be determined continually
and
improved with additional or replacement measurements over time.
A preferred method of determining the sensor bias, as outlined in figure
3, uses three separate inertial sensor measurements at different attitudes
(step 100). When a vehicle, or device containing an accelerometer, is
stationary the total force (fh ) acting on the accelerometer is due to gravity
and, accordingly, if scale, misalignment, noise and other error terms are
known or considered to be negligible, and the only significant error in
measurement is sensor bias ( ba ), then the following constraint must be
satisfied:
CA 02848150 2014-03-07
WO 2013/033753
PCT/AU2012/001007
13
II fh -b. 11= g (1)
where g is the magnitude of acceleration due to gravity. Accordingly, for
multiple arbitrary positions there are multiple equations:
II el -b. II g= (2)
1.62 b. 11g (3)
(4)
f - ba
II g(5)
=
Expanding equation (2) for a first inertial sensor measurement
results in:
(Abi -b.)2 +(f-hj = g2 = (6)
Each gravity measurement in equations (2)-(5) forms a sphere of
possible values for the bias and hence a first inertial sensor measurement can
. provide a first sphere of possible values for the bias.
With three measurements at three different attitudes, three spheres of
possible bias values (step 102) may be determined with the intersection of the
three spheres representing the bias of the sensor. Changes in attitude may
= occur as the result of driving the vehicle to different locations on non-
level
ground, such as a hill. For electronic devices, or the like, the attitude may
be
changed when the device is placed at different angles, such as a mobile
phone being placed upside down.
CA 02848150 2014-03-07
WO 2013/033753
PCT/AU2012/001007
14
In order to determine the sensor bias from equation (6), a second
inertial sensor measurement at a second attitude is obtained:
(frh2 ¨ bai, )2 +(f. yb2 ¨ ba y)2 +(f,b2 _b0,)2 = g2 (7)
and a third inertial sensor measurement at a third attitude is obtained (step
140):
(fxh3 ¨ ba,)2 + (.1' 3,b3 ¨ ba y)2 -1- (f,b3 _b0,)2 = g2 (8)
In determining the intersection of the three spheres (step 104), the
three inertial sensor measurements are considered (step 100) and, using
equations (6), (7), and (8), a sphere of possible bias values for each
measurement is determined (step 102). To determine the intersection (step
104), equation (7) is subtracted from equation (6) resulting in:
((Lb' )2¨ (f,b2)2 )+ 2k (f.:2 ¨ Lb' )-F
(Cf:' )2 ¨ (fP )2 )-F 2b0y(fb2 ¨ fybi )+
= ((Lb' )2 ¨ (f:b2
)2)+ 2k, (f,b2 ¨ fr,b1 Y= 0 = (9)
which is linear in lba x 13o3. kJ.
Subtracting equation (8) from (6) and (8) from (7) and rearranging into
matrix form results in:
-
-2 :2 ¨ frbi 2 :2 ¨ f:' 2(t2 ¨ JP - ba: -Ilfb2 112 ¨ 1 i fb 1 112
2 :3 _ ti 2Ç13 j:, 2 :3 _ fzbi bay = ii fb3 112
_lifbi 112 (10)
21J3 ¨ f:2 2 )1' 3 _f'12 2 zb 3 ¨ Ab2 baz _ II fb3 112 ¨ fifh2 ii2
CA 02848150 2014-03-07
WO 2013/033753
PCT/AU2012/001007
Equation (10) is then solvable using standard linear algebra techniques
from which the bias can be determined. More measurements, such as stored
measurements (step 30 from figure 1) can also be used in equation 10 by
subtracting the relevant sensor measurement equations in the same manner
5 as constructing equation (9) and solving using standard estimation
techniques, for example, using standard linear techniques such as the least-
squares technique.
If suitable measurements for bias determination are taken close to
each other in time, then the bias can be determined accurately. As stored
10 measurements age, however, their usefulness diminishes as the sensor
bias
changes with time and temperature. Different sensors change bias differently,
even from the same manufacturer, but in general sensor bias drifts relatively
slowly over time and relatively quickly with changes in temperature.
As sensor bias changes more notably with temperature, inertial sensor
15 measurements which are more than one to two hours old may need to be
discarded due to temperature variations. In this situation, it may be
difficult to
get sufficient inertial sensor measurements unless the sensor is stationary
frequently. Accordingly, in a preferred embodiment the temperature of the
sensor is preferably recorded when the inertial sensor measurement is
obtained. The temperature of the sensor at the time of measurement can be
utilised to estimate the temperature portion of the bias and, hence, inertial
sensor measurements of up to a much longer time period may be utilised
before being discarded. Effectively, the measurements will degrade primarily
due to age induced bias over time but it is envisaged that measurements of
CA 02848150 2014-03-07
WO 2013/033753
PCT/AU2012/001007
16
up to approximately a month, or even more, may be utilised without
significantly jeopardising the accuracy of the bias determination.
Inertial sensor measurements may be weighted to allow reduced
weight to be given to older measurements. A weighted least squares model
can be used to provide for use of measurements over a much longer period of
time such as, for example, a year. In effect, measurements for a particular
temperature can be retained and used for a long period of time, such as a
year, if degradation of the measurement due to time is accounted for.
The stationary period of no movement (for step 10) ideally needs to be
long enough for processing of the measurement to average readings in an
effort to reduce noise. Also, the measurement cannot be too long as other
factors may cause the measurement to drift from a true reading such as due
to flicker noise. The ideal length of time for a measurement is sensor
dependent but it is envisaged that a measurement in the range of 30 to 60
/5 seconds will be suitable. If movement of the sensor is resumed
during a
measurement, the measurement is discarded so that any portion of the
measurement due to the movement is does not influence the bias calculation.
In a preferred embodiment the method is implemented as part of a
navigation guidance system that includes a processor and memory, with the
memory being programmed to instruct the processor to carry out the method.
The method and system of the present invention advantageously
allows recalibration of a sensor to be performed without interruption to a
user
of the sensor. If sufficient measurements for sensor bias determination
CA 02848150 2014-03-07
WO 2013/033753
PCT/AU2012/001007
17
cannot be obtained (e.g. due to insufficient or inappropriate times when the
sensor is stationary) then the user may be prompted in an effort to obtain the
=
required measurements for recalibration. However, it is envisaged that this
will be unlikely and, in any event, the level of interruption will less than
if no
measurements had been taken automatically during stationary periods as less
measurements should be required from the user.
The method makes continual recalibration of an inertial sensor possible
resulting in increased accuracy of the sensor with minimal, if any,
interruption
of the user. In a vehicle navigation system, this reduces navigation errors
and
/0 offsets, and minimises downtime for recalibration of the sensor.
In the specification, reference to stationary and no movement of
serisors is intended to refer to no, or very little, movement of the sensor
relative to usual moving operation. For example, the sensor may move due to
other factors such as localised vibrations or the like.
In this specification, adjectives such as first and second, left and right,
top and bottom, and the like may be used solely to distinguish one element or
action from another element or action without necessarily requiring or
implying
am actual such relationship or order. Where the context permits, reference to
an integer or a component or step (or the like) is not to be interpreted as
being
limited to only one of that integer, component, or step, but rather could be
one
or more of that integer, component, or step etc.
The above description of various embodiments of the present invention
is provided for purposes of description to one of ordinary skill in the
related
=
CA 02848150 2014-03-07
WO 2013/033753
PCT/AU2012/001007
18
art. It is not intended to be exhaustive or to limit the invention to a single
disclosed embodiment. As mentioned above, numerous alternatives and
variations to the present invention will be apparent to those skilled in the
art of
the above teaching. Accordingly, while some alternative embodiments have
been discussed specifically, other embodiments will be apparent or relatively
easily developed by those of ordinary skill in the art. The invention is
intended
to embrace all alternatives, modifications, and variations of the present
invention that have been discussed herein, and other embodiments that fall
within the spirit and scope of the above described invention.
/o In this
specification, the terms 'comprises', 'comprising', 'includes',
'including', or similar terms are intended to mean a non-exclusive inclusion,
such that a method, system or apparatus that comprises a list of elements
does not include those elements solely, but may well include other elements
not listed.
=
=
