Autonomous Underwater Vehicle Navigation 14.2 Algorithms 349
Part B | 14.2
for future research may be to employ techniques from
acoustic time reversal [14.50] to achieve higher accuracy in range estimation while concurrently estimating
ocean acoustic propagation conditions.
14.2.3 Geophysical Map-Based Navigation
For some applications of AUVs, the use of acoustic beacons is undesirable or impractical. If an accurate a priori map of the environment is available,
one approach to globally-referenced position estimation is to use measurements of geophysical parameters, such as bathymetry, magnetic field, or gravitational anomaly [14.16, 17, 51, 52]. These approaches
are based on matching sensor data with an a priori
environment map, under the assumption that there is
sufficient spatial variation in the parameter(s) being
measured to permit accurate localization.
All forms of map-based navigation are motivated
by the desire to operate at an arbitrary location without
the additional expense or problems associated with the
installation of artificial beacons. In principle, the process appears straightforward: gather information about
the surrounding terrain and match that information to
an on-board map or database of terrain information.
When the vehicle has a match to the database, then it
knows its location on the map. This is analogous to the
method which humans use to navigate; we find our way
to our destination by locating and identifying landmarks
which are familiar to us – either from past experience or
via a map which has been constructed for our benefit.
In practice this form of navigation is not so simple.
The vehicle is attempting to navigate by matching a set
of sensed data with an a priori map or dataset of stored
data. Two key problems are the cost and difficulty of
generating the a priori maps and the computational
complexity of searching for a peak in the n-dimensional
correlation surface, where n is the number of dimensions in the map or sensor data set. Typically, map
making expense is governed by both the type of data
being collected and the desired resolution of that data.
Determining the map resolution has a direct effect on
the size and level of detail of the search needed to
locate the vehicle in space. Since the vehicle could
be in any of a large number of possible orientations
relative to the original dataset, the search must be performed over all possible locations and orientations. This
is a potentially large search space, necessitating some
simplifications and/or simplifying assumptions in order
to make the search more tractable. Typical simplifications are: restricting the types of map data stored (what
sensor values, how many different sensors), lowering
map resolution, patchy maps (maps of key areas only),
restricting vehicle orientations (to reduce the correlation problem), and using inertial navigation or dead
reckoning systems to limit the valid search area.
Evidence exists that geomagnetic navigation is employed by birds, fish, and other animals for migration
and general navigation [14.53]. The magnetic flux density of the earth varies according to latitude, the presence of man-made and natural anomalies, and even
one’s depth in the ocean, increasing from 6 to 30 nT=km
depth, depending on location [14.54]. Additionally,
there are small but predictable variations in the earth’s
magnetic flux from day to night, and large arbitrary
changes during magnetic storms; magnetic maps can be
rendered useless for the duration of such storms. Useful
magnetic maps, generated by satellites or surface ships,
can be employed by underwater vehicles by accounting for the daily field variations and by calculating the
effective magnetic field at depth using a Laplace field
equation, setting the boundary conditions at the ocean
surface [14.55].
Research into the nature of the earth’s gravitational
field has demonstrated that it is far from uniform and indeed possesses a varied topography [14.56, 57]. These
variations are due to a variety of factors, especially the
effects of local topography [14.58] and density inhomogeneities [14.59]. Variations in the earth’s gravitational
field on the ocean’s surface relative to a regular ellipsoidal model have been measured to be on the order
of 3050 mgal [14.60]. Gravity maps were originally
gathered on behalf of the US Navy for the purposes
of INS calibration [14.61]. To an INS, the effects
of a change in the local gravitational field are indistinguishable from accelerations of the vehicle itself.
Gerber [14.62] proposed the use of a gravity gradiometer as an aid to INS. Jircitano et al. extended this idea
to the AUV community, performing navigation simulations using a model of the Bell Aerospace Textron
Gravity Gradiometer System [14.60] with good preliminary results. The drawbacks to such a system are the
size, expense, and complexity of a gradiometer. In addition, the gradiometer must be mounted on an inertially
stabilized and vibrationally isolated platform, making
its use difficult on small, low-cost scientific AUVs.
Geophysical navigation algorithms have origins in
techniques of navigating at sea using depth soundings
that have been in use for centuries [14.31]. KamgarParsi has developed techniques for performing geophysical navigation that are based on fitting contour
lines to sensor data and matching these curves to an
a priori map using matching techniques from computer
vision [14.63, 64]. Lucido et al. have also investigated
the segmentation and registration of bathymetric profiles [14.65]. Tuohy et al. have investigated geophysical
navigation using maps of multiple geophysical parameters based on contour intersection methods [14.17]. The
Part B | 14.2
for future research may be to employ techniques from
acoustic time reversal [14.50] to achieve higher accuracy in range estimation while concurrently estimating
ocean acoustic propagation conditions.
14.2.3 Geophysical Map-Based Navigation
For some applications of AUVs, the use of acoustic beacons is undesirable or impractical. If an accurate a priori map of the environment is available,
one approach to globally-referenced position estimation is to use measurements of geophysical parameters, such as bathymetry, magnetic field, or gravitational anomaly [14.16, 17, 51, 52]. These approaches
are based on matching sensor data with an a priori
environment map, under the assumption that there is
sufficient spatial variation in the parameter(s) being
measured to permit accurate localization.
All forms of map-based navigation are motivated
by the desire to operate at an arbitrary location without
the additional expense or problems associated with the
installation of artificial beacons. In principle, the process appears straightforward: gather information about
the surrounding terrain and match that information to
an on-board map or database of terrain information.
When the vehicle has a match to the database, then it
knows its location on the map. This is analogous to the
method which humans use to navigate; we find our way
to our destination by locating and identifying landmarks
which are familiar to us – either from past experience or
via a map which has been constructed for our benefit.
In practice this form of navigation is not so simple.
The vehicle is attempting to navigate by matching a set
of sensed data with an a priori map or dataset of stored
data. Two key problems are the cost and difficulty of
generating the a priori maps and the computational
complexity of searching for a peak in the n-dimensional
correlation surface, where n is the number of dimensions in the map or sensor data set. Typically, map
making expense is governed by both the type of data
being collected and the desired resolution of that data.
Determining the map resolution has a direct effect on
the size and level of detail of the search needed to
locate the vehicle in space. Since the vehicle could
be in any of a large number of possible orientations
relative to the original dataset, the search must be performed over all possible locations and orientations. This
is a potentially large search space, necessitating some
simplifications and/or simplifying assumptions in order
to make the search more tractable. Typical simplifications are: restricting the types of map data stored (what
sensor values, how many different sensors), lowering
map resolution, patchy maps (maps of key areas only),
restricting vehicle orientations (to reduce the correlation problem), and using inertial navigation or dead
reckoning systems to limit the valid search area.
Evidence exists that geomagnetic navigation is employed by birds, fish, and other animals for migration
and general navigation [14.53]. The magnetic flux density of the earth varies according to latitude, the presence of man-made and natural anomalies, and even
one’s depth in the ocean, increasing from 6 to 30 nT=km
depth, depending on location [14.54]. Additionally,
there are small but predictable variations in the earth’s
magnetic flux from day to night, and large arbitrary
changes during magnetic storms; magnetic maps can be
rendered useless for the duration of such storms. Useful
magnetic maps, generated by satellites or surface ships,
can be employed by underwater vehicles by accounting for the daily field variations and by calculating the
effective magnetic field at depth using a Laplace field
equation, setting the boundary conditions at the ocean
surface [14.55].
Research into the nature of the earth’s gravitational
field has demonstrated that it is far from uniform and indeed possesses a varied topography [14.56, 57]. These
variations are due to a variety of factors, especially the
effects of local topography [14.58] and density inhomogeneities [14.59]. Variations in the earth’s gravitational
field on the ocean’s surface relative to a regular ellipsoidal model have been measured to be on the order
of 3050 mgal [14.60]. Gravity maps were originally
gathered on behalf of the US Navy for the purposes
of INS calibration [14.61]. To an INS, the effects
of a change in the local gravitational field are indistinguishable from accelerations of the vehicle itself.
Gerber [14.62] proposed the use of a gravity gradiometer as an aid to INS. Jircitano et al. extended this idea
to the AUV community, performing navigation simulations using a model of the Bell Aerospace Textron
Gravity Gradiometer System [14.60] with good preliminary results. The drawbacks to such a system are the
size, expense, and complexity of a gradiometer. In addition, the gradiometer must be mounted on an inertially
stabilized and vibrationally isolated platform, making
its use difficult on small, low-cost scientific AUVs.
Geophysical navigation algorithms have origins in
techniques of navigating at sea using depth soundings
that have been in use for centuries [14.31]. KamgarParsi has developed techniques for performing geophysical navigation that are based on fitting contour
lines to sensor data and matching these curves to an
a priori map using matching techniques from computer
vision [14.63, 64]. Lucido et al. have also investigated
the segmentation and registration of bathymetric profiles [14.65]. Tuohy et al. have investigated geophysical
navigation using maps of multiple geophysical parameters based on contour intersection methods [14.17]. The