Part B | 14.3
352 Part B Autonomous Ocean Vehicles, Subsystems and Control
which is synchronized to a global clock via GPS on the
surface [14.92–94].
Accurate time keeping has been central to precision navigation at sea throughout history [14.32]. Recent implementations of synchronous-clock one-waytravel-time acoustic navigation by Eustice et al. [14.92]
and Bahr et al. [14.95] have been enabled by using
a low-power temperature compensated crystal oscillator
(TCXO), from SeaScan Inc., which has a typical drift of
20 ns=s, corresponding to a drift in position of approximately 2:6 m=day [14.96]. A common timebase and
embedding a time-stamp indicating the broadcast time
of each outgoing transmission allow a receiving vehicle to determine the time which the signal took to reach
the receiver. With an accurate estimate of the speed of
sound in water, the vehicle can then determine the distance to the broadcasting vehicle.
Strategies
Two different kinds of strategies can be employed for
cooperative navigation. The first assumes a strict hierarchy where a group of AUVs acts as dedicated
navigation aids (NAs). Their sole purpose is to provide
navigation information to the mission AUVs (MAs).
The NAs are thus equipped with very sophisticated sensors such that the drift of their position estimate is
minimized. They may also operate very close to the surface such that they can easily surface to update their
position using a GPS fix. In this scenario, only NAs
broadcast navigation information and serve as mobile
LBL-beacons to all MAs in communication range. As
the sole purpose of the NAs is to maximize the MAs’
navigation accuracy, their path can be adapted to best
accomplish that goal [14.97].
The second strategy does not impose any particular hierarchy. Each vehicle occasionally broadcasts
its position estimate and incorporates other overheard
broadcasts. This approach does not require an extra
set of dedicated vehicles, but due to the stochastic nature of received updates it is very difficult to predict,
let alone guarantee, the performance of the navigation
improvement of this cooperation strategy. In addition,
the fact that each vehicle may transmit and receive
navigation information may lead to cyclical updates.
This occurs when a vehicle A incorporates information
from a vehicle B and at a later point in time vehicle B
then incorporates information from A. The dependencies arising between the position estimates may lead to
overconfidence in the position estimate of each vehicle
and must be carefully mitigated [14.98].
Algorithms
By treating the information received from a cooperating AUV as the noisy observation of landmark with
an uncertain position, we can draw on a vast amount
of algorithms and techniques developed by the SLAM
community. Their work provides various frameworks
to represent a state estimate (in this case a position)
together with the uncertainty of this estimate. It also
provides several methods to incorporate landmark observations, in this case received broadcasts from cooperating vehicles. Several of these SLAM techniques
have been adapted for cooperative navigation [14.95,
99, 100].
14.3 Summary
Figure 14.8 shows typical navigation system performance for current AUVs based on five typical vehicle
configurations. The sensor suite employed for an AUV
depends on the navigation accuracy required for the
mission as well as the available power, space, and the
cost constraints.
14.3.1 Glider with Very Low Power
Sensor Suite
Autonomous ocean gliders must operate for extended
periods of time without being able to recharge their
batteries [14.9, 101]. As a result, power consumption
is the limiting factor for the selection of navigation
sensors, and the navigation suite of a glider usually
consists of a GPS, an AHRS, and a pressure gauge.
While submerged the glider uses the AHRS combined with a vehicle model to estimate its heading
and forward velocity and dead-reckon its position. The
high noise and the unobservable variables in the vehicle model lead to a very high drift of 30% or even
more if strong currents are present. On the surface,
the vehicle resets its position estimate using GPS. The
navigation accuracy achieved is typically more than
adequate for providing data into ocean circulation models [14.102].
14.3.2 Low-Cost AUV Sensor Suite
A low-cost AUV such as the IVER uses a flow meter to
obtain a measurement of their forward speed u
V
x . This
information combined with an AHRS leads to a significant improvement of the navigation accuracy when
compared to that of a glider.
352 Part B Autonomous Ocean Vehicles, Subsystems and Control
which is synchronized to a global clock via GPS on the
surface [14.92–94].
Accurate time keeping has been central to precision navigation at sea throughout history [14.32]. Recent implementations of synchronous-clock one-waytravel-time acoustic navigation by Eustice et al. [14.92]
and Bahr et al. [14.95] have been enabled by using
a low-power temperature compensated crystal oscillator
(TCXO), from SeaScan Inc., which has a typical drift of
20 ns=s, corresponding to a drift in position of approximately 2:6 m=day [14.96]. A common timebase and
embedding a time-stamp indicating the broadcast time
of each outgoing transmission allow a receiving vehicle to determine the time which the signal took to reach
the receiver. With an accurate estimate of the speed of
sound in water, the vehicle can then determine the distance to the broadcasting vehicle.
Strategies
Two different kinds of strategies can be employed for
cooperative navigation. The first assumes a strict hierarchy where a group of AUVs acts as dedicated
navigation aids (NAs). Their sole purpose is to provide
navigation information to the mission AUVs (MAs).
The NAs are thus equipped with very sophisticated sensors such that the drift of their position estimate is
minimized. They may also operate very close to the surface such that they can easily surface to update their
position using a GPS fix. In this scenario, only NAs
broadcast navigation information and serve as mobile
LBL-beacons to all MAs in communication range. As
the sole purpose of the NAs is to maximize the MAs’
navigation accuracy, their path can be adapted to best
accomplish that goal [14.97].
The second strategy does not impose any particular hierarchy. Each vehicle occasionally broadcasts
its position estimate and incorporates other overheard
broadcasts. This approach does not require an extra
set of dedicated vehicles, but due to the stochastic nature of received updates it is very difficult to predict,
let alone guarantee, the performance of the navigation
improvement of this cooperation strategy. In addition,
the fact that each vehicle may transmit and receive
navigation information may lead to cyclical updates.
This occurs when a vehicle A incorporates information
from a vehicle B and at a later point in time vehicle B
then incorporates information from A. The dependencies arising between the position estimates may lead to
overconfidence in the position estimate of each vehicle
and must be carefully mitigated [14.98].
Algorithms
By treating the information received from a cooperating AUV as the noisy observation of landmark with
an uncertain position, we can draw on a vast amount
of algorithms and techniques developed by the SLAM
community. Their work provides various frameworks
to represent a state estimate (in this case a position)
together with the uncertainty of this estimate. It also
provides several methods to incorporate landmark observations, in this case received broadcasts from cooperating vehicles. Several of these SLAM techniques
have been adapted for cooperative navigation [14.95,
99, 100].
14.3 Summary
Figure 14.8 shows typical navigation system performance for current AUVs based on five typical vehicle
configurations. The sensor suite employed for an AUV
depends on the navigation accuracy required for the
mission as well as the available power, space, and the
cost constraints.
14.3.1 Glider with Very Low Power
Sensor Suite
Autonomous ocean gliders must operate for extended
periods of time without being able to recharge their
batteries [14.9, 101]. As a result, power consumption
is the limiting factor for the selection of navigation
sensors, and the navigation suite of a glider usually
consists of a GPS, an AHRS, and a pressure gauge.
While submerged the glider uses the AHRS combined with a vehicle model to estimate its heading
and forward velocity and dead-reckon its position. The
high noise and the unobservable variables in the vehicle model lead to a very high drift of 30% or even
more if strong currents are present. On the surface,
the vehicle resets its position estimate using GPS. The
navigation accuracy achieved is typically more than
adequate for providing data into ocean circulation models [14.102].
14.3.2 Low-Cost AUV Sensor Suite
A low-cost AUV such as the IVER uses a flow meter to
obtain a measurement of their forward speed u
V
x . This
information combined with an AHRS leads to a significant improvement of the navigation accuracy when
compared to that of a glider.