IEEE Robotics & Automation Magazine - June 2019 - 97

thermoclines. They also have a limited range and can be used
only in confined areas (LBL) or require a support vessel following the UUV (SBL or USBL) [19].
As mentioned previously, navigation capabilities can also
be improved with exteroceptive sensors (optical or sonar) that
identify specific landmarks in the environment and use them
to localize the UUV. If a map is available, this approach is
known as map-based localization [20]. When no map is available, AUVs can perform SLAM, in which the vehicle concurrently builds a map of relevant features and uses it to navigate.
The latter approach is most effective when the vehicle revisits
the same area multiple times, enabling loop-closing and a
global reduction of the uncertainty in vehicle position [21].
While SLAM is very effective in land and aerial robotics, it is
still a challenge to use it in the subsea domain [22] because of
the relative paucity of unique features underwater, the short
range of high-resolution optical sensors, and the low resolution of long-range acoustic sensors. The biggest challenge that
remains is to navigate in an unknown environment without a
priori information or external position updates and without
using a very precise (and costly) INS or DVL.
Guidance [23] focuses on high-level trajectory planning,
while control is devoted to the low-level kinematics and
dynamics control of actuators to achieve guidance objectives
[24]. High-performance and achievable control systems are
the current state of the art [25]. Indeed, while guidance and
control for a single vehicle are mature domains, cooperative
navigation and control of networked UUVs remain challenges.
To the best of our knowledge, no commercial system based on
multiple AMVs exists. In [26], adaptive ocean sampling was
first tested with a fleet of gliders. In a recent EU-funded project, MORPH [27], significant advancements in coordinated
control have been achieved, while in another European project, Widely Scalable Mobile Underwater Sonar Technology
(WiMUST) [28], cooperative control was designed and implemented in a survey operation involving two autonomous surface catamarans and six AUVs. The WiMUST application, i.e.,
acoustic-based geophysical and geotechnical surveys, required
the vehicles to carry streamers of hydrophones whose positions needed to be synchronized and known.
Challenges in guidance and control of multiple vehicles
mainly reside in the bottleneck presented by using the same
limited-bandwidth physical medium for both navigation and
communication, i.e., acoustic. Thus, algorithms need to be
robust in handling frequent packet loss and multipath arrival
of data. Beyond the theoretical aspects, from a purely practical perspective, a significant barrier is the difficulty of
launching and recovering a large number of vehicles in any
kind of weather.
Communications
Underwater communications are a challenge. Currently,
acoustic communications can be used for navigational
updates from the surface (SBL or USBL) or the bottom (LBL)
and to communicate a UUV's status to operators. UUVs can
also receive messages from other vessels for control and

coordination. Acoustic communications allow for longer
ranges than optical communications; recently, however, optical communications solutions have been developed thanks to
the availability of powerful LEDs and lasers combined with
high-sensitivity avalanche photodiodes [29]. The advantage is
that the bandwidth is typically much higher than acoustic
communications (up to 500 Mb/s for a range of tens of meters
compared to Kb/s over a few kilometers).
As mentioned, acoustic sensors underwater require good
calibration, which is sometimes a practical challenge. The low
bandwidth of acoustic
communications and the
The biggest challenge
short range of optical
communications drastithat remains is to
cally limit the level of
coordination and supervinavigate in an unknown
sion possible. This is a
challenge for the developenvironment without
ment of under water
acoustic networks [30].
a priori information or
Moreover, these networks
are normally associated
external position updates
with permanent deployments because making
and without using a very
them portable is challenging. At the same time, the
precise (and costly) INS
underwater communications challenges have proor DVL.
moted the development
of more autonomy inside
the vehicles. Unlike other
domains, underwater roboticists have no choice but to embed
autonomy to enable useful and safe operations.
Autonomy and Planning
The first UUVs had only basic levels of autonomy and executed strictly prescribed missions. They were equipped with
a number of safety behaviors that would become activated in
the case of problems and would generally lead to the termination of the mission [31]. More complex autonomy levels
were slowly introduced for autonomous inspection [4].
Complex mission-planning algorithms were able to adapt the
initial plan based on new data gathered or changes in environmental conditions [32]. These were applied to ocean sampling [33] and autonomous intervention [34]. Extensions to
multivehicle coordination and planning were explored and
demonstrated in actual large-scale experiments using a collection of UUVs and USVs for marine science [35] and
defense. We are, therefore, on the cusp of a new era in which
large numbers of autonomous systems can be deployed and
collaborated to perform complex, long-term missions.
However, a number of challenges remain. First, as the
complexity of the embedded planning abilities grows and
increasingly involves some form of machine learning, so
does the need to keep the operator situated and maintain
operator trust in the embedded intelligence. Moreover,
JUNE 2019

IEEE ROBOTICS & AUTOMATION MAGAZINE

IEEE Robotics & Automation Magazine - June 2019

Table of Contents for the Digital Edition of IEEE Robotics & Automation Magazine - June 2019