Part B | 22.1
500 Part B Autonomous Ocean Vehicles, Subsystems and Control
schemes for cooperative target tracking with measurable effectiveness. This trading methodology can be
used to evaluate the effectiveness of different schemes,
and to predict the effectiveness of a given scheme under
changing environmental and operational constraints.
The trade-offs are, in general, effectiveness versus efficiency (for a specific scenario, time to find a target
versus number of assets). Applying the methodology to
a specific scenario avoids the computational limitations
of brute-force processing methods. Two example scenarios are discussed stemming from multistatic sonar
and maritime surveillance.
Cooperative target tracking using autonomous
ocean vehicles is a multidisciplinary endeavor that
benefits from recent progress in distributed sensor networking, networked robotics, collaborative signal processing, control and game theory, and interactions of
multiagent communities [22.3–11]. However, maritime
applications have unique challenges including transient
environmental conditions, high noise levels at sensors,
sparse coverage over large areas, intermittent and low
bandwidth communication channels, navigation uncertainty, and limited energy availability.
Communication is essential for coordinating the behavior of individuals or agents. Communication can be
preplanned or adaptive. To minimize communication
bandwidth, two questions must be answered: What information should be shared and with whom should this
information be shared? More specifically, for the targettracking task: How much information is sufficient to
allow data association and pattern recognition? Given
the assumption of intelligent targets with detection and
avoidance capability, communication events should be
as rare as possible. From a system design point of view,
the communication process should minimize interdependency, since interdependency can result in unpredictable and perhaps uncontrollable system behavior.
Here, the target-tracking team consists of an array of
mobile sensors distributed throughout an ocean domain
of interest. If the spatial domain is large, and the number of mobile sensors relatively small, then part of the
solution for robust tracking is determining curves of intercept for sensors along a projected track. Such curves
can be mapped covertly with minimal target reaction to
the sensing or they can be manipulated through deliberate interaction with the object. The tracking process can
be divided into two phases:
A search phase (I) in which the cooperative sensor
interaction is used to fuse measurements in time and
space to increase the signal-to-noise ratio of potential target signatures.
A holding phase (II) in which the highest ranked
signals from the first phase are correlated with
tracks and the tracking team adapts to best follow
the tracks.
In Phase I, the performance of a single sensor can
be enhanced to increase range, or multiaspect measurements made to better determine a signature.
In Phase II, a track can be followed with a single-sensing system whose speed and maneuverability
matches that of the target (track and trail), or with a network of sensors distributed over the domain in a handoff
mode.
In both phases, if the sensors are stationary and the
detection ranges are modest relative to the domain size,
the number of sensors must be large. If instead the sensors have mobility and can adapt to the environment and
to the projected tracks, then their effectiveness increases
and the number of sensors is reduced. Further effectiveness gains can be achieved if the sensors can adapt
to one another (cooperative behavior). This chapter focuses on the latter approach to tracking performance
enhancement.
When the target is reactive (perceives the tracking
and maneuvers evasively), cooperative target tracking
becomes a game. Each player in this game has a given
system capability. Both must be modeled consistent
with the stochastic nature of the control and estimation
processes operative during game execution.
Cooperative target tracking over realistic maritime
time and space scales is a complex adaptive system.
The specific application dictates the level of cooperation needed:
From situations where the cooperation is not really
necessary to achieve sufficient performance, and
the built-in cooperation dependencies introduces an
overhead which hampers the overall system. Divide and conquer is rather the better heuristic then.
Stated differently: If the cooperation introduces interdependencies, the risk that these interdependencies produce is that an unstable and unpredictable
overall system should be monitored.
To situations where cooperation is mandatory, especially against stealth targets, in nonstationary
environmental conditions. In this case, single sensors or platforms are not capable of detecting targets
or holding tracks, and collaboration is the only
way to achieve a sufficient overall system effectiveness. The interconnectivity between the collaborating platforms is key to achieving the gain in
effectiveness: One system is helping the other or
one platform is covering the potential poor performance of another platform. This covering can be
done simply (but not very efficiently) by adding
more assets, scanning, and acting independently. Or
it can be done by implementing synchronization
500 Part B Autonomous Ocean Vehicles, Subsystems and Control
schemes for cooperative target tracking with measurable effectiveness. This trading methodology can be
used to evaluate the effectiveness of different schemes,
and to predict the effectiveness of a given scheme under
changing environmental and operational constraints.
The trade-offs are, in general, effectiveness versus efficiency (for a specific scenario, time to find a target
versus number of assets). Applying the methodology to
a specific scenario avoids the computational limitations
of brute-force processing methods. Two example scenarios are discussed stemming from multistatic sonar
and maritime surveillance.
Cooperative target tracking using autonomous
ocean vehicles is a multidisciplinary endeavor that
benefits from recent progress in distributed sensor networking, networked robotics, collaborative signal processing, control and game theory, and interactions of
multiagent communities [22.3–11]. However, maritime
applications have unique challenges including transient
environmental conditions, high noise levels at sensors,
sparse coverage over large areas, intermittent and low
bandwidth communication channels, navigation uncertainty, and limited energy availability.
Communication is essential for coordinating the behavior of individuals or agents. Communication can be
preplanned or adaptive. To minimize communication
bandwidth, two questions must be answered: What information should be shared and with whom should this
information be shared? More specifically, for the targettracking task: How much information is sufficient to
allow data association and pattern recognition? Given
the assumption of intelligent targets with detection and
avoidance capability, communication events should be
as rare as possible. From a system design point of view,
the communication process should minimize interdependency, since interdependency can result in unpredictable and perhaps uncontrollable system behavior.
Here, the target-tracking team consists of an array of
mobile sensors distributed throughout an ocean domain
of interest. If the spatial domain is large, and the number of mobile sensors relatively small, then part of the
solution for robust tracking is determining curves of intercept for sensors along a projected track. Such curves
can be mapped covertly with minimal target reaction to
the sensing or they can be manipulated through deliberate interaction with the object. The tracking process can
be divided into two phases:
A search phase (I) in which the cooperative sensor
interaction is used to fuse measurements in time and
space to increase the signal-to-noise ratio of potential target signatures.
A holding phase (II) in which the highest ranked
signals from the first phase are correlated with
tracks and the tracking team adapts to best follow
the tracks.
In Phase I, the performance of a single sensor can
be enhanced to increase range, or multiaspect measurements made to better determine a signature.
In Phase II, a track can be followed with a single-sensing system whose speed and maneuverability
matches that of the target (track and trail), or with a network of sensors distributed over the domain in a handoff
mode.
In both phases, if the sensors are stationary and the
detection ranges are modest relative to the domain size,
the number of sensors must be large. If instead the sensors have mobility and can adapt to the environment and
to the projected tracks, then their effectiveness increases
and the number of sensors is reduced. Further effectiveness gains can be achieved if the sensors can adapt
to one another (cooperative behavior). This chapter focuses on the latter approach to tracking performance
enhancement.
When the target is reactive (perceives the tracking
and maneuvers evasively), cooperative target tracking
becomes a game. Each player in this game has a given
system capability. Both must be modeled consistent
with the stochastic nature of the control and estimation
processes operative during game execution.
Cooperative target tracking over realistic maritime
time and space scales is a complex adaptive system.
The specific application dictates the level of cooperation needed:
From situations where the cooperation is not really
necessary to achieve sufficient performance, and
the built-in cooperation dependencies introduces an
overhead which hampers the overall system. Divide and conquer is rather the better heuristic then.
Stated differently: If the cooperation introduces interdependencies, the risk that these interdependencies produce is that an unstable and unpredictable
overall system should be monitored.
To situations where cooperation is mandatory, especially against stealth targets, in nonstationary
environmental conditions. In this case, single sensors or platforms are not capable of detecting targets
or holding tracks, and collaboration is the only
way to achieve a sufficient overall system effectiveness. The interconnectivity between the collaborating platforms is key to achieving the gain in
effectiveness: One system is helping the other or
one platform is covering the potential poor performance of another platform. This covering can be
done simply (but not very efficiently) by adding
more assets, scanning, and acting independently. Or
it can be done by implementing synchronization