among the diffe rent vehicles and it is required in a wide
spectrum of applications. For instance, for objects mo nitor-
ing, several synchronized perceptions of the objects could be
required. In addition, spatial coordination of UAVs deals
with the sharing of the space among them to ensure that each
UAV will be able to perform safely and coherently regarding
the plans of the other UAVs, and the potential dynamic and/
or static obstacles.
Regarding cooperation, according to Cao et al., (1997),
given some task specified by a designer, a multiple-robot
system displays cooperative behavior if, due to some under-
lying mechanism (i.e., the “mechanism of cooperation”),
there is an increase in the total utility of the system. The
cooperation of heterogeneous vehicles requires the integra-
tion of sensing, control, and planning in an appropriated
decisional architecture. These architectures can be either
centralized or decentralized depending of the assumptions
on the knowledge's scope and accessibility of the individual
vehicles, their computational power, and the required scal-
ability. A centralized approach will be relevant if the com-
putational capabilities are compatible with the amount of
information to process, and the exchange of data meets both
the requirements of speed (up-to-date data) and expressivity
(quality of information enabling well-informed decision-
taking). On the other hand, a distributed approach will be
possible if the available knowledge within each distributed
vehicle is sufficient to perform “coherent” decisions, and this
required amount of knowledge does not endow the distrib-
uted components with the inconveniences of a centralized
system (in terms of computation power and communication
bandwidth requirements). One way to ensure that a minimal
global coherence will be satisfied within the whole system is
to enable communication between the vehicles of the system,
up to a level that will guarantee that the decision is globally
coherent. One of main advantages of the distributed approach
relies on its superior suitability to deal with the scalability of
the system.
2.2 Classification of multi-UAV architectures
Multi-UAV systems can be classified from different points of
view. One possible classification is based on the coupling
between the UAVs:
1. Physical coupling. In this case, the UAVs are connected
by physical links and then their motions are constrained
by forces that depend on the motion of other UAVs. Th e
lifting and transportation of loads by several UAVs lies
in this categor y (Bernard et al., 2011). The main problem
is the motion coordinated control taking into account the
forces constraints. From the point of view of motion
planning and collision avoidance, all the members of the
team and the load can be considered as a whole. As the
number of vehicles is usually low, both centralized and
decentralized control architectures can be applied.
2. Formations. The vehicles are not physically coupled but
their relative motions are strongly constrained to keep the
formation. Then, the motion planning problem can be also
formulated considering the formation as a whole. Regard-
ing the collision avoidance problem within the team, it is
possible to embed it in the formation control strategy.
Scalability properties to deal with formations of many
individuals are relevant and then, decentralized control
architectures are usually preferred (Turpin et al., 2012).
Techniques in this area can be applied to control coordi-
natedmotions of vehicles even if they are not in formation.
3. Swarms. They are homogeneous teams of many vehi-
cles in which interactions generate emerging collective
behaviors. The resulting motion of the vehicles does not
Figure 1. A team of helicopters and an airship were demonstrated in forest fire-fighting activities in the EU-funded COMETS Project (Ollero
et al., 2005).
348 Multi-Vehicle Cooperation and Coordination
评论0