Resource allocation schemes in multi-vehicle cooperation systems 115
Multi-vehicle formation control has several advan-
tages, including increased instrument resolution, im-
proved efficiency, reduced cost, reconfiguration abil-
ity, and overall system robustness
[7]
. It also has
broad applications, for example, search and rescue
in hazardous environments, or area coverage and re-
connaissance in military missions. The key prob-
lems arising from the study of multi-vehicle forma-
tion control are listed below.
2.1.1 Formation selection
Control performance is closely related to the forma-
tion selection. Various factors impact the formation
selection, such as the number of vehicles and the type
of target. Several basic formations for a team of vehi-
cles are shown in Fig. 1, and these are: Line, in which
vehicles travel line abreast; column, in which vehicles
travel one after the other; diamond, in which vehi-
cles travel in a diamond; and wedge, in which vehi-
cles travel in a “V” formation
[8]
. The line formation
is appropriate for a linear motion path, the column
formation is suitable for a curved motion path, and
the diamond and wedge formations maintain a poly-
gon formation among the vehicles, which maintains
a relatively stable structure. Additionally, vehicles
must change their formation during a task to adapt
to complex mobile surroundings.
2.1.2 Formation maintenance
Two steps are required to maintain the formation
during movement. First, the target position of each
vehicle is determined according to their current sur-
roundings. Next, the control command is generated
on the basis of a particular control strategy, and the
vehicles are instructed to move to the target position
in a certain formation.
So far, the three typical multi-vehicle formation
control approaches are the leader-follower approach,
the behavior-based approach and the virtual struc-
ture approach, and each approach has its strength
and weakness. A brief introduction of these three
approaches is as follows.
• Leader-follower approach: The basic idea of the
leader-follower approach is that a particular vehicle
in a group of vehicles will be specified the leader,
while the others will be the followers and will fol-
low the leader at a certain distance
[9]
. This ap-
proach can be expanded based on the above descrip-
tion, meaning that more than one vehicle may be
the leader. Therefore, different network topologies
can be formed according to the relative position of
the leader and the followers. By applying this ap-
proach to formation control, cooperation is realized
by sharing the mutual leader. The strength of this
approach is that the behavior of the whole cluster
can be controlled, as long as the behavior or the path
of the leader is provided. The weakness of the sys-
tem lies in the lack of explicit formation feedback,
which means that the follower may not be able to
follow the leader if it goes too fast, and the forma-
tion cannot be maintained if the leader loses effi-
cacy. Corresponding measures can be adopted aimed
at the above weakness, such as applying feedback
linearization and specifying another vehicle as the
leader while the previous one is out of control.
• Behavior-based approach: The basic idea of the
behavior-based approach is firstly to set a primary
behavior for the vehicle, which includes obstacle
avoidance, target achievement and formation main-
tenance under general conditions. When the sensor
of the vehicle accepts external stimuli, the correct
behavior will be selected according to the informa-
tion input, and the vehicle will react in the way that
best meets the intention. In this approach, the coop-
eration is realized by sharing the positions and states
among the vehicles. The advantage of the behavior-
based approach is that it can easily obtain the ap-
propriate control strategy when several competitive
targets exist. The system can also give explicit for-
mation feedback, due to the fact that each vehicle
reacts according to the other vehicles’ positions. Al-
though this approach can realize distributed control,
it still cannot clearly define the cluster behavior and
conduct mathematical analysis.
• Virtual structure approach: The virtual struc-
ture approach uses the movement of a rigid body,
with varying degrees of freedom for reference. When
a rigid body moves with varying degrees of freedom,