Abstract:
The eld of multi-agents presents highly e cient solutions to problems
that may be too complex for single arti cial agents to address e ectively.
By simply breaking down a complicated agent function into simpler sub-
functions that can be coordinated either centrally or in a distributive en-
vironment, multi-agent systems can be applied to a wide array of problem
domains. One such domain is robotics where autonomous decision-making
is imperative and simultaneous learning functions are signi cantly gaining
importance. Therefore, the combination of the aforementioned disciplines
of agents and robotics provides unique opportunities in terms of achieving
advanced functionalities for wide variety of applications.
Conversely, this arrangement results in a signi cant number of challenges
as well. For instance, robotics generally involves hardware limitations that
can prohibit the design of the multi-agent system. A similar case is explored
here where a multi-agent system is developed for highly maneuverable quad-
copters intended for the possible surveillance of a particular area for emer-
gency evacuation purposes in case of disaster. However, the limited
ight
times and load bearing capacity of the quad copter present major hurdles
towards the architectural design.
The basic idea remains bridging the cyber world to physical world in
the most e cient way. Therefore the designed multi-agent system manages
group of robots in an unstructured environment resulting from catastrophic
happenings such as earthquakes. Hence, the robots are responsible for sur-
veying a disaster a ected area and locating the maximum number of trapped
and/or injured people as part of the immediate response. The major feature
of proposed architecture is that it e ectively manages group of quad copters
within a cloud framework for increased reliability and ease of accessibility.
The proposed multi-agent system for cloud base management of collab-
orative robots has been tested using simulation, Several experiments with
varying earthquake magnitude and corresponding scale of devastation have
been carried out; whereby the results show that proposed architecture e ec-
tively manages group of robots given an unstructured environment.
