Abstract:
This Project aims to target Multi-agent Intelligent Electro-mechanical Systems (MIEMS), a
relatively new field which focuses on making intelligent autonomous agents, capable of acting
autonomously, to behave in a collaborative manner. The basic problem proposed is to apply the
concepts of multi-agent autonomous systems to develop a team of robots, which could effectively
play a game of soccer in a real-time dynamic environment. Robot Soccer based environment is best
suitable for multi-agent artificial intelligence because of its adversarial, collaborative and dynamic
nature. For development, software system is further divided in multiple modules. Major modules
include listener module, prediction module for predicting future state of environment, strategy
module for intelligent strategy planning and cooperative play, path planning for individual robots
and transmitter module for wireless communication. Different algorithms are implemented and
tested in different modules and best suitable and functional algorithms are incorporated in the final
system. Parallel processing techniques has been employed to cater for high computational
requirements due to real time complex decision making. System designed and developed in this
project is capable of decision making for multiple agents, path planning to avoid obstacles during
the motion of agents in field and reacting to the change in environment in no time because of
pipelined threaded architecture of system. Thread safety has been given prime importance to make
system fail safe. Multi-threaded application architecture makes it possible to respond to real time
environment by sending commands to hardware agents at a rate of more than 30 commands per
second.
