Information Technology Reference
In-Depth Information
- Finally the creation of the ScriptBasedXXXXAgent , which utilises the parser and
CodeGeneratorVisitor to populate all the agent code into the class, thus allowing
the creation of agents within the language.
- To fully connect the agent type with Agent Factory we must create the XXXXAr-
chitectureFactory which is responsible for automatic creation of agents using the
ScriptBasedXXXXAgent class, automatic creation is achieved through the associ-
ation of the file extension to the agent type.
- A supplementary step in the process is the creation of the MigrationBasedXXXXA-
gent which is similar to the ScriptBasedXXXXAgent in that it allows the creation
of an agent of the language on the system. Rather that through the use of the parser
it uses the state information from another platform to instantiate the agent.
3.5
Eclipse Integration
At this point the language is fully developed and functional, however the final step of
Eclipse integration is recommended as it makes designing agents within the language
easier. As discussed in Section 2.4 there are three components to the eclipse integration,
the builder, editor and new file wizard. An Eclipse plugin is provided which provides
the structure and an example implementation of all the features described.
- The builder is implemented by integrating the parser developed in Section 3.4 and
display the errors through eclipse's built in marker system.
- The editor requires only the modification of the example implementation such that
the correct keywords, labels and operators are highlighted.
- Finally the new file wizard is created through the modification of the example and
the provision of a sample agent file.
4
Evaluation
Scientific evaluation of this system is not an easy task, initial evaluation has been com-
pleted and further evaluation is planned.Informally it is the opinion of the authors that
the provided components simplify the process of developing AOP languages. The ini-
tial evaluation comprised of the use and comparison of two languages created using the
framework during participation in the 2010 Multi Agent Contest 2 [11].
The contest scenario consisted of developing a multi-agent system to solve a cooper-
ative task in a dynamic environment. The environment was a grid-like world in which
virtual cows move around collectively in one or more herds exhibiting a swarm-like be-
havior. There were two corrals, each belongs to one of the two agent teams. The teams
of agents competed to control the behavior of animals and lead them to their own corral.
The winning agent team being the one which scored highest.
The two languages used were AF-TeleoReactive (AF-TR), which is based on Nils
Nilsson's Teleo-Reactive formalism [9], and AF-AgentSpeak (AF-AS), an implementa-
tion of AgentSpeak(L). The scenario is very much suited to having two types of agents,
a leader agent and a herder agent. The CLF allowed us do develop modules which could
2 http://www.multiagentcontest.org/2010
Search WWH ::




Custom Search