Abstract :

In this paper we present a Colored Petri Net (CPN) for a multi-agent 
application. In particular we modeled the PacketWorld. The packet-world 
consists of a number of different packets that are scattered over a 
rectangular grid. Agents that live in this virtual world have to collect 
those packets and bring them to their corresponding destination.  
Despite its obvious simplicity the packet-world offers a rich set of 
fundamental characteristics for a broad range of multi-agent systems. 
In order to perform their job in an efficient way the agents have to 
cooperate. Therefore the agents need a number of social skills. We use 
the packet-world as a case to get a deeper insight into the 
fundamentals of agents' social behavior. Our approach is to start from
a very basic model of the agent world and then adding social aspects 
in a gradual way. In view of a flexible integration of new 
functionality, we decided to model the agent world as a combination of 
modules. In this paper we first  present a CPN for a basic model of the 
packet-world. This model consists of agents that can only interact 
through passive objects in the environment. Because interaction is the 
basis of any multi-agent system, we have incorporated basic  
infrastructure for agent coordination straight away into our basic 
model. Then we extend the model, making it possible for the agents 
to communicate information with each other. On the one hand this 
requires a communication module for each agent, on the other hand 
the environment must be extended with a message routing module. 
The advantages of using CPNs are obvious: (i) we get a clear conceptual 
view on the evolution of agents and the environment when we integrate 
social functionality and (ii) CPNs support neat verification and 
formalization of our findings.