background image
being logged in a proprietary manner by the various MMOG vendors. This makes it hard to
develop data mining and CRM approaches that can used for several MMOGs. So the primary
research problem is: How to enable a scalable and flexible simulation environment for
testing out approaches for player usage logging in MMOGs?
.
Since the environment needs to simulate autonomous MMOG participants (i.e. players
and non-personal characters), we choose to use autonomous intelligent agents as the primary
abstraction. Due to several types and numbers of agents we select the Multi-Agent System
(MAS) approach from Distributed Artificial Intelligence as the underlying architecture [17].
The rest of this paper is organized as follows. Section 2 describes and discusses the Zereal
MMOG simulation platform, section 3 describes related work, and finally the conclusion.
2
Zereal Platform
The overall goal is to create a MMOG simulation platform that provides a (coarse) simulation
of active players that can be used to test various approaches for player usage logging.
2.1
Overall Architectural Choices
Creating a realistic activity level with thousands of concurrent players and NPCs is considered
to be of greater importance than having a very accurate simulation of each player's and NPC's
behavior. However, the architecture should also provide a flexible interface for testing out
various intelligence mechanisms.
Requiring support for thousands of concurrent players and NPCs encourages the selection
of a distributed or parallel architecture, this combined with the prior selection of agents as the
main abstraction encourages use of mobile agents for the representation of players and NPCs.
Due to the relative simplicity of implementation and deployment of message-based parallel
systems we choose a platform of type message-based parallel mobile agent architecture. A
second motivation for parallel message-based systems is that they can run on most current
supercomputers and cluster systems. Parallel cluster systems are frequently used to serve
MMOGs, see figure 1 for a typical industry configuration [4, 26, 7].
2.2
Simulation Platform Requirements
The six fundamental requirements for our MMOG simulation platform are:
1. Virtual World Model
2. Item Model
3. Player Models
4. Non-Personal Character Models
5. Scalability
6. Logging of Player behavior
Paper C
81

<< - < - > - >>