Name: Decision making architectures for bots
You will learn: Java, behaviour modelling
Difficulty: moderate
Excitement: Explore different ways of coding bot’s behaviour!
Description: Programming a bot's behaviour from scratch in Java or other general purpose programming languages is not always the best option. There are several already implemented software architectures that are better suited for expressing the bot's decision making process. Some of these architectures originate from cognitive modelling (e.g. SOAR, ACT-R) others from agent modelling (e.g. 2APL, POSH).

The goal of this work will be to connect one of these architectures to the Pogamut 3 toolkit and to create a non trivial example bot using this architecture.

There are four distinct themes. Each can be done by one student:

Theme 1
SOAR - This project is concerned with integrating the jSoar with Pogamut. The result would be a jSoar-Pogamut agent for UT2004 driven by Soar rules. The focus of work will be in implementating an adapter which will mediate senses/actions between Pogamut and Soar.
Further readings: [[http://sitemaker.umich.edu/soar/home|SOAR]], [[http://code.google.com/p/jsoar/|jSOAR]]

Theme 2
2APL – 2APL is an agent oriented programming language capitalizing on the belief-desire-intention (BDI) paradigm. Similar to previous themes, the focus of this work will be on connecting 2APL to Pogamut, creating an example bot and exploring the limitations of this approach.
Further readings: [[http://www.cs.uu.nl/2apl/|2-APL]]