Please use this identifier to cite or link to this item:
|Title:||Modelling Uncertainty in 3APL|
|Publisher:||Open Universiteit Nederland|
|Abstract:||In this thesis, the author investigates how the agent programming language 3APL can be enhanced to model uncertainty. Typically, agent programming languages such as 3APL that are based on beliefs, goals and intentions use logical formulae to represent their beliefs and reason on them. These formulae are either true or false (i.e. they are believed or not), and this limits the use of such agent programming languages in practical applications. While a lot of research has been done on the topic of reasoning with uncertainty the possible use of these methods in the field of agent programming languages such as 3APL has not been given much attention. The author investigates several methods (with a focus on Bayesian networks and Dempster-Shafer theory), and show that Dempster-Shafer theory is a promising method to use in agent programming. Particulary appealing in this theory is the ability to model ignorance, as well as uncertainty. Nevertheless, the combinatorial explosion of its combination rule and the issue of inconsistency (which are addressed in the thesis) are serious disadvantages of this theory for its practical application to agent programming. The author investigates a possible mapping of Dempster-Shafer sets to belief formulae in 3APL. With restrictions on the mass functions and on the frame of discernment, Dempster-Shafer theory is a convenient way to model uncertainty in agent beliefs. Because, with certain restrictions, mass values can be computed based on the beliefs in the belief base, we do not need to keep a combined mass function of n beliefs in memory and update it with each belief update. Therefore there is no combinational explosion. The author proposes a syntax and semantics for 3APL with uncertainty, and demonstrate a prototype Prolog implementation of the calculation of the certainty of a logical formula given a certain belief base.|
|Appears in Collections:||MSc Computer Science|
Files in This Item:
|INF_20050801_kwisthout.pdf||740.89 kB||Adobe PDF||View/Open|
Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.