Skip to content Home Contact Mobile MyRMIT Library A-Z
RMIT UniversityResearch Repository
 

Maintenance goals in intelligent agents

Duff, S 2009, Maintenance goals in intelligent agents, Masters Thesis, School of Computer Science and Information Technology, RMIT University.

Document type: Thesis
Collection: Theses
Attached Files
Name Description MIMEType Size Downloads
Duff.pdf Thesis application/pdf 578.82KB 135

Title Maintenance goals in intelligent agents
Author(s) Duff, S
Year 2009
Abstract One popular software development strategy is that of intelligent agent systems. Agents are often programmed by goals; a programmer or user defines a set of goals for an agent, and then the agent is left to determine how best to complete the goals assigned to them. Popular types of goals are achievement and maintenance goals.

An achievement goal describes some particular state the agent would like to bring about, for example, being in a particular location or having a particular bank balance. Given an achievement goal, an agent will perform actions that it believes will lead it to having the achievement goal realised.

In current agent systems, maintenance goals tell an agent to ensure that some condition is always kept satisfied, for example, ensuring that a vehicle stays below a certain speed, or that it has sufficient fuel in its fuel tank. Currently, maintenance goals are reactive, in that they are not considered until after the maintenance condition has been violated. Only then does the agent begin to perform actions to restore the maintenance condition.

In this thesis, we have discussed methods by which maintenance goals can be made proactive. Proactive maintenance goals may cause an agent to perform actions before a maintenance condition is violated, when it can predict that a maintenance condition will be violated in the future. This can be due to changes to the environment, or more interestingly, when the agent itself is performing actions that will cause the violation of the maintenance condition. Operational semantics that clearly demonstrate the functionality and operation of proactive maintenance goals have been developed in this thesis. We have experimentally shown that agents with proactive maintenance goals will reduce the amount of resources consumed in a variety of error-prone environments. This includes scenarios where the agent's beliefs are less than the true values, as well as when the beliefs are in excess of the true values.
Degree Masters Thesis
Institution RMIT University
School, Department or Centre School of Computer Science and Information Technology
Keyword(s) intelligent agents
maintenance goals
prediction
artificial intelligence
 
Versions
Version Filter Type
Access Statistics: 79 Abstract Views, 135 File Downloads  -  Detailed Statistics
Created: Thu, 25 Nov 2010, 16:45:42 EST