Behavior composition optimisation

Yadav, N 2014, Behavior composition optimisation, Doctor of Philosophy (PhD), Computer Science and Information Technology, RMIT University.


Document type: Thesis
Collection: Theses

Attached Files
Name Description MIMEType Size
Yadav.pdf Thesis application/pdf 1.58MB
Title Behavior composition optimisation
Author(s) Yadav, N
Year 2014
Abstract The behavior composition problem involves automatic synthesis of a controller that is able to “realize” (i.e., implement) a desired target specification by suitably controlling a collection of already available, partially controllable, behaviors running in a partially predictable shared environment. A behavior in our context refers to an already existing functionality such as the logic of a device, a service, a standalone component, etc; whereas a target specification represents the desired non-existent functionality that is meant to be obtained through the available behaviors. Previous work in behavior composition has exclusively aimed at synthesising exact controllers, those that bring about the desired specification completely. One open issue has resisted principled solutions: if the target specification cannot be completely implemented, is there a way to realize it “optimally”? In this doctoral thesis, we propose qualitative and quantitative optimisation frameworks that are able to accommodate composition problems that do not admit the “perfect” coordinating controller. In the qualitative setting, we rely on the formal notion of simulation to define realizable fragments of a target specification and show the existence of a unique supremal realizable fragment for a given problem instance. In addition, we extend the qualitative framework by introducing exogenous uncontrollable events to represent observable contingencies. In the quantitative setting, we provide a decision theoretic approach to behavior composition by quantifying the uncertainties present in the domain. In all cases, we provide effective techniques to compute optimal solutions and study their computational properties.
Degree Doctor of Philosophy (PhD)
Institution RMIT University
School, Department or Centre Computer Science and Information Technology
Keyword(s) Behavior Composition Problem
Automatic Synthesis
Versions
Version Filter Type
Access Statistics: 221 Abstract Views, 201 File Downloads  -  Detailed Statistics
Created: Fri, 19 Sep 2014, 17:29:07 EST by Maria Lombardo
© 2014 RMIT Research Repository • Powered by Fez SoftwareContact us