Page 1 |
Save page Remove page | Previous | 1 of 117 | Next |
|
small (250x250 max)
medium (500x500 max)
large ( > 500x500)
Full Resolution
All (PDF)
|
This page
All
Subset |
A PLANNER-INDEPENDENT APPROCH TO HUMAN-INTERACTIVE
PLANNING
by
Hyeok-Soo Kim
A Dissertation Presented to the
FACULTY OF THE GRADUATE SCHOOL
UNIVERISITY OF SOUTHERN CALIFORNIA
In Partial Fulfillment of the
Requirements for the Degree
DOCTOR OF PHILOSOPHY
(COMPUTER SCIENCE)
May 2008
Copyright 2008 Hyeok-Soo Kim
Object Description
| Title | A planner-independent approach to human-interactive planning |
| Author | Kim, Hyeok-Soo |
| Author email | hyeoksoo@usc.edu |
| Degree | Doctor of Philosophy |
| Document type | Dissertation |
| Degree program | Computer Science |
| School | Viterbi School of Engineering |
| Date defended/completed | 2007-12-14 |
| Date submitted | 2008 |
| Restricted until | Unrestricted |
| Date published | 2008-03-05 |
| Advisor (committee chair) | Gratch, Jonathan |
| Advisor (committee member) |
Rosenbloom, Paul Narayanan, Shrikanth |
| Abstract | This dissertation introduces a novel approach to the problem of human-interactive planning (HIP), in which a human user and an intelligent planning agent collaboratively develop plans. Many AI researchers have realized the potential of human-in-the-loop in plan development. Not only does joint plan development between human users and planning systems overcome some limitations of fully automated AI planning systems, but it also increases user s trust in the end product as well as complements the intuitive planning skills of human experts with the superior bookkeeping capabilities of automated planning systems.; Putting human users in the loop of plan development, however, creates many additional requirements that differ from what classical AI planning systems require in general, including a communication protocol in which the system and a human user can communicate each other, and additional components (e.g., natural language processing (NLP)) to support these communications. Integrating these different components is also another big issue in developing HIP systems. Due to the fact that they are traditionally studied in isolation, the integration is hard, but the individual update and modification is much harder than that. It causes isolation and disconnection of HIP research from mainstream planning research and contributes to the obsolescence of the planning technologies underlying HIP systems.; This work shows a way to effectively exploit classical AI planning techniques within the context of human-interactive planning. It can be achieved by changing the target of communication. In our approach, the communication of the underlying planning system with other modules, including human users, happens in the surface level, i.e., it's about the inputs to the planning system, not about the detail of inside the planning process. This effort makes it possible to directly utilize classical one-shot planning techniques -- that take a fixed set of goals, initial state, and a domain theory -- to support the hierarchical, gradual, and exploratory nature of human-interactive planning. Because this approach doesn't require to control inside the planning system, it is planner-independent -- in that it could incorporate any classical planning technique. It recasts the problem of human-interactive planning as a search through a space of possible inputs to a classical planning system. This work can bridge HIP and classical AI planning communities that have traditionally evolved in parallel so that they can leverage each other's development and improvement. Although ideally suitable for intelligent agents that must plan interactively and collaboratively with humans, the technique suggests ways to enhance multi-agent planning in general. |
| Keyword | AI planning; human-computer interaction; collaborative planning; mixed-initiative planning |
| Language | English |
| Part of collection | University of Southern California dissertations and theses |
| Publisher (of the original version) | University of Southern California |
| Place of publication (of the original version) | Los Angeles, California |
| Publisher (of the digital version) | University of Southern California. Libraries |
| Type | texts |
| Legacy record ID | usctheses-m1039 |
| Rights | Kim, Hyeok-Soo |
| Repository name | Libraries, University of Southern California |
| Repository address | Los Angeles, California |
| Repository email | http://www.usc.edu/isd/libraries/services/ask_a_librarian/email/ |
| Filename | etd-Kim-20080305 |
| Archival file | uscthesesreloadpub_Volume23/etd-Kim-20080305.pdf |
Description
| Title | Page 1 |
| Full text | A PLANNER-INDEPENDENT APPROCH TO HUMAN-INTERACTIVE PLANNING by Hyeok-Soo Kim A Dissertation Presented to the FACULTY OF THE GRADUATE SCHOOL UNIVERISITY OF SOUTHERN CALIFORNIA In Partial Fulfillment of the Requirements for the Degree DOCTOR OF PHILOSOPHY (COMPUTER SCIENCE) May 2008 Copyright 2008 Hyeok-Soo Kim |
Comments
Post a Comment for Page 1
