RADAR Reflective Agent with Distributed Adaptive Reasoning Automated
RADAR (Reflective Agent with Distributed Adaptive Reasoning) Automated Assistant for Crisis Management www. radar. cs. cmu. edu
Purpose • Automation of office-management tasks, such as scheduling, e-mail handling, and resource allocation • Help not only in routine situations, but also under crisis conditions
Outline • Overview of RADAR • Resource allocation • Future challenges More information • See www. radar. cs. cmu. edu • Talk with Radar researchers
Outline • Overview of RADAR • Resource allocation • Future challenges
PAL video Three-minute video Military-setting motivation for RADAR (Carnegie Mellon) and CALO (SRI).
Project size Largest research project in CMU’s School of Computer Science. • Five departments Language Technologies (LTI) Computer Science Department (CSD) Institute for Software Research International (ISRI) Human-Computer Interaction Institute (HCII) Center for Automated Learning and Discovery (CALD) • Eighty people Twenty-nine faculty members Twenty-seven graduate students Twenty-four others • Five years (2003– 2008)
Project size Largest research project in CMU’s School of Computer Science. Advantages • Multiple research areas • Collaboration opportunities • Potential of a major impact Drawbacks • Coordination challenges • Frequent deliverables
Challenges • Intelligent performance of • office-management tasks Collaboration with a human administrator Dialog with users • • Continuous learning of new • knowledge and strategies Integration of multiple tasks
Research areas • Artificial intelligence • Machine learning • Natural-language processing • Human-computer interaction • Architecture and integration
Main components Planning and co-ordination of the system’s high-level actions.
Main components Web Master Helps create and maintain web sites.
Main components Web Master E-Mail Organizer Helps filter, sort, and prioritize messages.
Main components Web Master E-Mail Organizer Calendar Manager Helps keep track of appointments and negotiate meeting times among multiple users.
Main components Web Master E-Mail Organizer Calendar Manager Briefing Assistant Helps compile reports based on multiple data sources.
Main components Web Master E-Mail Organizer Calendar Manager Briefing Assistant Resource Allocation
Outline • Overview of RADAR • Resource allocation • Future challenges
Purpose Automated allocation of office resources, in both routine and crisis situations. • Assignment of offices • Reservation of conference rooms • Allocation of furniture, computers, and other office equipment
People Faculty Jaime Carbonell jgc@cs. cmu. edu Resource allocation (AI and learning) Eugene Fink e. fink@cs. cmu. edu Resource allocation (AI and learning) Bob Frederking ref@cs. cmu. edu E-mail understanding (Natural language) Grad students Ulas Bardak cyprus@cs. cmu. edu Richard Wang rcwang@cs. cmu. edu Research staff Greg Jorstad gregjor@cs. cmu. edu
RADAR/Space video Six-minute video Initial system for automated assignment of offices.
Initial results A prototype system for automated allocation of offices. • Effective allocation of office resources • Processing of natural-language requests • Interface for a human administrator
Outline • Overview of RADAR • Resource allocation • Future challenges
Motivating task Scheduling of talks at a conference, and related allocation of rooms and equipment, in a crisis situation. • Initial allocation plan • Unexpected major change in • space availability; for example, closing of a building Continuous stream of minor changes; for example, schedule changes and unforeseen equipment needs
Automated reasoning • Temporal reasoning • Uncertainty tolerance • Preference elicitation • Collaboration with a human administrator
Learning Integrated learning of new knowledge and strategies. • From experience • From observation • From instruction
Integration High-level planning Task manager Integrated learning RADAR Resource Allocation RADAR Calendar Manager RADAR E-Mail Organizer RADAR Web Master Users Integrated RADAR Briefing Assistant
Integration User dialog manager Natural language processing Knowledge base and inferences High-level planning Integrated learning RADAR Resource Allocation Users Resource allocation group RADAR Calendar Manager RADAR E-Mail Organizer RADAR Web Master Integrated RADAR Briefing Assistant
Tasks and skills • Development of AI, learning, and natural-language algorithms • Solving open-ended problems • Implementation and integration
More information AI and learning Jaime Carbonell www. cs. cmu. edu/~jgc jgc@cs. cmu. edu Newell Simon Hall 4519 Eugene Fink www. cs. cmu. edu/~eugene e. fink@cs. cmu. edu New Simon Hall 4521 Language Bob Frederking www. cs. cmu. edu/~ref ref@cs. cmu. edu Newell Simon Hall 4617 www. radar. cs. cmu. edu
- Slides: 28