Taskable Reactive Agent Communities

The focus of Taskable Reactive Agent Communities (TRAC) project was to develop mixed-initiative technology to enable humans to supervise and manage teams of agents as they perform tasks in dynamic environments. TRAC technology would enable users to task agent communities in a high-level language tha...

Full description

Saved in:
Bibliographic Details
Main Authors: Myers, Karen L, Martin, David L, Morley, David N
Format: Report
Language:English
Subjects:
ADAPTIVE CONTROL SYSTEMS
Administration and Management
AUTOMATION
AUTONOMOUS AGENTS
Computer Programming and Software
COMPUTER PROGRAMS
CONFIGURATION MANAGEMENT
HIGH LEVEL ARCHITECTURE
HIGH LEVEL LANGUAGES
MAN MACHINE SYSTEMS
Operations Research
OPTIMIZATION
PE63760E
PROJECT MANAGEMENT
WUAFRLAGENT007
Online Access:Request full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:The focus of Taskable Reactive Agent Communities (TRAC) project was to develop mixed-initiative technology to enable humans to supervise and manage teams of agents as they perform tasks in dynamic environments. TRAC technology would enable users to task agent communities in a high-level language that provides both descriptions of goals to be satisfied and boundaries on agent behavior. The TRAC approach can be viewed as providing a form of high-level process management technology that enables flexible human control of agent communities. In particular, the TRAC framework is intended to support applications where processes are distributed and automated, but where human guidance can improve performance and increase flexibility. The technical work for the project focused on identifying useful categories of guidance for directing agents, designing languages for expressing guidance, and developing techniques for operationalizing guidance into constraints that influence agent behavior. Because users may provide incompatible or unsatisfiable guidance to agents, the work also encompassed techniques for detecting and resolving conflicting guidance. The TRAC implementation was used as the basis for a demonstration system called TIGER (TRAC Intelligence Gathering and Emergency Response), which focused on the problem of multiagent intelligence gathering in the wake of a simulated natural disaster. Within TIGER, a human supervisor can delegate tasks to agents while providing guidance to control their runtime behavior.