Organization of Multi-Agent Teams

1
Organization of Multi-Agent Teams

• Eric Matson
• Multi-agent and Cooperative Robotics Lab
• Department of Computing and Information Sciences
• Kansas State University
• 234 Nichols Hall, Manhattan, KS 66506
• matson_at_cis.ksu.edu

2
Agenda

• Introduction
• Organization
• Agent Organization Model
• Domains and Implementation
• Conclusions
• Summary
• Current and Future Research
• Questions and Answers

3
Introduction

• What is organization?
• defined an organization as a task applied to
  the efforts of two or more people, a system of
  communication, a means of problem solving, and a
  means of facilitating decision making Dale 67.
  Dale goes on to further describe organization
• (1) Determining what must be done if a given aim
  is to be achieved.
• (2) Dividing the necessary activities into
  segments small enough to be performed by one
  person.
• (3) Providing a means of coordination, so that
  there is no wasted effort and members of the
  organization do no get into each other's way.
• Why is organization important?
• How can we use agent organizations?

4
Organization

1. The mission goal must be defined from an
   intelligent agent or human source.
2. The goal is used to generate plans to satisfy all
   goals.
3. The plans are used to dynamically create
   decomposition structures to develop the
   definition of roles, rules, relationships and
   tasks.
4. The available members skills and capabilities
   are inventoried and considered.
5. The roles are matched to agents, based on
   capabilities, to roles to accomplish goals.
6. The organization begins to work for
   accomplishment of goals

5
Implementation Goal

• Develop organizational model for implementing a
  sustainable cooperative agent team capable of
  long term execution
• Self Organization and Reorganization
• Self Reconfiguration
• Cooperative
• Sustainable
• Repair
• Adaptable

6
Agent Organization Model

• Structure
• Elements of an organization
• Agents, Roles, Goals, Laws
• State
• Instance of an organization
• Transition
• State to State
• Initial Organization
• Re-organization

7
Structure
8
State
Reorganization Instances
Goals Roles Agents Capability Relationships Laws


Goal Satisfaction
State 0
State 1
State 2
State N
Organization Duration Initial State -gt Goal
Satisfaction
9
Transition

• Function of transition from one state to a new
  state
• Examples
• Hiring a new person for a development team
• Substituting a player during a basketball game
• Otransition (O,F,d,Sn,Soptimal,Spossible)
• Transition function d(O,F,Sn) ? S'

10
Example
Goal Search and Rescue
Sonar
Role Rescuer Must move to where victim is
located, pick up and return victim to base
location
Role Capability Requirements
Active Grasp
Roll
Query Who can play role?
NO
YES
Robot 1
Robot 2
Roll
Sonar
Roll
Sonar
Active Grasp
Agent Capability
11
Problem Domains
Organizational Model
Generic
Robotic Teams
EII
AIS
Human
Battle Intelligence
Cooperative
Data Integration
Current Projects
Future Projects
Adaptive
Specific
12
AIS Overview

• Tactical and Strategic Battlefield Intelligence
• Timely
• Continuous or Ad Hoc
• Sensor Networks to Provide Information
• Sensor Loss Does Not Reduce the Information
  Requirements
• Adaptive Information System (AIS) must
• Overcome the Loss of Sensor Instances
• Provide Continuous Information Flow

13
AIS Implementation

• Battlefield Contains Many Sensor Types
• Human
• Mechanical
• Organize, Synthesize and Merge Streams
• Structure
• Agent Types
• Data Sensor
• Synthesis
• Query
• Query
• Self Organization

14
AIS Conclusions

• Combination of AIS and Organization Models can
• Provide a Flexible System
• Overcome Sensor Loss
• Compensate Through Self-Organization
• Simulation (100K Scenarios)
• Worked to reorganize when possible
• Actual Battlefield
• Less success due to Physical

15
Robotic Teams
agentTool agentMom
Complex Capabilities And Teamwork
Environmental Learning
Behaviors
Higher Level Function
In Progress
Organization Computation
Capability Abstraction Layer
Integration Connectors
Existing
Generic Integration Layer
Java
Java
Java
Java
Java
Aria C
Aria C
C
Mini Java
??
ActivMedia AmigoBot
ActivMedia Pioneer
Parallax Javelin
Nomadic
16
Robot Team Conclusions

• Agent Organization can be applied to robotics
• Self-reconfiguration
• Adaptability
• Sustainable
• Architecture supports many robotic instances,
  regardless of implementation

17
Current Research

• Capability classification of robotic agents
• Computational
• Physical
• Integrate
• Concepts of computational organization theory
  into teamwork theories
• Methods for developing teams of cooperative
  robots.
• Organizational rules to define proper
  organizational function.
• Extend general-purpose reorganization algorithms
• Incorporate organizational design and knowledge
  into agent-oriented software methods
• Multi-agent systems design
• Application to Cooperative Robotics.

18
Future Research

• Effects of attrition/recovery on sustainable
  teams (current model)
• Decision points for organizations
• Goal satisfaction
• Goal relaxation
• Effective reorganization function
• Structural composition of capabilities
• Capability synthesis algorithms
• Developing capabilities for one (or more) robotic
  team member to
• Repair
• Restore
• Mixed-initiative programs
• Employment of human agents, as part of the
  cooperative organization
• Human capability to elevate cooperative goal
  satisfaction strategies

19
Summary

• Goal is to implement agent teams using
  organizational model with the following
  characteristics
• Recoverable
• Adaptable
• Sustainable
• Flexibility with broad domain application
• Adaptive Information Systems (AIS)
• Robotic teams
• Current working architecture to support the model
• Developing higher level organizational function
• Environmental learning
• Reasoning about effective organization function

20
Questions
?