Multi Agent Coordination and Planning

1
Multi Agent Coordination and Planning

• Michael T. Cox
• Computer Science Engineering Department
• Wright State University
• Dayton, OH 45435
• www.cs.wright.edu/mcox

2
Overview

• Agents, DAI, and Planning
• Coordination and Negotiation
• Goal Transformations Planning
• Agent-Based Mixed-Initiative Planning
• Conclusion

3
Agents

• PAGE Characterization
• Perception, Action, Goals, Environment
• Knowledge Representations
• Autonomy and Specialization
• Cognitive Agents
• Social Agents

4
Distributed Artificial Intelligence

• Distributed Artificial Intelligence (DAI) is the
  study, construction, and application of systems
  in which several interacting, intelligent agents
  pursue some set of goals

5
Two Types of DAI Systems

• Multiagent Systems
• Several agents coordinate their knowledge and
  activities by reasoning about the problem solving
  process
• Distributed Problem Solving
• A particular problem is solved by dividing tasks
  among a number of generally equivalent nodes who
  divide and share knowledge about the problem

6
Planning

• Search through Space of Plan Solutions
• Initial State, I, and Goal State, G
• Operator Representations for Actions
• Preconditions
• Effects
• Variable Bindings
• Plan as Sequence of Operators from I to G

7
PlanGoal Tree
8
Critical Issues

• How to enable agents to decompose their tasks and
  goals (and allocate sub-goals and sub-tasks to
  other agents) and synthesize partial results
• How to enable agents to represent and reason
  about the actions, plans, and knowledge of other
  agents in order to interact with them

9
Communication Coordination

• To be effective, multiple agents must coordinate
  and communicate to achieve their goals
• Given goal differences negotiation is a strategy
  for collaborative agents

10
Negotiation

• A process by which a joint decision is reached by
  2 agents, each trying to reach in individual
  goal
• Goal shifts until close enough for agreement

11
Negotiation Protocol
12
Planning as Search
S1
G1
I
S2
G2
G
S3
13
Goal Transformation as Movement in a Goal Space

• Over Predicate Abstraction Hierarchy
• Over Argument Abstraction Hierarchy
• Over Argument Component Hierarchy
• Partonomy
• Over Argument Values
• Enumerated or Numerical

14
The Bridges Problem

• Each river has three bridges
• Plane destroy 1 bridge damage others
• Operationalize Goal of River Impassable
• Operationalize Reduced Transp. Capacity
• Quantify Damage (50 reduction) and Destruction
  (100 reduction)

15
R1
R2
16
Cost/Benefit Analysis

• Requesting Agents Benefit is Value of Achieving
  Goal less Value of Achieving (Trans (Goal))
• Requesting Agents Cost is Value of the Servicing
  Agents Loss

17
Utility Calculation

• Given
• Transformation Factor F
• Goal Priorities P1 P2
• Num Bridges B1 B2
• Num Resources R1 R2
• P1(100B1-R1 (100-F)-P2(B2- R2(100-F)))

18
Goal Transformation Decision
19
R1
G2

• Ground
• g1 Maintain Advantage
• g2 Control R2
• g3 Secure City 1
• Air
• G1 Air Superiority
• G2 Air Interdiction R1
• G3 Ground Support

g2
R2
City 1
g3
G3
20
Task Goals

• What is the trade-off between G2 and G3?
• How does this tradeoff effect g2 and g3?
• Which goal transformation is selected?
• Which resources are reassigned?

21
Future Research

• Change Cost/Benefit to Negotiation
• Agent-Based Mixed-Initiative Planning
• Planning for the Planning Interface

22
Conclusion

• Many Research Areas at Intersection of Agent
  Technology, DAI, and Planning
• Notion of Goal Transformations Helpful
  Integrating These Areas
• More Questions than Answers