Model Integrated Computing and Autonomous Negotiating Teams for Autonomic Logistics

1
Model Integrated ComputingandAutonomous
Negotiating TeamsforAutonomic Logistics

• G.Karsai (ISIS)
• J. Doyle (MIT)
• G. Bloor (Boeing)

2
Roadmap

• Project domain and goals
• Architecture concept
• Technology reviews Java agent frameworks
• Illustrative demo for framework capabilities
• Modeling (and generation) of agents
• Domain analysis
• Plans

3
(No Transcript)
4
Domain and Goals Autonomic Logistics

• Negotiation techniques for problem solving
  typical in resource management for maintenance
  logistics operations
• Model-integrated approach for component
  generation (agents and interfaces) and system
  integration
• Demonstration on real-life problems

5
MICANTS Concept
MIPS Environment
Models of apps agents etc.
Negotiating a globally beneficial solution
Models
Model Int.
Agent Space
Adapter
Adapter
Adapter
Adapter
Logistics App/Dbase (Legacy)
Logistics App/Dbase (Legacy)
Logistics App/Dbase (Legacy)
Logistics App/Dbase (Legacy)
6
Review and evaluation of relevant
technologiesJava Co.

• Implementation language for network-aware
  computing rich set of libraries
• XML a way for embedding ACL speech-acts
• Jini (simple) service locator mechanism

7
Review and evaluation of relevant
technologiesAgent frameworks
8
Review and evaluation of relevant
technologiesAgent frameworks

• BeeGent (Toshiba)
• Modeling tool for building new agent behaviors
• XML-based communication infrastructure
• MIT MetaGlue
• Infrastructure for the Intelligent Room
• Higher-level organization based on subsumption
  architecture
• Robust communication and service location
  capabilities

9
Review and evaluation of relevant
technologiesZeus Agent Building Toolkit

• Agent framework with predefined agent
  architecture
• Modeling generation tool run-time tools
• Coordination protocols are modifiable
• Predefined Agent building paradigm
• Infrastructural support
• Messaging Service lookup Debugging
• Built-in contract-net (replaceable)
• http//www.labs.bt.com/projects/agents/zeus/index.
  htm

10
Review and evaluation of relevant
technologiesZeus Agent Building Toolkit
11
Review and evaluation of relevant
technologiesZeus Agent Building Toolkit
12
Illustrative Demonstration Example Allocating
and negotiating for a resource
Transfer(8)MT2
Maintenance Task 1
Allocate (1)
Offer(5)
Accept(2)
Reject(6) Accept(7)
Resource
Allocate (3)
Maintenance Task 2
Refuse(4)MT1
13
Modeling (and generation)

• Agent behavior models
• Reactive models (stimulus/response)
• Interaction protocol models (e.g. contract-net)
• Problem-solving process models (planning/schedulin
  g/negotiation phase/etc.)
• Prototype
• Interaction protocol models

14
Interaction Protocol Modeling
15
Domain Analysis

• NAVSUP business processes
• Boeing work on maintenance/logistics support
• Autonomic logistics vision scenario
• Autonomic response to fault events detected
  on-board the aircraft
• Facility location resource allocation scheduling

16
Negotiating Preferences

• Agent preferences visible to owners
• Qualitative model explanations
• Quantitative details and tradeoffs
• Preferences changeable by owners
• Directly in qualitative and quantitative models
• Indirectly through rejection of alternatives
• Learning preferences from negotiation behaviors

17
Plans

• Refinement of the modeling paradigm
• Model interpreters that generate code for the
  framework
• Modeling of legacy apps
• Legacy app integration infrastructure
• Further domain analysis (site visits w/ customer)
• Refinement of demo scenario