Application of Agent Technology Dr Thomas E Potok

Application of Agent Technology Dr. Thomas E. Potok Collaborative Technologies Research Center Computer Science and Mathematics Division Oak Ridge National Laboratory Lockheed Martin Energy Research

Collaborative Technologies Research Center (CTRC) } Computer Science and Mathematics Division } Pioneering research in – Agent technology – Information integration – Cluster analysis – Software engineering } We have successfully developed systems for Lockheed Martin, the Department of Energy, and the Defense Logistics Agency. } Approach – Small Entrepreneurial team of researchers and software developers – Broad range of collaborators, including • LM, DLA, DOE, University of Tenn, NIST, CMU, NCSU, NTRC 2

Key Projects } SURGE - Supplier Utilization through Responsive Grouped Enterprises – DLA funded to drastically reduce cost/delivery time for military spares – Software agents and grouping technology used to define part families } MABES - Manufacturing Agent Based Emulation System – LMTAS to rapidly model fundamental changes to manufacturing systems – Software agents to analyze the impact of changes to manufacturing lines } CME - Collaborative Management Environment – DOE funded to provide significant improvement in research funding – Information integration used to gather, search over, and report on heterogeneous information from a number of national laboratories 3

Future Technology Trend Internet Telephone Face to Face 4 GE A TS N

Successful Projects }We have extensive expertise in agent development }Began working with agent technologies in 1980 s Supply Chain Management Agent System Manufacturing Emulation Agent System Collaborative Decision Support System Neural Nets for Recovery Boiler Control Neural Nets for Bankruptcy Prediction Neural Nets for Spring-back Prediction Collaborative Design System Neural Nets for Resistance. Spot Welding Neural Nets for Material Mix Optimization Genetic Algorithms for Chemical Synthesis Knowledge-based Systems - Manufacturing Advisors }Over 10 successful projects within the last 5 years Knowledge-based Systems for Constructability }Collaborations with leading Design and Analysis of Computer Experiments Knowledge-based Computer Systems Calibration 1985 5 1990 1995 agent experts 2000

Recent Accomplishments } Guest researcher at NIST for standardization of agent frameworks } Delivered a multi-agent part grouping system for DLA } Press Release: Lockheed Martin Completes First Phase in Applying 'Agent-Based' Software to JSF } Papers and Presentations – – Invited presentation to MIT’s Lean Aerospace Initiative Forum An invited paper to the IEEE Internet Computing Journal Presented two multi-agent papers at the ISAS'99 conference Paper accepted by International Journal of Flexible Automation and Integrated Manufacturing. – Paper accepted by Flexible Automation and Intelligent Manufacturing Conference 6

What are Agents? . . . Software entities that assist people and act on their behalf. . . Software “robots” Traditional Software Object Proactive detect changes in their environment and react to those in a timely manner by answering to events and initiating actions Behavior Goal-driven have a purpose and act in accordance with that purpose until it is fulfilled Communicative able to interact and communicate with users and other agents State Autonomous Learning have the ability to learn from experiences in their environment 7 can have control over their own actions and be able to work and launch actions independent of the user or other actors

Simple Agent Example Agent, find me the book “War and Peace, ” and I need it tomorrow Amazon 2 Days $18. 50 Barnes and Nobel 1 Day $21. 75 Borders NA . . . Does the agent understand buying books? Form a plan to buy the book Execute the plan B. Dalton Learn for next time 8 1 Day $20. 25 Order the book

SURGE/MABES

Process Overview Projected or Actual Parts Need Group Parts Form Lean Cells 01

SURGE Grouping Agents Agent Group Themselves Part Families Agent Mediator Parts represented by Agents 11

Results on C-130 Parts Input Data Grouping Results ro G ou Gr up 5 Group 1 G ro ro G up 4 p 2 up 3 12

C 130 Grouping Results Why two groups? Common Processes Group “ 3” Processes Group “ 5” Processes Why are extrusions spread over two groups? 13

Results Cell 1 Group 2 Group 1 Gr oup 2 up ro G 3 Cell 2 Group 1 Group 3 Wire Harness Data forms 3 groups 14 Cell 3 Possible cell layout

SURGE 1 st Phase Results } Investment Total Savings – Initial investment $3. 3 M – Agent investment $812 K of the $3. 3 M – 1 st phase duration of 9 months } Return Agent Investment { – 7, 832 of ~130 K spare parts grouped, 4400 Parts Bid – $7. 0 M savings in Inventory Reduction (30%) – $5. 8 M Savings in reduced pricing (23%) – 58% Reduction in lead times (from 220 to 93 days) } Total – $12. 8 M in savings – Significant reduction in lead times 15

Structured Text Application for DOE p 5 ro Natural groups 3 rou p 4 Group 2 16 up Groups by Program Managers G Group 1 rou G G DOE Prog. Mgr 1 DOE Prog. Mgr 2 DOE Prog. Mgr 3 DOE Prog. Mgr 4 DOE Prog. Mgr 5

MABES: Analytic Model Manufacturing Rules Minimization Throughput At 17 Planes Per month: - Machine utilization 85% - Cost is $65 M - Span time is 48 Days Process Metrics 17 Activity Metrics

MABES: Dynamic Model Queues and Task Centers Pull/Push/Takt Animation 18 Identify • Bottlenecks • Network Sensitivity • Outages

Visibility into the Supply Chain Manufacturing Operations Theater Operations PARTNER SUPPLIER 3 rd TIER SISTER DIVISION SUPPLIER 2 nd TIER SUPPLIER PARTNER FABRICATION - SUBASSEMBLY - FINAL - DELIVERY 1 st TIER SUPPLIER How does a problem here affect operations here 19

SURGE/MABES Summary } An advanced logistics optimization system } Significant research breakthroughs in clustering technology } Provides significant savings and leadtime reductions to DLA 20

The Collaborative Management Environment Database Management Human Computer Interaction Meta-modeling Languages Object-oriented Technologies Scalable Algorithms Security Collaborative Management Environment Information Integration 21 XML Ames Lab Berkeley Lab Fermi Lab Los Alamos Lab Sandia Lab Livermore Lab Oak Ridge Lab Software Engineering

Current situation } Problem – Field Work Proposals (FWPs) submitted to DOE are in paper books – Weeks and thousands of dollars are spend in collating, copying, binding, and shipping these books – The books provide very limited query and search capability } Approach – Developed an FWP “ontology” for several national laboratories – Pioneered use of the Extended Markup Language (XML) as a means of storing, querying, and presenting FWP information. • Simple data storage technology • Very low costs to the labs, integration work done by CME team • Very well received article at XML’ 98, In. Forum’ 99, Inter. Lab’ 99. 22

Sample FWP 23

Same XML Data Type Definition Tag Definitions 24 Tagged Document

CME System 25

CME Summary } One common picture to DOE } Integrated, collaborative, and distributed information in a secure web-based environment } Innovation – Use of Extended Markup Language (XML) for low-cost information integration – Staged Schema migration CME information model } DOE evaluating CME to be a corporate system 26

Summary – We have expertise and experience with developing • Advanced agent-based systems • Collaborative decision support systems – We are pioneering in agent and information integration technologies – Proven track record of success 27