David Pardoe Peter Stone The University of Texas at Austin Department of Computer Sciences...

32
David Pardoe Peter Stone The University of Texas at Austin Department of Computer Sciences TacTex-05: A Champion Supply Chain Management Agent

Transcript of David Pardoe Peter Stone The University of Texas at Austin Department of Computer Sciences...

Page 1: David Pardoe Peter Stone The University of Texas at Austin Department of Computer Sciences TacTex-05: A Champion Supply Chain Management Agent.

David Pardoe

Peter Stone

The University of Texas at AustinDepartment of Computer Sciences

TacTex-05: A Champion Supply Chain Management Agent

Page 2: David Pardoe Peter Stone The University of Texas at Austin Department of Computer Sciences TacTex-05: A Champion Supply Chain Management Agent.

Supply Chain Management

• Research goal: automate the process• Trading Agent Competition (TAC SCM)• Many challenges• TacTex-05 (2005 winner) - agent composed of

several interacting components:– prediction

– optimization

– adaptation

Page 3: David Pardoe Peter Stone The University of Texas at Austin Department of Computer Sciences TacTex-05: A Champion Supply Chain Management Agent.

Outline

• Summary of TAC SCM

• TacTex-05 agent design

• Adaptive aspects of TacTex-05

• Competition results and experiments

• Conclusion

Page 4: David Pardoe Peter Stone The University of Texas at Austin Department of Computer Sciences TacTex-05: A Champion Supply Chain Management Agent.

TAC SCM

• Agents compete as manufacturers

• 220 simulated days per game (15s each)

Page 5: David Pardoe Peter Stone The University of Texas at Austin Department of Computer Sciences TacTex-05: A Champion Supply Chain Management Agent.

Component Procurement

• Supplier’s production capacity fluctuates

• Prices depend on supplier’s free capacity

Page 6: David Pardoe Peter Stone The University of Texas at Austin Department of Computer Sciences TacTex-05: A Champion Supply Chain Management Agent.

Customer Negotiation

• 16 computer types in 3 segments

• Daily number of RFQs fluctuates

Page 7: David Pardoe Peter Stone The University of Texas at Austin Department of Computer Sciences TacTex-05: A Champion Supply Chain Management Agent.

Factory Scheduling

• Limited production capacity

• Daily storage cost for all inventory

Page 8: David Pardoe Peter Stone The University of Texas at Austin Department of Computer Sciences TacTex-05: A Champion Supply Chain Management Agent.
Page 9: David Pardoe Peter Stone The University of Texas at Austin Department of Computer Sciences TacTex-05: A Champion Supply Chain Management Agent.

Outline

• Summary of TAC SCM

• TacTex-05 agent design

• Adaptive aspects of TacTex-05

• Competition results and experiments

• Conclusion

Page 10: David Pardoe Peter Stone The University of Texas at Austin Department of Computer Sciences TacTex-05: A Champion Supply Chain Management Agent.
Page 11: David Pardoe Peter Stone The University of Texas at Austin Department of Computer Sciences TacTex-05: A Champion Supply Chain Management Agent.

Demand Model

• Goal: predict future customer demand

• Bayesian approach adapted from DeepMaize (Kiekintveld et al. 2004)

Page 12: David Pardoe Peter Stone The University of Texas at Austin Department of Computer Sciences TacTex-05: A Champion Supply Chain Management Agent.

Order Probability Predictor

• Want to predict P(order | offer price)

• Linear predictor for each computer type

Page 13: David Pardoe Peter Stone The University of Texas at Austin Department of Computer Sciences TacTex-05: A Champion Supply Chain Management Agent.

Demand Manager

• Given resources and predictions, determine:– production schedule– deliveries– offers on all of today’s RFQs

• All done with greedy scheduling algorithm

Page 14: David Pardoe Peter Stone The University of Texas at Austin Department of Computer Sciences TacTex-05: A Champion Supply Chain Management Agent.
Page 15: David Pardoe Peter Stone The University of Texas at Austin Department of Computer Sciences TacTex-05: A Champion Supply Chain Management Agent.
Page 16: David Pardoe Peter Stone The University of Texas at Austin Department of Computer Sciences TacTex-05: A Champion Supply Chain Management Agent.
Page 17: David Pardoe Peter Stone The University of Texas at Austin Department of Computer Sciences TacTex-05: A Champion Supply Chain Management Agent.

Supplier Model• Estimate each supplier’s free capacity from offers• Use estimates to predict future offer prices

Page 18: David Pardoe Peter Stone The University of Texas at Austin Department of Computer Sciences TacTex-05: A Champion Supply Chain Management Agent.

Supply Manager: What to Order

• Goal: maintain a threshold inventory

Page 19: David Pardoe Peter Stone The University of Texas at Austin Department of Computer Sciences TacTex-05: A Champion Supply Chain Management Agent.

Supply Manager: When to Order

• Given a desired delivery, when to send RFQ?

• Assume today’s price pattern holds

Page 20: David Pardoe Peter Stone The University of Texas at Austin Department of Computer Sciences TacTex-05: A Champion Supply Chain Management Agent.

Outline

• Summary of TAC SCM

• TacTex-05 agent design

• Adaptive aspects of TacTex-05

• Competition results and experiments

• Conclusion

Page 21: David Pardoe Peter Stone The University of Texas at Austin Department of Computer Sciences TacTex-05: A Champion Supply Chain Management Agent.

Adaptation

• Different opponents lead to different situations

• Adapt by modifying predictions

• Make use of game logs

Page 22: David Pardoe Peter Stone The University of Texas at Austin Department of Computer Sciences TacTex-05: A Champion Supply Chain Management Agent.

Two Areas of Adaptation

• Initial orders and endgame sales

• Important, but difficult to reason about

• Agents may handle as special cases

• Update predictions during these periods

Page 23: David Pardoe Peter Stone The University of Texas at Austin Department of Computer Sciences TacTex-05: A Champion Supply Chain Management Agent.

Outline

• Summary of TAC SCM

• TacTex-05 agent design

• Adaptive aspects of TacTex-05

• Competition results and experiments

• Conclusion

Page 24: David Pardoe Peter Stone The University of Texas at Austin Department of Computer Sciences TacTex-05: A Champion Supply Chain Management Agent.

Final Results

• Adaptation important:– ordered 95,000 components on first day

– SouthamptonSCM: 22,000; Mertacor: 18,000

Page 25: David Pardoe Peter Stone The University of Texas at Austin Department of Computer Sciences TacTex-05: A Champion Supply Chain Management Agent.
Page 26: David Pardoe Peter Stone The University of Texas at Austin Department of Computer Sciences TacTex-05: A Champion Supply Chain Management Agent.

Experiments

• Experiments analyzing agent components

• Use TAC Agent Repository

• Compare modified versions of TacTex-05

• Test adaptation against different opponents

Page 27: David Pardoe Peter Stone The University of Texas at Austin Department of Computer Sciences TacTex-05: A Champion Supply Chain Management Agent.

Results

• Start-game adaptation– competition results very atypical

• End-game adaptation– beats fixed strategies in experiments

• Predictive models:– supplier price predictions most important

• Often better to wait to order components– tradeoff: price vs demand certainty

Page 28: David Pardoe Peter Stone The University of Texas at Austin Department of Computer Sciences TacTex-05: A Champion Supply Chain Management Agent.

Outline

• Summary of TAC SCM

• TacTex-05 agent design

• Adaptive aspects of TacTex-05

• Competition results and experiments

• Conclusion

Page 29: David Pardoe Peter Stone The University of Texas at Austin Department of Computer Sciences TacTex-05: A Champion Supply Chain Management Agent.

Related Work

• Many TAC SCM agent descriptions– SouthamptonSCM – He et al. 2006– Mertacor – Kontogounis et al. 2006– DeepMaize – Kiekintveld et al. 2006– CMieux – Benisch et al. 2006

• Available from TAC website http://www.sics.se/tac

Page 30: David Pardoe Peter Stone The University of Texas at Austin Department of Computer Sciences TacTex-05: A Champion Supply Chain Management Agent.

TAC News

• 2006 TAC SCM competition complete– Won by TacTex-06– Most important addition: use learning to

predict future changes in computer prices

• TAC in 2007: 3 games– TAC Classic– TAC SCM– New market design game

Page 31: David Pardoe Peter Stone The University of Texas at Austin Department of Computer Sciences TacTex-05: A Champion Supply Chain Management Agent.

Conclusion

• Introduced TAC SCM• Described TacTex-05

– prediction– optimization– adaptation

• Future work– additional learning, adaptation– focus on component price prediction, ordering

Page 32: David Pardoe Peter Stone The University of Texas at Austin Department of Computer Sciences TacTex-05: A Champion Supply Chain Management Agent.

Thank You!