啤酒游戏.ppt

《啤酒游戏.ppt》由会员分享，可在线阅读，更多相关《啤酒游戏.ppt（29页珍藏版）》请在淘文阁 - 分享文档赚钱的网站上搜索。

1、Artificial Agents Play the Beer Game Eliminate the Bullwhip Effect and Whip the MBAsSteven O.KimbroughD.-J.WuFang ZhongFMEC,Philadelphia,June 2000;file:beergameslides.ppt2021/9/211The MIT Beer Game PlayersRetailer,Wholesaler,Distributor and Manufacturer.GoalMinimize system-wide(chain)long-run averag

2、e cost.Information sharing:Mail.Demand:Deterministic.CostsHolding cost:$1.00/case/week.Penalty cost:$2.00/case/week.Leadtime:2 weeks physical delay2021/9/212Timing1.New shipments delivered.2.Orders arrive.3.Fill orders plus backlog.4.Decide how much to order.5.Calculate inventory costs.2021/9/213Gam

3、e Board2021/9/214The Bullwhip EffectOrder variability is amplified upstream in the supply chain.Industry examples(P&G,HP).2021/9/215Observed Bullwhip effect from undergraduates game playing 2021/9/216Bullwhip Effect Example(P&G)Lee et al.,1997,Sloan Management Review2021/9/217Analytic Results:Determ

4、inistic DemandAssumptions:Fixed lead time.Players work as a team.Manufacturer has unlimited capacity.“1-1”policy is optimal-order whatever amount is ordered from your customer.2021/9/218Analytic Results:Stochastic Demand(Chen,1999,Management Science)Additional assumptions:Only the Retailer incurs pe

5、nalty cost.Demand distribution is common knowledge.Fixed information lead time.Decreasing holding costs upstream in the chain.Order-up-to(base stock installation)policy is optimal.2021/9/219Agent-Based ApproachAgents work as a team.No agent has knowledge on demand distribution.No information sharing

6、 among agents.Agents learn via genetic algorithms.Fixed or stochastic leadtime.2021/9/2110Research QuestionsCan the agents track the demand?Can the agents eliminate the Bullwhip effect?Can the agents discover the optimal policies if they exist?Can the agents discover reasonably good policies under c

7、omplex scenarios where analytical solutions are not available?2021/9/2111Flowchart2021/9/2112Agents Coding StrategyBit-string representation with fixed length n.Leftmost bit represents the sign of“+”or“-”.The rest bits represent how much to order.Rule“x+1”means“if demand is x then order x+1”.Rule se

8、arch space is 2n-1 1.2021/9/2113Experiment 1a:First CupEnvironment:Deterministic demand with fixed leadtime.Fix the policy of Wholesaler,Distributor and Manufacturer to be“1-1”.Only the Retailer agent learns.Result:Retailer Agent finds“1-1”.2021/9/2114Experiment 1bAll four Agents learn under the env

9、ironment of experiment 1a.ber rule for the team.All four agents find“1-1”.2021/9/2115Result of Experiment 1bAll four agents can find the optimal“1-1”policy2021/9/2116Artificial Agents Whip the MBAs and Undergraduates in Playing the MIT Beer Game2021/9/2117Stability(Experiment 1b)Fix any three agents

10、 to be“1-1”,and allow the fourth agent to learn.The fourth agent minimizes its own long-run average cost rather than the team cost.No agent has any incentive to deviate once the others are playing“1-1”.Therefore“1-1”is apparently Nash.2021/9/2118Experiment 2:Second CupEnvironment:Demand uniformly di

11、stributed between 0,15.Fixed lead time.All four Agents make their own decisions as in experiment 1b.Agents eliminate the Bullwhip effect.Agents find better policies than“1-1”.2021/9/2119Artificial agents eliminate the Bullwhip effect.2021/9/2120Artificial agents discover a better policy than“1-1”whe

12、n facing stochastic demand with penalty costs for all players.2021/9/2121Experiment 3:Third CupEnvironment:Lead time uniformly distributed between 0,4.The rest as in experiment 2.Agents find better policies than“1-1”.No Bullwhip effect.The polices discovered by agents are Nash.2021/9/2122Artificial

13、agents discover better and stable policies than“1-1”when facing stochastic demand and stochastic lead-time.2021/9/2123Artificial Agents are able to eliminate the Bullwhip effect when facing stochastic demand with stochastic leadtime.2021/9/2124Agents learning 2021/9/2125The Columbia Beer GameEnviron

14、ment:Information lead time:(2,2,2,0).Physical lead time:(2,2,2,3).Initial conditions set as Chen(1999).Agents find the optimal policy:order whatever is ordered with time shift,i.e.,Q1=D(t-1),Qi=Qi-1(t li-1).2021/9/2126Ongoing Research:More BeerValue of information sharing.Coordination and cooperatio

15、n.Bargaining and negotiation.Alternative learning mechanisms:Classifier systems.2021/9/2127SummaryAgents are capable of playing the Beer GameTrack demand.Eliminate the Bullwhip effect.Discover the optimal policies if exist.Discover good policies under complex scenarios where analytical solutions not available.Intelligent and agile supply chain.Multi-agent enterprise modeling.2021/9/2128A framework for multi-agent intelligent enterprise modeling 2021/9/2129