CMieux is Carnegie Mellon University’s entry in the Supply Chain Trading Agent Competition. The agent reached the TAC-SCM finals in 2007 and 2008, won the CS50 Exhibition tournament organized in April 2006 and the 2008 edition of TAC-SCM procurement challenge (see July 2008 press release)

CMieux implements adaptive techniques for supply chain procurement and bidding along with an architecture that enables it to tightly coordinate component procurement, production planning and customer bidding activities:

  • Adaptive Bidding: CMieux’s bidding technology revolves around a newly developed ε-optimal solution to a continuous knapsack problem, where a seller with a finite number of goods has to choose discriminatory prices for customers with probabilistic valuations. Customer valuations are constantly updated based on information learned from previous encounters with competing agents. This enables CMieux to constantly adapt its bidding behavior as market conditions change. This technique has been shown to outperform previously proposed bidding solutions.
  • Adaptive Procurement: Our procurement technique is based on a search algorithm that optimizes the utility of procured components, taking into account target demand for finished products, inventory costs associated with holding components for longer than needed, market forecast for component availability and prices as well as the opportunity costs of having mismatched components. By constantly updating its market forecast for both component availability and prices, CMieux is able to anticipate future shortages and take advantage of projected price fluctuations for different components.
  • Tight Coordination: While traditional supply chain management architectures rely on decoupled views of procurement, planning and bidding, CMieux focuses on constantly coordinating these activities. As a result, the agent bids on products it is capable of assembling and procures components it expects to be able to sell quickly and for a high enough profit margin.
© 2001-2010 Norman M. Sadeh