Estratégias de participação em leilões combinatoriais aplicadas em um problema de transporte de derivados de petróleo

Abstract

Recent researches have shown that approaches based on multi-agent systems (MAS) and market mechanisms like auctions are efficient on the resolution of planning problems in supply chains. This work uses the combinatorial auction-based MAS paradigm and participation strategies in auctions for solving the problem of transporting oil derivatives of PETROBRAS - Petroleo Brasileiro S/A, called Simplified Problem of Transporting Oil Derivatives (SPTOD), which is characterized as a planning problem in supply chains. In combinatorial auctions, the winner determination is a NP-Complete problem without approximation algorithms, whose computational cost increases with the number of bids received by the auctioneer. In this context, this work aims at enlarging the scope of application of combinatorial-auction mechanisms in planning problems by using a heuristic strategy for participation in auctions, besides serving as a support tool for decision-making process by specialists of industrial oil. The results were drawn from several scenarios where the MAS was used with the proposed strategy (global evaluation of needs) and with other strategies for comparison (general – all possible auctions – and greedy – only one auction). The results show that the use of the proposed strategy reduces the processing time when compared to the general strategy and that the quality of the solution is preserved in comparison with the general and greedy strategies. Other contributions of this work are the development of a MAS to realize the planning of transporting oil derivatives between producing and consuming basis having as negotiation model the combinatorial auction-based mechanism and a proposition of a decentralized model where several combinatorial auctions can be run simultaneously.