[pt] Este trabalho apresenta um sistema multi-agente para negociação em leilões simultâneos de bens relacionados. A dissertação descreve a arquitetura multiagente, como também a análise e desenvolvimento de estratégias para negociação em leilões simultâneos, onde são desejados bens relacionados, em contrapartida com bens isolados. Alguns problemas bem conhecidos em negociação foram identificados no projeto da arquitetura do sistema, tais como predição de preços, alocação de bens, tomadas de decisão, raciocínio sob incerteza e segmentação de demanda. Cada agente que compõe o sistema trata um destes subproblemas. Isto torna possível a aplicação de diferentes técnicas de computação para resolver os subproblemas separadamente e depois combinar as soluções. Utilizou-se o Trading Agent Competition (TAC) para exemplificar as técnicas examinadas. O TAC foi escolhido para testar as heurísticas desenvolvidas por apresentar um conjunto de características que se enquadram adequadamente no domínio em exame. Cada heurística desenvolvida foi testada e seus resultados comparados com edições anteriores do TAC. O sistema multi- agente apresentou uma boa performance em cenários competitivos testados usando o servidor TAC. / [en] This work presents a multi-agent system to trade in simultaneous auctions of related goods. The dissertation describes the multi-agent architecture, and also the analysis and development of strategies for trading in simultaneous auctions, where the purchase of combined goods is required. Some well known problems in trading were identified in order to design the architecture, such as price prediction, good allocation, decision making, reasoning under uncertainty, and demand segmentation. Each agent that composes the system is concerned with one of those trading subproblems. This makes possible to apply different computational techniques to separately solve the subproblems and then combine the solutions. The Trading Agent Competition (TAC) is used to illustrate our approach. TAC was chosen to test the developed heuristics since it presents a set of characteristics that adequately fits the problem domain. Each heurist developed was tested and had its results compared to TAC previous editions. Finally, the system shows a high performance on very competitive scenarios tested by using the TAC server environment.

The thesis contains the theoretical basis for study of possibilities of pricing information goods. The source areas are microeconomics, marketing, competitive advantage and economics of information goods. The model of market is created with constraints defined on the ground of theoretical basis. The thesis analyzes requirements that define the system that supports the choice of pricing strategy. It includes detailed design of the prototype of such system. The prototype is designed with robustness because of the future improvements. The design describes the prototype's input parameters and their transformation into useful outputs that cover basic characteristics of information goods. The designed prototype is implemented. The thesis includes demonstration of the prototype and possible directions for improvements that lead to validity of proposed model of market.