A Decentralized Agent-Based Platform For Automated Trade And Its Simulation

For the last few years the internet has profoundly affected the retail trade of standarized consumer goods, from books and CDs to intercontinental flights. For a single product, consumers enjoy a wide choice of offers and can easily compare prices making use of the decreased effort of information retrieval and of software agent technology. But the interaction between the consumer (or his agent) and retailer is tipically very limited: it mainly consists of obtaining price statements and eventually sending orders. For the future we envision a much more sophisticated trade on the internet benefiting both, consumers and retailers: agents entering into actual negotiations with each other would be able to act on consumers' or retailers' behalf, locating specific products or variants, discussing terms of delivery or special conditions, and performing the transactions automatically, based on their owners's preferences. An extended model of this is presented in MarketSpace[1].We believe this automated retail trade will be based on a decentralized dynamic system of autonomous software agents with asynchronous mechanisms for communication and transactions. In the present work, our first aim is to propose a design for such an agent system, discuss its specific structural demands and develop a technical concept for its implementation. Based on the simulation system DMarks[4] we develop an integrative project DMarks II consisting of: 1. a prototypical implementation for studing this general automated trading system, using Java technology; 2. an extension of this prototype to a distributed multi-agent simulator for market scenarios.Using the DMarks II simulator we then examine developments arising from widespread application of the DMarks II platform prototype in real markets. Namely, we want to shed light on price developments and the power of seller strategies of different complexity in a mixed buyer population, with an increasing fraction of fully informed buyers. Here we make use of results achieved by Greenwald and Kephart[2,3].