A Broker based architecture to control robots in a simulated environment Luis Moniz 1 , Paulo Urbano 1 , and Helder Coelho 1 Faculdade de Ciˆ encias de Lisboa Campo Grande, 1749-016 Lisboa, Portugal {hal,pub,hcoelho}@di.fc.ul.pt Abstract. We present an extension to the Aglets platform in order to support brokering services agencies in a transparent process for the agent designer. This architecture is constructed to provide a set of functionali- ties, not originally available in the core Aglet environment. These include, white page service, group formation and yellow page support in a dis- tributed and mobile agent environment. The extension was developed in three directions, new environment agents, new agent architectures and an extended agent live cycle model. We also extend the Aglet message model in order to support KQML like directives and speech acts. With these modifications, we encapsulate the Aglets environment in a set of new facilities that provide the designer with a set of higher-level tools to develop his agents. We use this platform to control a team of robots in a simulated environment, providing a complex communication channel to the robots, extending the original features providing a white pages service and a GPS like facility. 1 Introduction The number and variety of agent-based systems have been growing rapidly in the last few years. The advent of e-commerce and the widespread use of Internet made the agents one of the hot technologies in the present, and as a result, the computer software industry is beginning to use this technology to develop its own products. The absence of an industrial norm of what is or could be classified as an agent gives rise to the abuse of the term and concept, and consequently, to the appearance of a wide variety of different heterogeneous and incompatible systems [10, 9, 14]. The main objective of some of these systems is to provide the user with a simulated environment (for instance the Swarm environment [3]) where his agents are to be executed, but almost none provides tools to aid the development of the agents. Other example of these systems is the RoboSoccer [6] environment, although it provides the user with a complex simulated execution environment, it is not simple to directly use the agents developed in this system in a real application, the specificity of the environment limits the application of the agents produced.