Designing rational agents that can work with others

随着万维网和电子商务的到来，软件的趋势是向开放系统、跨应用程序的更多集成以及能够适应变化的系统发展。作为对此的回应，许多开发人员开始采用面向代理的体系结构，其中系统由代理组成，这些自主实体可以灵活地交互，例如通过协商，同时朝着他们的目标工作并对环境中的变化做出反应。然而，如果不通过使用基于知识的体系结构、自动规划等来赋予一些代理一些智能，这种灵活性是无法实现的。在这个项目中，我们将开发新的技术来开发能够与其他代理有效工作的智能代理：关于其他代理的知识和目标的推理方法，基于其他代理执行的操作和上下文识别其他代理的意图和他们正在遵循的计划的方法，以及通过与其他代理有效交互来生成实现其目标的计划的方法。我们还将把这些功能整合到代理编程语言中，以支持构建可以与其他代理一起工作的代理。

With the advent of the World Wide Web and electronic commerce, trends in software are towards open systems, more integration across applications, and systems that can adapt to change. In response to this, many developers are starting to adopt agent-oriented architectures, where a system is composed of agents, autonomous entities that can interact in flexible ways, for instance through negotiation, while working towards their goals and reacting to changes in the environment. However, such flexibility cannot be achieved without imparting some intelligence to some of the agents through the use of knowledge-based architectures, automated planning, etc. In this project, we will develop new techniques for developing intelligent agents that can work effectively with other agents: methods for reasoning about other agents' knowlege and goals, for recognizing the intentions of other agents and what plans they are following based on the actions that they perform and the context, and for generating plans that achieve their goals by interacting effectively with others. We will also incorporate these capabilities into agent programming languages that supports building such agents that can work with others.