Resource-Bounded Reasoning in Intelligent Systems: New Directions & Opportunities

资源受限推理的新兴领域涉及灵活计算方法的开发、组合和元级控制，这些方法允许用少量资源(如时间、内存或信息)来换取计算结果的价值收益。更好地理解这些技术对于构建用于自动计划、诊断、决策和控制的有用的智能系统至关重要。这项研究项目有两个主要组成部分。一个是开发资源受限的技术，用于从异类、动态的信息源网络中自动收集信息。该方法扩展了经典的信息价值理论，将信息可靠性、检索成本、实时操作和不确定性等因素的影响形式化。第二个密切相关的活动通过构建能够在不确定和可变计算约束下实时运行的计划系统来改进当前的计划技术。

The emerging field of resource-bounded reasoning is concerned with the development, composition, and meta-level control of flexible computational methods that allow small quantities of resources, such as time, memory, or information, to be traded for gains in the value of computed results. A better understanding of these techniques is essential for the construction of useful intelligent systems for automated planning, diagnosis, decision making, and control. There are two primary components to this research project. One is the development of resource-bounded techniques for automated information gathering from a network of heterogeneous, dynamic information sources. The approach extends the classical theory of information value by formalizing the effect of such factors as information reliability, retrieval costs, real-time operation, and uncertainty. A second and closely related activity improves current planning technology by constructing planning systems that can operate in real-time under uncertainty and under variable computational constraints.