Optimization of Real-Time Rule-Based Expert Systems

本研究的重点是发展的科学基础和实施实用的工具，优化和综合基于规则的专家系统，以满足指定的响应时间的限制。 基于规则的实时专家系统是嵌入式人工智能系统，越来越多地用于安全关键应用，如飞机航空电子设备，医疗监测仪器，智能机器人和航天器。 除了功能正确性之外，这些系统还必须满足严格的时序约束。 在这些系统中错过最后期限的结果可能是灾难性的。 如果一个给定的基于规则的系统不能在有限的时间内提供足够的性能，那么它必须被优化或重新合成。 本计画的第一部分探讨了几种优化专家系统规则库的方法（基于状态空间和基于语义）。 该项目的第二部分研究了match阶段的优化，该阶段具有高度不可预测的运行时间。 一个系统的方法来解决优化问题的发展开辟了新的途径，以进一步提高基于规则的系统在时间关键的环境中的运行时性能。 这项技术将在未来的复杂信息和计算系统中发挥关键作用，如多媒体工具、虚拟现实系统、智能机器人和高带宽网络，其中智能接口是必不可少的，时间成为越来越重要的资源。

This research focuses on developing the scientific foundation and implementing practical tools for optimizing and synthesizing rule-based expert systems to meet specified response time constraints. Real-time rule-based expert systems are embedded artificial intelligence systems increasingly used in safety-critical applications such as airplane avionics, medical monitoring instruments, smart robots, and space vehicles. In addition to functional correctness, these systems must also satisfy stringent timing constraints. The result of missing a deadline in these systems may be catastrophic. If a given rule-based system cannot deliver an adequate performance in bounded time, then it has to be optimized or resynthesized. The first part of the project investigates several approaches (state-space-based and semantics-based) for optimizing the rule base of expert systems. The second part of the project investigates the optimization of the match phase, which has a highly unpredictable runtime. The development of a systematic methodology to tackle the optimization problem opens up new avenues to further enhance the runtime performance of rule-based systems in time-critical environments. This technology will play a key role in tomorrow's complex information and computing systems such as multimedia tools, virtual reality systems, smart robots, and high-bandwidth networks, where an intelligent interface is essential and time becomes an increasingly critical resource.