A two-stage inference strategy in real-time expert systems

摘要

Expert systems are one of the most active branches in AI applications. Real-time “expert systems” are examples of real-time systems (such as real-time control systems, real-time service systems, etc.). The most important problem is response speed in real-time systems. To solve this problem we present a two-stage inference strategy. The first stage, called the pre-process stage, compiles production rules creating a link-list vector. The second stage, called the real-time inference stage, uses the link-list vector to save the inference time. This strategy has produced obvious effects in our real-time expert system for machine-tool heat error compensation.