提案-验证通用推理及在铁路选线中的应用

Proposing-Testing Generic Reasoning and Its Application in Railway Location

为开发可直接利用软件工程中事实知识并能在智能铁路选线系统中重用的推理机,提出并实现了提案-验证通用推理模型.用面向对象技术表示知识,使用规则层次模型,把知识表示为事实知识、约束知识、启发知识、策略知识和目标验证知识;采用双层形式化模型,把知识可阅读和可执行形式联系起来;把问题求解知识表示为推理控制知识,用目标验证知识描述任务目标,实现推理机与控制策略分离.推理基于数据驱动方式,利用反射技术实现动态模式匹配和规则执行;用分类组织知识和按领域特征排列知识对象的冲突消解机制,使推理有序进行.提出的方法已成功应用于新建铁路的线路平面自动生成.

To develop an inference engine that can be reused in an intelligent railway location system and directly utilize factual knowledge in software engineering, a proposing-testing generic inference model was put forward. In this model, knowledge is represented using the object-oriented technology. By using a hierarchical model, rules are described as factual knowledge, constraint knowledge, heuristic knowledge, tactful knowledge and goal verifying knowledge, knowledge＇s executable form is related to its readable form by using a bi-formal model. Problem-solving method is described as inferential control knowledge, and task verifying knowledge is used to describe task goal. So reasoning control strategy can be separated from an inference engine. The inference engine takes data-driven as its control strategy. Dynamic pattern matching and rule executing are fulfilled using the reflection technology. By classifying knowledge based on knowledge categories and knowledge domain characteristics, conflict resolution mechanism is constructed. Therefore, the reasoning process is controllable. This method has been successfully used in railway location to automatically generate new railway horizontal alignment.