为探究自主式交通系统（Autonomous Transportation System, ATS）的演化机理，基于本体的知识建模与基于产生式规则的知识推理分析自主式交通系统要素的相互作用及演化特征。首先，收集交通领域文本数据，经预处理后形成语料库。构建狄利克雷分布（Latent Dirichlet Allocation, LDA）模型和BERT-BiLSTM-CRF模型用于从语料库中抽取自主式交通的概念和实例。其次，为构建自主式交通的知识模型，基于本体理论，定义概念间的语义关系并填充实例。再次，依据知识模型构建了交叉口的自动驾驶场景要素网络，并基于产生式规则的知识推理方法，设立了3条规则进行推理。最后，计算了推理前后要素网络指标，包括关系数量、平均度、平均路径长度和图密度。结果表明，该知识推理方法挖掘了463条隐含关系，平均度由1.418增至2.347，平均路径长度由1.940减小为1.745，图密度由0.003上升为0.005。借助知识推理方法，总结出自主式交通系统的演化机理，即网络形态从稀疏发展为稠密、网络效率逐步提升的过程。

In order to explore the evolution mechanism of the Autonomous Transportation System (ATS), this paper analyzed the interactions and evolution features of ATS elements based on know-ledge modeling of ontology and knowledge reasoning of production rules. First, text data in the transportation domain was collected and pre-processed to form a corpus. Latent Dirichlet Allocation (LDA) model and BERT-BiLSTM-CRF model were built to extract concepts and instances of ATS from the corpus. Second, semantic relations among concepts were defined and instances were populated based on ontology theory to construct a knowledge model of ATS. Third, the element network for automatic driving at intersections was created according to the knowledge model, and three rules based on the knowledge reasoning of production rules were set up to reason the element network. Finally, the element network metrics before and after reasoning were calculated, including the number of relations, average degree, average path length, and graph density. The results show that the knowledge reasoning method mines 463 implicit relations, the average degree increases from 1.418 to 2.347, the average path length decreases from 1.940 to 1.745, and the graph density ascends from 0.003 to 0.005. By using knowledge reasoning, this paper summarized the evolution mechanism of ATS, i.e., the process of development for network morphology evolving from sparse to dense and the gradual increase in network efficiency.