With the growing popularity and application of knowledge-based artificial intelligence, the scale of knowledge graph data is dramatically increasing. As an essential type of query for RDF graphs, RegularPath Queries (RPQs) have attracted increasing research efforts, whichexplore RDF graphs in a navigational manner. Moreover, path indexeshave proven successful for semi-structured data management. However,few techniques can be used effectively in practice for processing RPQon large-scale knowledge graphs. In this paper, we propose a novel indexing solution named FPIRPQ (Frequent Path Index for Regular PathQueries) by leveraging Frequent Path Mining (FPM). Unlike the existing approaches to RPQs processing, FPIRPQ takes advantage of frequentpaths, which are statistically derived from the data to accelerate RPQs. Furthermore, since there is no explicit benchmark targeted for RPQsover RDF graph yet, we design a micro-benchmark including 12 basicqueries over synthetic and real-world datasets. The experimental resultsillustrate that FPIRPQ improves the query efficiency by up to orders ofmagnitude compared to the state-of-the-art RDF storage engine.