Understanding V2V Driving Scenarios Through Traffic Primitives

Understanding driver interaction behavioral semantics has potential benefits to autonomous car’s decision-making design. This article presents a framework of analyzing various encountering behaviors through decomposing driving encounter sequential data into small building blocks, called traffic primitives, using Bayesian nonparametric learning (BNPL) approach. offers flexible way gain semantic insights complex encounters without any prerequisite knowledge behavior categories. Its effectiveness is then validated 976 naturalistic from which more than 4000 primitives were learned with the BNPL After that, dynamic time warping method integrated $k$ -means clustering developed cluster all these extracted groups. Experimental results identify 20 kinds capable representing essential components in our database. Based on results, we conclude that proposed primitive-based analysis could prove useful for vehicle applications.