Extracting Traffic Primitives Directly From Naturalistically Logged Data for Self-Driving Applications

Abstract

Developing an automated vehicle, that can handle complicated driving scenarios and appropriately interact with other road users, requires the ability to semantically learn and understand driving environment, oftentimes, based on analyzing massive amounts of naturalistic driving data. An important paradigm that allows automated vehicles to both learn from human drivers and gain insights is understanding the principal compositions of the entire traffic, termed as traffic primitives. However, the exploding data growth presents a great challenge in extracting primitives from highdimensional time-series traffic data with various types of road users engaged. Therefore, automatically extracting primitives is becoming one of the cost-efficient ways to help autonomous vehicles understand and predict the complex traffic scenarios. In addition, the extracted primitives from raw data should 1) be appropriate for automated driving applications and also 2) be easily used to generate new traffic scenarios. However, existing literature does not provide a method to automatically learn these primitives from large-scale traffic data. The contribution of this paper has two manifolds. The first one is that we proposed a new framework to generate new traffic scenarios from a handful of limited traffic data. The second one is that we introduce a nonparametric Bayesian learning method -a sticky hierarchical Dirichlet process hidden Markov model -to automatically extract primitives from multidimensional traffic data without prior knowledge of the primitive settings. The developed method is then validated using one day of naturalistic driving data. Experiment results show that the nonparametric Bayesian learning method is able to extract primitives from traffic scenarios where both the binary and continuous events coexist.

Hierarchical Reinforcement Learning Framework Towards Multi-Agent Navigation

“…Based on the characteristics of the continuity of agent motion, we use HMM to implement this module. [29] and [30] have also proved the competence of HMM to extract primitives from raw data. HMM is a time series model containing two types of Markov processes.…”

A Tempt to Unify Heterogeneous Driving Databases using Traffic Primitives

“…In order to provide well-labeled multi-dimensional traffic data, traffic primitive that represents principal compositions of driving scenarios should be recognized and extracted [15]. Powerful unsupervised learning-based technologies have been developed to achieve this.…”

“…Wang, et al [21] investigated three different nonparametric Bayesian approaches to analyze drivers' carfollowing styles. Apart from previous research focusing on driver's behavior, wang, et al [15] proposed a framework to extract primitives from multi-scale high-dimension raw traffic data and generate traffic scenarios. In order to automatically extract primitives with less subjective intervention and without prior knowledge then find and cluster the analogous, they introduced nonparametric Bayesian learning based on the sticky HDP-HMM method.…”

“…where F (·) is the emission function and θ pt is the emission parameter. Because the intrinsic dynamic of traffic is changing and open ended, the dimension of the parameter space regarding hidden states in the model is believed to be infinite [15], [22]. In order to represent such situation, a prior probability distribution on an infinite-dimensional space is introduced.…”

“…By using nonparametric Bayesian approaches in this paper, we can not only create a unify index for traffic data but also use this index as label. Since the nonparametric Bayesian method segments and classifies the data according to its inherent characteristic [15], the extracted primitives can represent meaningful real world traffic scenarios. After all, a well-designed Relational Database Management System (RDBMS) [16] is utilized to uniformly arrange and index big data according to defined driving scenarios.…”