Defect-Repairable Latent Feature Extraction of Driving Behavior via a Deep Sparse Autoencoder.

Hailong Liu,Kazuhito Takenaka,Takashi Bando,Tadahiro Taniguchi

doi:10.3390/s18020608

Abstract

Data representing driving behavior, as measured by various sensors installed in a vehicle, are collected as multi-dimensional sensor time-series data. These data often include redundant information, e.g., both the speed of wheels and the engine speed represent the velocity of the vehicle. Redundant information can be expected to complicate the data analysis, e.g., more factors need to be analyzed; even varying the levels of redundancy can influence the results of the analysis. We assume that the measured multi-dimensional sensor time-series data of driving behavior are generated from low-dimensional data shared by the many types of one-dimensional data of which multi-dimensional time-series data are composed. Meanwhile, sensor time-series data may be defective because of sensor failure. Therefore, another important function is to reduce the negative effect of defective data when extracting low-dimensional time-series data. This study proposes a defect-repairable feature extraction method based on a deep sparse autoencoder (DSAE) to extract low-dimensional time-series data. In the experiments, we show that DSAE provides high-performance latent feature extraction for driving behavior, even for defective sensor time-series data. In addition, we show that the negative effect of defects on the driving behavior segmentation task could be reduced using the latent features extracted by DSAE.

Highlights

Driving behavior can be measured as multi-dimensional time-series data, i.e., the driving behavior data generated by a variety of sensors in a vehicle, e.g., the speed of the wheels, steering angle, and accelerator opening rate
In our view, reconstructing driving behavior data from the extracted time-series of latent features is important for evaluating the feature extraction model, especially for the deep sparse autoencoder (DSAE), which extracts the time-series of latent features by minimizing the reconstruction error with an encode-decode process
Note that the convergence speed is controlled by the learning rate when DSAE with a back propagation method (DSAE-back propagation (BP)) is used, which means that a comparison of the convergence speed between DSAE with a fixed-point iterative method (DSAE-FP) and DSAE-BP would have no significance

Summary

Introduction

Driving behavior can be measured as multi-dimensional time-series data, i.e., the driving behavior data generated by a variety of sensors in a vehicle, e.g., the speed of the wheels, steering angle, and accelerator opening rate. In most studies about driving behavior analysis, the feature vectors that are regarded as input data of the analysis method were selected manually from the measured sensor data [7,8,9] and designed manually [10]. Compared with manual selection and manual feature design, we can apply an automatic feature extraction method for obtaining low-dimensional time-series data that can represent multi-dimensional data from many sensors. We selected an unsupervised feature extraction method that can extract low-dimensional time-series of latent features from multi-dimensional sensor time-series data automatically

Objectives

Methods

Results

Discussion

Conclusion