Early Prediction of Driver's Action Using Deep Neural Networks

Early Prediction of Driver's Action Using Deep Neural Networks

Shilpa Gite (Symbiosis Institute of Technology, Pune, India) and Himanshu Agrawal (Symbiosis Institute of Technology, Pune, India)
Copyright: © 2019 |Pages: 17
DOI: 10.4018/IJIRR.2019040102

Abstract

Intelligent transportation systems (ITSs) are one of the most widely-discussed and researched topic across the world. The researchers have focused on the early prediction of a driver's movements before drivers actually perform actions, which might suggest a driver to take a corrective action while driving and thus, avoid the risk of an accident. This article presents an improved deep-learning technique to predict a driver's action before he performs that action, a few seconds in advance. This is considering both the inside context (of the driver) and the outside context (of the road), and fuses them together to anticipate the actions. To predict the driver's action accurately, the proposed work is inspired by recent developments in recurrent neural networks (RNN) with long short term memory (LSTM) algorithms. The performance merit of the proposed algorithm is compared with four other algorithms and the results suggest that the proposed algorithm outperforms the other algorithms using a range of performance metrics.
Article Preview
Top

1. Introduction

With the initial acceptance of various autonomous vehicle projects, there is an ongoing research focus on improving the existing prototypes on autonomous vehicles to be used on wide scale in near future. The continuous work on autonomous vehicles is driving research community to develop better prediction algorithms to predicting the driving behaviour. Road accidents are mainly attributed due to range of reasons including distraction of driver in terms of eating, talking while driving, fatigue, drunk driving, poor vision of road surface, driving after medication, over speeding and sudden lane change behaviour (Shia et al., 2014). Drowsiness or fatigue breaks the driver concentration while driving which results into loss of decision making functionality for controlling the car. From basic research & polls presented by (Vasudevan et. al 2012), it is suggested that for continuous driving case, driver experiences fatigue after every 2-3 hours and hence the control over the steering is affected. It is further reported that, the driver drowsiness is more likely to occur during midnight hours followed by after lunch afternoon hours as compared to other time slots in a day. These studies have further concluded that the alcohol and drugs also lead to the loss of driver concentration. Across the world, many countries presented their own statistics on accidents that happened because of driver’s distraction or fatigues. As reported by (Vasudevan et. al 2012), driver drowsiness and distraction is the leading cause (20-30%) of all accidents. Furthermore, (Rezaei & Klette, 2014) emphasized on fatigue and had classified driver fatigues types such as muscle fatigue, cognitive fatigue, and sensory fatigue. The fatigues like sensory and muscular are easily addressable, but there is no technique to address the cognitive fatigue. Therefore, anticipation of driver’s action could help in minimizing road accidents.

A significant research work has been conducted to predict and analyse driver’s behaviour or action related information. (Dong W. et al., 2016) discussed different driving styles using deep learning approach including Convolutional Neural Network (CNN) and Recurrent Neural Network (RNN). They worked on driving behaviours but other driving contexts like street level, street shape and climate has not been considered that additionally impact driving conduct. (Yan C. et.al. 2016) have described driving posture in regard to driver hand position, with high-level state features extracted progressively using the convolution layers and the max-pooling layers. They have not taken night time conditions. (Saito Y. et al., 2016) discussed identification of driver drowsiness and accordingly prevent lane departure accidents (Gindele T. et al., 2015) proposed learning driver behaviour models from traffic observations for decision making and efficient planning. They estimated and anticipated activity circumstances, building predictable probabilistic models of driver’s communications with the environment by the use of hierarchical dynamic Bayesian model, and partially observable Markov decision process. The framework utilizes many concrete solutions for Probability and decision making and they all are tightly coupled, so this yield additional time complexity. Some of these previous methods to assist drivers are based on blind spot detection and lanes keeping that are being used to alert drivers if they perform dangerous action. But with such methods, sometimes it becomes too late or you may not provide sufficient time to alerting driver to prevent or control the dangerous actions (Jain et al., 2016). Recent approaches are using deep learning-based framework to anticipate driver’s action few seconds in advance, thereby avoiding the possibility of an accident. In one of the recent research works, (Jain et al., 2016) proposed a deep learning framework by redesigning Recurrent Neural Networks (RNN) and Long Short Term Memory (LSTM), the proposed work is inspired by this research work. The contribution of the paper is summarised as below:

Complete Article List

Search this Journal:
Reset
Open Access Articles: Forthcoming
Volume 10: 4 Issues (2020): 2 Released, 2 Forthcoming
Volume 9: 4 Issues (2019)
Volume 8: 4 Issues (2018)
Volume 7: 4 Issues (2017)
Volume 6: 4 Issues (2016)
Volume 5: 4 Issues (2015)
Volume 4: 4 Issues (2014)
Volume 3: 4 Issues (2013)
Volume 2: 4 Issues (2012)
Volume 1: 4 Issues (2011)
View Complete Journal Contents Listing