Loading...
Please wait, while we are loading the content...
Monitoring distraction through smartphone naturalistic driving experiment
| Content Provider | Semantic Scholar |
|---|---|
| Author | Yannis, George Tselentis, Dimitrios I. Vlahogianni, Eleni I. Argyropoulou, Anastasia |
| Copyright Year | 2017 |
| Abstract | for 6 International Naturalistic Driving Research Symposium, The Hague, Netherlands, 7-9 June 2017 Monitoring distraction through smartphone naturalistic driving experiment George Yannis, Dimitrios I. Tselentis,Eleni I. Vlahogianni, Anastasia Argyropoulou Department of Transportation Planning and Engineering, National Technical University of Athens, 5 Heroon Polytechniou str., GR-15773 Athens, *geyannis@central.ntua.gr Summary: Accurate monitoring of driving distraction has been proved to be a difficult task in the past. However, the rapid technological progress, especially in telematics and Big Data analytics, along with the increased penetration of information technologies to the drivers (e.g. smartphones), provide new potential for driving behaviour monitoring and analysis. Concerning driving distraction, one of the most influencing factors that has a significant effect on driver behaviour, is mobile phone usage. The objective of this paper is twofold, to analyse and assess the impact of mobile phone usage on driving characteristics such as the number of harsh events that occur while driving and to investigate the predictability of mobile phone usage. An innovative data collection scheme is implemented in this research by recording driving behaviour analytics in real time, using smartphone device sensors. Over a hundred drivers participated in the designed experiment during a 4-months timeframe. The number of harsh events occurred when driving was found to be influenced by several factors including mobile phone usage. Additionally, mobile phone usage is found to be predictable by a few factors such as driving duration and percentage of speed limit exceedance. The results of this research substantiate herein that distraction through smartphone has a serious impact on driving characteristics and subsequently on the relative crash risk. Further analysis of the data collected is implemented through statistical and econometric techniques and is leading to the quantification of the factors influenced by mobile usage that causes an alteration in driving risk. Background: Accurate monitoring of driving distraction has been proved to be a difficult task in the past. The rapid technological progress, especially in telematics, and Big Data analytics, along with the increase in the information technologies’ penetration and use by drivers (e.g. smartphones), provide new potential for driving behaviour monitoring and analysis. First results from related applications [1,2,3,4,5] have confirmed the efficiency and usefulness of such big data collection schemes. As for driver’s distraction, one of the most influencing factors that has a significant effect on driver behaviour is mobile phone usage [6]. Since mobile usage is a standard part of everyday driving process and is expected to increase over the years [7], its impact on driving behaviour in traffic and road safety is particularly important and should be further investigated. Literature so far has showed that when the driver is using the phone while driving his/her behaviour alters significantly. Therefore, mobile usage is banned in many countries [8] as the distraction caused is considered the main risk while driving [9,10,11]. Objective: The objective of this paper is twofold, to analyse and assess the impact of mobile phone usage and therefore driving distraction on driving characteristics as well as to examine the predictability of mobile phone usage. More specifically, by continuously collecting data from smartphone devices while driving, this study aims to examine the way that driving metrics recorded such as harsh events (braking, acceleration, cornering) are influenced by driving distraction in the form of mobile phone usage and therefore predicting the number of Monitoring distraction through smartphone naturalistic driving experiment 2 harsh events that take place. Moreover, exploiting the information collected regarding the number of harsh events that took place in each trip, a variable representing the number of harsh events per distance travelled is calculated to account for driver's performance. It should be highlighted that this study is investigating the macroscopic driving characteristics within a trip and as a result all indicators that were taken into consideration such as harsh events and mobile usage might have not been recorded simultaneously. Furthermore, the potential of predicting mobile phone usage while driving through recording of driving related metrics is also examined herein. Method: An innovative data collection scheme using a Smartphone Application that has been developed by OSeven was exploited for the purpose of this research. Driving behaviour analytics is recorded in real time, using smartphone device sensors. Over a hundred drivers participated in the designed experiment during a 4-months timeframe and a large database of several thousand trips is created. The solid integration platform for collecting, transferring raw data and recognizing the driving behaviour metrics via ML algorithms is also developed by OSeven. This ensured a smooth transition from the data collection to the data analysis procedure. All data received is evaluated and filtered when deemed to be necessary. The steps of the standard procedure developed that is followed every time a new trip is recorded by the App, are clearly shown in Figure 1. Figure 1 OSeven data handling chart Driving measures collected include indicatively distance travelled, speed, accelerations, braking, steering, cornering and smartphone usage (dialling, talking, texting etc.) in different driving environments (urban, rural, highway). During data processing, new variables were created in order to define the time of the day driving (daylight, morning rush, afternoon rush). A correlation matrix to check the correlation between variables used in the analysis and pivot charts of the data collected were created, and illustrated bellow. Monitoring distraction through smartphone naturalistic driving experiment 3 Figure 2Average Speeding per road type Figure 2 shows the average percentage of time driving over the speed limit per road type where it is evident that the percentage of exceedance of SL is lower for highways than urban and rural roads. Figure 3 shows the average mobile phone usage percentage (duration of mobile usage / driving time) per road type demonstrating again a lower percentage of mobile usage for highways. A linear regression model is developed to model the driving characteristics that influence the number of harsh events per distance travelled including mobile phone usage that is one of the components of driver distraction. The influence that each variable has on the number of harsh events occurring in each trip separately are quantified and consequently, the relative risk for each trip can be identified. A binary logistic model is also utilized to predict the situation of using or not the mobile phone while driving through the observation of different driving measures. Figure 3Mobile Usage per road type 0,00% 2,00% 4,00% 6,00% 8,00% 10,00% 12,00% 14,00% Pe rc en ta ge o f t im e dr iv in g ov er th e sp ee d lim it urban rural highway 0,00% 1,00% 2,00% 3,00% 4,00% 5,00% 6,00% 7,00% 8,00% M ob ile U sa ge Urban Rural Highway Monitoring distraction through smartphone naturalistic driving experiment 4 Results: Overall, a change in driving behaviour and more specifically in the number of harsh events occurred such as harsh brakings, accelerations and cornerings is proved to be predictable using as an indicator the mobile phone usage while driving. The linear regression model that originated from the above analysis, illustrates a significant dependence between driver's distraction (total harsh events / total distance) and percentage of mobile usage, the standard deviation of speed, the average exceedance of the speed limit as a percentage of the speed limit, the driving period during a day (morning, afternoon rush) and the total duration of the trip. As shown in Table 1, the most significant predictors among the others is found to be trip duration and average speed limit exceedance. Table 1Linear Regression model output for the estimation of Harsh Events Coefficients Model Unstandardized Coefficients Standardized Coefficients T Sig. |
| File Format | PDF HTM / HTML |
| Alternate Webpage(s) | https://www.nrso.ntua.gr/geyannis/wp-content/uploads/geyannis-pc260c.pdf |
| Alternate Webpage(s) | https://www.nrso.ntua.gr/geyannis/wp-content/uploads/geyannis-pc260.pdf |
| Language | English |
| Access Restriction | Open |
| Content Type | Text |
| Resource Type | Article |
