Modeling the Effects of Advanced Driver Assistance Systems on Driver Behavior
Analytics
35 views ◎12 downloads ⇓
Abstract
About 38,000 fatalities are reported every year in the United States and traffic crashes (referred to as crashes) are the leading cause of deaths among people up to 54 years. Additionally, economic loss due to crashes is estimated equal to $380 million, annually, in direct medical bills. Further, new vehicles are added to the roads every year increasing the traffic exposure and vehicle miles traveled. Driver errors are the leading cause of crashes and contribute to about 94% of crashes. Automobile manufacturers are striving to enhance the vehicles to eliminate driver errors, which can help avoid the major chunk of crashes. These enhancements include development of various types of advanced driver assistance systems (ADAS) that are designed to assist or in some cases also take over certain driving maneuvers. They include lane departure warning (LDW), blind spot warning (BSW), over speeding warning (OSW), lane keep assist (LKA), front collision warning (FCW), adaptive cruise control (ACC), and automatic emergency braking (AEB). Each of these features are focused at addressing a particular task of driving, thereby, reducing the driving load on the driver and also enhancing safety.The ADAS are expected to reduce crashes and yet a 14% increase in crashes was observed from the year 2014 to the year 2016. On the other hand, the acceptance levels of ADAS among drivers are questionable. Many surveys determined that the drivers are unaware of the applications and limitations of ADAS. To catalyze the issue, drivers admitted to blindly trusting such features which makes the problem critical. Hence, there is a need to understand how the driver behavior is influenced when driving a vehicle with ADAS compared to when driving a vehicle without ADAS. The National Advanced Driver Simulator (NADS) miniSimTM driver simulator was used to capture driver behavior in this research. Three different driving scenarios namely; urban, rural and freeway scenarios were developed to test on the drivers (participants) with varying weather and lighting conditions. Other variables like the demographic characteristics of the participants were also considered for analyzing and modeling the data. This enabled an extensive analysis of the effects of ADAS on driver behavior while also magnifying the applicability of this research. The research can be categorized into four vital stages. The first stage is to develop appropriate driving scenarios to test the effects of ADAS in all driving conditions experienced in the real-world. The second stage involves the careful selection of participants such that the sample population is an accurate representation of the general population. The third stage involves the data processing and analysis of data to derive meaningful results. The fourth stage involves the identification of changes in driver behavior and applying them to propose any possible changes that could further enhance ADAS. LDW was observed to reduce lane departure events in all the three scenarios (rural, urban, and freeway). OSW reduced the average and maximum speeds making driving less aggressive in rural and urban scenarios only, indicating they were not as effective in the freeway scenario. Similarly, BSW was also observed to affect the brake pedal force and influence aggressive driving. Providing two advanced features at a time also affected brake pedal force indicating they were effective in influencing aggressive driving. Further, none of the warning features were observed to influence the participant following behavior as the average headway difference between with and without ADAS was not found to be statistically significant. Driving behavior improved further when vehicles with automated features like ACC and LKA were provided individually or in combination to the participants. Automated features improved braking, vehicle handling, and lane-following behaviors in all the three driving scenarios. However, more aggressive car-following behavior was observed with the automated features. The variation in driving behavior among participants when provided with automated features reduced drastically. The effects of automated features were influenced by the type of driving scenario. The intervention of ADAS with driving tasks led to safer driving conditions. The driving safety improved with the level of assistance provided to the drivers. While the ADAS is effective in meeting their intended objectives, they seem to inadvertently affect other driving behaviors. The type of driving scenario (rural, urban, or freeway) also influenced the way an advanced feature affects the driver behavior. Braking behavior is predominantly affected by the presence of an advanced feature in most cases, which also influenced vehicle handling events like lane-following, turning, and car-following in some cases. Lighting and weather conditions had similar effects on driver behavior when not provided with any advanced features, when provided with warning features, and when provided with advanced features as well. Longer headways were observed in nighttime conditions and rainy conditions. However, less aggressive lane-following, braking, and vehicle handling behavior was observed. Also, more speeding was observed on freeways in clear weather. Male drivers displayed aggressive driving maneuvers when provided with both warning and automated features. On the other hand, female drivers maintained smaller headways in urban scenario and longer headways in rural and freeway scenario. Similarly, drivers aged under 25 years maintained smaller headways in urban scenario but maintained longer headways in rural and freeway scenarios. Further, drivers aged above 25 years showed more aggressive braking and speeding behavior with both warning and automated features in urban scenario. The type of ADAS provided, the type of driving scenario, the lighting and weather conditions, as well as the age and gender of the participants affected the driver’s behavior. The nature of the effects of ADAS, however, varied by the type of driving scenario. Further, the effects of all these factors varied when segregated by the type of ADAS (warning or automated feature) provided compared to when not provided with any advanced features. The effects of both warning and automated features varied when provided individually and in combination. However, warning features had limited behavioral changes when provided in combination, but automated features displayed evidently different driving behavioral changes in combination and individually. Based on the observations made from this research, it is suggested to accommodate both operational and safety standpoints while developing ADAS. Further, developing adaptive ADAS, formulating educational policies, and developing methods to collect naturalistic driving data are also emphasized. The findings can be used to define vehicle parameters within microscopic simulation software and mimic the effect of vehicles with and without advanced features on transportation system performance. Additional samples can be collected and other advanced features may also be tested and compared using the driver simulator.