Dynamically Adapting Driver Assistance Systems Using Driver Monitoring Technology
Advanced driver assistance systems (ADAS), are rapidly becoming standard across manufacturer lineups. These technologies (SAE Level 1; SAE, 2014) provide either steering or acceleration assistance to the driver, such as lane departure warnings (LDW) that alert the driver that the vehicle is approaching a marked lane boundary without signaling. These technologies have been referred to as “Guardian Angel” systems in that they can compensate for driver errors, which account for over 90% of fatal vehicle crashes (NHTSA, 2017). Research suggests that many ADAS technologies, such as forward collision warnings (FCW), could significantly reduce the number of crashes (Jermakian, 2011). Unlike FCW, however, evidence is at best mixed on the potential crash reduction associated with lane departure warnings. Data from the Highway Loss Data Institute failed to show a reduction in crash claims for LDW-equipped vehicles. One contributing factor may be driver acceptance of LDW. Research by the Insurance Institute for Highway Safety showed that, of a sample of 1,000 ADAS-equipped vehicles, only 45% had the systems activated. These deactivation rates are linked to driver acceptance issues and lack of trust in the systems. Most LDW are triggered before a vehicle crosses a lane boundary, increasing the potential that these alerts could be seen as nuisance alarms. Nuisance alarms have a direct negative impact on user trust and subsequent willingness to use technology (Bliss & Acton, 2003).
One potential solution is adaptive automation, which dynamically adjusts the level of automation based on the state of the operator (Scerbo, 2008). The goal of adaptive automation is to tailor the amount of technological support to the dynamic needs of the operator. For example, a LDW might fire if a driver is distracted or otherwise impaired but not when he is attentive. Extensive research shows a benefit of adaptive automation over static automation across tasks such as air traffic control and display monitoring (Parasuraman et al., 1999). Recent technological innovations make it possible to monitor the driver’s state via in-cabin camera, which could be used as the input into an adaptive system. A recent SaferSim automated vehicle study used a production driver monitoring system to provide feedback to distracted drivers in highly-automated vehicles (Gaspar et al., 2018).
The goal of this proposal is examine whether driver state information can be used to create adaptive lane keeping systems and examine the impact on driver performance, acceptance, and trust. This research builds on previous Safer-Sim research to address the focus area of Automated Vehicle Technology.