Modeling the Effects of Intelligent Driver Assistance on Driver Behaviors

Abstract

Many studies have concluded that activities like answering a cell phone call and entering an address while driving could potentially distract the driver. This thesis presents the design and experimental evaluation by simulation of four Intelligent Driver Assistance (IDA) modes to improve driver safety and efficiency. In order to verify the effectiveness of these modes, we conducted four experiments on a driver simulator called the Cognitive-Computational Driving Model (CCDM). The four IDA modes are: 1) Limited Cognitive Resources: None of the driver’s workload channels (such as vision, speech, cognition etc.) are allowed to exceed the baseline values needed for normal driving. 2) Safe Mode driving: Distracting events such as cell phone calls are terminated, if a safety-critical event like a vehicle cutting in front occurs. 3) Dynamic delay: Distracting events (such as entering GPS destination by voice) are delayed or disabled temporarily, if a safety-critical event (such as an obstacle in lane ahead) occurs. 4) Static delay: Distracting events (e.g., cell phone rings) are delayed by a certain time interval if the driver's workload is high. We measured various aspects of the driver such as cognitive load, time to complete safety-critical goals, ratio of interface related actions to driving related actions, reaction times, visual perception delays, etc. with and without intelligent driver assistance. After analyzing the data, we found that in the simulated scenarios, with intelligent driver assistance the driver could concentrate more on safety-critical tasks and finished them earlier compared to when intelligent driver assistance was turned off. Cognitive overload was reduced in all modes except the dynamic delay mode. Visual overload was less with limited cognitive resources and static delay modes. Also, we found that among the four modes of IDA, the static delay mode was better than the rest of the modes in terms of safety, efficiency, and visual and cognitive resource usage of the driver. We integrated the kernel of simulation engine with Microsoft Excel 2007. This allowed us to pipe the simulation output data at various intervals to the Excel application. With the office-automation feature of CLI/C++, we have been able to automatically generate driver’s workload charts/graphs from each simulation run.