Investigating pedal errors and multi-modal effects: Driving testbed development and experimental analysis

These days, Intelligent Driver Assistance Systems (IDAS) for driving safety enhancement have an increasingly important role. Although there have been a large amount of related research studies in this area, developing an efficient assistance system, which can perceive and assist the driver in a non-intrusive and naturalistic manner, is still an open question. Due to the diverse characteristics and safety issue of driving, one initial and major difficulty in studying and developing such systems is to have adequate infrastructure testbeds with multi-modal sensing and displays. In this paper, we introduce two multi-modal driving testbeds (including both a real-world vehicle and a driving simulator) that we have been developing for years in our laboratory. Based on these testbeds, we describe two novel joint audio visual driving experiments and databases that we have built for investigating driver pedal misapplication error phenomenon in traffic accident prevention. We will then discuss our analysis towards understanding some factors influencing driver pedal errors including driver workload, sequential effect, and cue modality (i.e. audio visual stimuli) as well as a possibility to predict and mitigate pedal errors to some extent using our approach for modeling and prediction of driver foot behavior.