Examination of accident prevention measures due to incorrect brake and accelerator pedal pression (third report) (Analysis of driver behavior when the brake and accelerator pedals are pressed incorrectly)

Abstract

In the natural sciences, a common methodology is to observe phenomena, formulate a hypothesis, and verify it. However, human errors such as pedal mis-operation occur very infrequently, making it difficult to apply this methodology. Therefore, we established a method to artificially generate a pedal mis-operation situation on a driving simulator. Specifically, the method is remotely changing the pedal position so that the driver would not notice it. In this experiment, we acquired video and driving data. In this paper, we extracted driver features from the accumulated data and investigated them. The results showed that aging slows down the transition from accelerator pedal mis-operation to correct brake operation. In addition, it was found that elderly drivers tended to press the accelerator pedal again after mishandling the accelerator pedal. It is inferred that the higher incidence of mis-step accidents among younger drivers in the statistics is due to the higher frequency of pedal misapplication errors. On the other hand, it was found that the influence of driving frequency and gender was small. We confirmed that driver characteristics can be divided into "ability axis" and "personality and driving habit axis" by principal component analysis. Furthermore, decision tree analysis was used to analyze the driver features that allow the driver to operate the brakes correctly even if there is a accelerator pedal mis-operation.

Keywords

• accident analysis
• pedal errors
• driver behavior
• safety
• machine learning