Identifying and Modelling Pedestrian Encounters that Drivers Think Warrant Alerts English Free

  • Category Technical paper
  • Edition SIA
  • Date 10/07/2012
  • Author Jan-Erik Källhammer - Autoliv
  • Language English
  • Type PDF file (175.14 Ko)
    (Downloadable immediately on receipt of online payment)
  • Number of pages 5
  • Code R-2012-02-34
  • Fee Free

In-vehicle active safety systems that alert the driver to potential hazards or incidents have the potential to save lives and reduce the likelihood of crashes by making the driver aware of vulnerable road users and of traffic situations that may develop into a fatal accident. This paper discusses a program of research aimed at uncovering drivers’ responses to night-time traffic situations involving pedestrians and to system-generated alerts to those situations. The system couples a far infra-red (FIR) sensor, pedestrian recognition algorithms, alerting logic, and a console display. At issue is the design of the alerting logic. Our aim is to tune the system to issue alerts when and only when drivers believe the situation warrants one. By coupling field operational tests (FOT) and laboratory observations, we have identified a set of contextual variables that appear to influence drivers’ assessment of the acceptability of system alerts to pedestrian encounters. A nominal characterization of pedestrian location and two quantitative measures of pedestrian motion explain a large share of the variance in driver ratings.