Abstract
This paper describes a modeling method for predicting a walking route of a pedestrian in a stochastic manner. We consider one of the most typical situations where a pedestrian walks along to a sidewalk, and then some obstacle exists in front of the pedestrian. To represent the walking route of the pedestrian during the avoidance action, a stochastic model is suitable than deterministic one. The stochastic model is derived from the walking experiment where a pedestrian avoids some obstacle in natural walking. Based on the loci obtained from the experiment, the pedestrians walking speed and walking direction at any local area was approximated by Gaussian and Beta distribution function, respectively. As a result the walking route of a pedestrian can be represented in a stochastic manner. The estimated output of the model is examined by comparing with two real walking loci that were obtained from near-miss incident database. One examination scene is avoidance of a parked vehicle, and the other is of parked bicycle on the roadside. By the numerical simulation, we obtained the results that the both real walking routes are included within the 3-sigma ranges of the estimated output of 500 trials.
