Prediction for Obstacle Avoidance Route of Pedestrian using Stochastic Distributions

Abstract

This report proposes a modeling method for predicting a pedestrian walking route in stochastic manner. To represent pedestrian motion in naturally, 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. Using the loci obtained from the experiment, we first represent the walking speed and walking direction at local region in the form of histogram. Then we approximate the histogram by continuous distributions. As a result, we obtain the stochastic model representing the natural motion of a pedestrian. The effectiveness of the proposed method is examined by comparing the predicted locus with real walking one.