Pedestrian path changing behaviour prediction models for urban mid blocks

Abstract

Pedestrian’s behaviour at a crosswalk is a critical influencing factor in determining pedestrian safety in urban areas. A facility will be ineffective when most pedestrians fail to use it effectively; moreover, using a location without a pedestrian crossing facility further increases the risk of involvement in accidents. Although several studies have been done to identify pedestrian behaviour at the crosswalk, only a few have been on path changing behaviour. The objective of this study is to analyse the pedestrian behaviour at midblock and to develop a model predicting the path changing behaviour. The pedestrian behaviour study was conducted at Kozhikode city and location with zebra marking and without zebra marking is considered simultaneously. The study shows that male pedestrians make more path change when compared to female pedestrians. About 88.6% pedestrians make path change at no zebra marking area, whereas only 60.5% in locations with zebra marking. This indicates that pedestrians tend to make a perpendicular or straight crossing when there is a facility like zebra marking. Crossing time is more for path changing pedestrian than the straight moving pedestrian and waiting time is lesser for pedestrian using zebra marking area. Path changing behaviour prediction is done by binary logit and machine learning techniques. Study shows that waiting time has negative influence on path changing behaviour and stages of crossing shows a positive influence on path changing behaviour. Increase in waiting time by 50% will result in 40% decrease of path changing behaviour.

Keywords

• Pedestrian Path change behaviour
• Urban road midblock
• Binary logit
• Neural network
• Naïve bayes
• Random forest