Pedestrian crossing decision prediction based on behavioral feature using deep learning

Sidharta, Hanugra Aulia; Yuniamo, Eko Mulyanto; Kindhi, Berlian Al; Pumomo, Mauridhi Hery

doi:10.1109/isitia52817.2021.9502243

Cited by 1 publication

(4 citation statements)

References 12 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…In this study, a technique is used that sets the learning rate as a function of the epoch. As training goes further, optimization step size becomes smaller, as shown in (15), considering both convergence rate and training accuracy. The initial rate r 0 can be set to 0.1.…”

Section: Pedestrian Feature Estimationmentioning

confidence: 99%

“…To quantify pedestrians' features is a complicated problem since the behaviour logic in a pedestrian-vehicle interaction scenario is intrinsic and not intuitive [14]. Some studies treat pedestrian feature as a vector consisting of several physical quantities including velocities and behaviour labels [15,16]. Another method is to use one unified quantity to represent the overall feature.…”

Section: Introductionmentioning

confidence: 99%

“…Vermuyten et al [26] listed some evaluation methods based on deep learning. For example, Sidharta et al [15] introduced a multi-layer perception (MLP) framework for assessing pedestrians' features. Yang et al [27] used attention-based neural networks to predict pedestrian's intentions.…”

Section: Introductionmentioning

confidence: 99%

“…For example, Sidharta et al. [15] introduced a multi‐layer perception (MLP) framework for assessing pedestrians’ features. Yang et al.…”

Section: Introductionmentioning

confidence: 99%

See 3 more Smart Citations

A differentiated decision‐making algorithm for automated vehicles based on pedestrian feature estimation

Wang

Huang

Zhang³

et al. 2023

IET Intelligent Trans Sys

View full text Add to dashboard Cite

One critical difficulty to high‐level automated driving is the decision‐making process of automated vehicles in complicated traffic environments, especially in situations mixed of pedestrians and vehicles. This paper proposes a differentiated decision‐making algorithm to promote passing capability and efficiency in mixed traffic conditions. First, the behavioural characteristic of pedestrians, denoted as the pedestrian feature index, is estimated by a multi‐layer perception module input with quantitative analysis of pedestrian action. Based on estimation results, the decision algorithm merges pedestrian feature index into intelligent driver model and adjusts corresponding parameters, which used to be unchangeable so that the ego‐vehicle can make differential decisions according to various pedestrian features. Validation on the PIE dataset shows that the accuracy of pedestrian feature estimation is ensured. A simulation scenario is established utilizing cellular automata, and the results indicate that the proposed decision‐making algorithm can greatly improve passing efficiency under safety and manoeuvrability prerequisite.

show abstract

Section: Pedestrian Feature Estimationmentioning

confidence: 99%