Core Concepts
Using machine learning to predict pedestrian crossing behavior at unsignalized crossings improves accuracy and reduces errors.
Abstract
The content discusses predicting pedestrian behavior when interacting with vehicles at unsignalized crossings using machine learning. It explores factors influencing interaction outcomes, such as time to arrival, waiting time, and personality traits. The study uses a distributed simulator dataset and develops models for predicting pedestrian crossing decisions, initiation time, and duration. Results show improved prediction accuracy and reduced errors compared to baseline models.
Structure:
I. Introduction
Importance of understanding pedestrian behavior in automated driving.
II. Methodology
Data collection from a distributed simulator study.
III. Results and Discussions
A. Crossing Decision Prediction:
Models' performance comparison for zebra and non-zebra crossings.
Impact of time to arrival on prediction accuracy.
Important features influencing crossing decisions.
B. Crossing Initiation Time Prediction:
Comparison of prediction errors among different models.
Box plots showing model predictions compared to ground truth.
C. Crossing Duration Prediction:
Evaluation of prediction errors for crossing duration.
Box plots illustrating model predictions for crossing duration.
IV. Conclusion
Summary of findings and contributions in improving predictive models for pedestrian-vehicle interactions.
Stats
Compared with the logistic regression baseline model, our proposed neural network model improves the prediction accuracy and F1 score by 4.46% and 3.23%, respectively.
Our model also reduces the root mean squared error (RMSE) for CIT and CD by 21.56% and 30.14% compared with the linear regression model.