Predicting pedestrian crossing behavior with attention-based encodings of vehicle and pedestrian states

Date

2025-8-12

Author

Kumral, Mehmet Ali

Metadata

Show full item record

This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License.

Item Usage Stats

2646
views

1573
downloads

Cite This

This thesis proposes a new method to predict the road-crossing intention of pedestrians with the aim of increasing the driving safety of autonomous vehicles. The developed method offers a architecture with the consideration of modularly integratable pedestrian action states, vehicular motion states, and pedestrian positioning to improve the prediction performance. Information hidden in the change of states is extracted by utilizing temporal self-attention and combined with the pedestrian positioning features to form a method that is both lightweight and fast. A comparison to existing methods shows notable improvements, especially in accuracy, recall, area under the receiver operating characteristic curve, and computational resource usage.

Subject Keywords

Autonomous vehicles, Pedestrian intention prediction, Machine learning, Attention networks, Feature fusion

URI

https://hdl.handle.net/11511/115591

Collections

Graduate School of Natural and Applied Sciences, Thesis