Work Highlights

• Focuses on predicting both risky and safe lane changes using a novel approach combining Knowledge Graphs and Bayesian inference.
• Utilizes CARLA simulator for collecting the CRASH (CARLA Risky-lane-change Anticipation in Simulated Highways) dataset.
• Achieves 91.5% f1-score for risky lane changes with anticipation time that extends to 4 seconds.
• Attains 90.0% f1-score for safe lane changes with anticipation time that extends to 4 seconds.
• Successfully validated the prediction of sudden lane changes in CARLA simulation environments.
• Implements Retrieval Augmented Generation (RAG) for explainable predictions.

Designed Knowledge Graph

LC_KG_reduced (1).mp4

Interactive Playground for Exploring Our Knowledge Graph

https://embed.kumu.io/2d1ea6b4d77ed852e8ae62458362abaa

Results

Click the button below to view additional supplementary video demonstrations.

IDM WITHOUT the integration of the model predictions

https://www.youtube.com/embed/te4tm0hyDxQ?si=LxbvOl57EfCT_azz

IDM WITH the integration of the model predictions

[http://<iframe width="560" height="315" src="https://www.youtube.com/embed/UPmzsjOJY2k?si=A0rccI_7xjlo9ZK-" title="YouTube video player" frameborder="0" allow="accelerometer; autoplay; clipboard-write; encrypted-media; gyroscope; picture-in-picture; web-share" referrerpolicy="strict-origin-when-cross-origin" allowfullscreen></iframe>](http://<iframe width="560" height="315" src="https://www.youtube.com/embed/UPmzsjOJY2k?si=A0rccI_7xjlo9ZK-" title="YouTube video player" frameborder="0" allow="accelerometer; autoplay; clipboard-write; encrypted-media; gyroscope; picture-in-picture; web-share" referrerpolicy="strict-origin-when-cross-origin" allowfullscreen></iframe>)