Research papers · Posted 2024.04.20 · Publication / acceptance: author-reported
Augmenting Safety-Critical Driving Scenarios while Preserving Similarity to Expert Trajectories
Augmenting Safety-Critical Driving Scenarios while Preserving Similarity to Expert Trajectories
Overview
The method clusters expert trajectories to identify rare safety-critical groups, then geometrically combines similar examples. Safety-filtered augmentations are added to imitation-learning data and evaluated through closed-loop driving performance.
Publication and version information
Publication, acceptance or venue information is author-reported on arXiv. The exact source wording is preserved below.
Author-supplied source information
Author-supplied information on arXiv.
Accepted to 35th IEEE Intelligent Vehicles Symposium, 2024
Version: arXiv v1 — manuscript revision