Jul 2026
Time-to-Collision Based Dynamic Obstacle Avoidance Using Pretrained Vision Models for Robots in Unstructured Environments
This work presents a data-efficient, interpretable method for vision-based dynamic obstacle avoidance that operates entirely on real-world data, avoiding the sim-to-real transfer problem inherent in simulation-trained policies.
Erik Jagnandan, M. Haile, Gregory Barber et al.
· arXiv.org · 0 citations