Nihal Gunukula

Conference paper describing PathCluster, an approach that generates pedestrian groups to improve efficiency and safety of robot navigation in dense crowds (IEEE IROS publication).

Replication and ablation study improving and analyzing MEDIRL for modeling pedestrian behavior and socially-aware navigation in HRI contexts; includes reproducibility efforts and ablation experiments.

Proposes an inverse reinforcement and curriculum learning model (MIRACLE) that uses gamified data collection to train deep inverse reinforcement learning for human-like robot navigation, with application to emergency scenarios.

Technical/engineering blog post hosted on Phyvant discussing limitations of a common AI search technique.