iGibson

Stanford's interactive indoor simulation platform, used for navigation and mobile-manipulation research in household settings.

iGibson was developed by Stanford's Vision and Learning Lab (Fei-Fei Li, Silvio Savarese, and colleagues), as an interactive version of the earlier Gibson environment. Version 1.0 (IROS 2021) provides 15 fully interactive scenes reconstructed from real homes, totaling 108 rooms, where furniture, cabinet doors, drawers, and other rigid and articulated objects can all be pushed or opened by the robot; it can output RGB, depth, segmentation, and lidar sensor signals, with physics based on the Bullet engine, and it comes with its own motion planner. Version 2.0 (CoRL 2021) added object states such as temperature, wetness, cleanliness, and being sliced, and supports collecting human demonstrations with VR, used for the BEHAVIOR household-task benchmark. Stanford's BEHAVIOR-1K later switched to OmniGibson, built on NVIDIA Isaac Sim, and most new projects have since moved to that instead.

ExampleIn an iGibson household scene, a mobile robot is made to navigate to the kitchen and open a cabinet door, using the built-in motion planner to generate a collision-free arm trajectory.

Also called

Interactive Gibson, iGibson 1.0, iGibson 2.0

Related

OmniGibson · BEHAVIOR-1K (BEHAVIOR Challenge) · Habitat · AI2-THOR · Mobile Manipulation · PyBullet

Sources

iGibson project page (Stanford SVL)
iGibson 1.0 (arXiv 2012.02924)
BEHAVIOR (Stanford) — OmniGibson on Isaac Sim

As of

2026-09