We introduce Habitat 2.0 (H2.0), a simulation platform for training virtual robots in interactive 3D environments and complex physics-enabled scenarios. We make comprehensive contributions to all levels of the embodied AI stack - data, simulation, and benchmark tasks.

We introduce Habitat 2.0 (H2.0), a simulation platform for training virtual robots in interactive 3D environments and complex physics-enabled scenarios. We make comprehensive contributions to all levels of the embodied AI stack – data, simulation, and benchmark tasks. These large-scale engineering contributions allow us to systematically compare deep reinforcement learning (RL) at scale and classical sense-plan-act (SPA) pipelines in long-horizon structured tasks, with an emphasis on generalization to new objects, receptacles, and layouts. We find that (1) flat RL policies struggle on HAB compared to hierarchical ones; (2) a hierarchy with independent skills suffers from'hand-off problems', and (3) SPA pipelines are more brittle than RL policies.

The Habitat 2.0 simulation platform is for training virtual robots in interactive, physics-enabled 3D environments. We present the new ReplicaCAD 3D data set of apartments; the Habitat 2.0 simulator, capable of simulating 850x faster than real time; and the Home Assistant Benchmark, a suite of common tasks for assistive robots.