Built-In Tasks

SIMPLE ships 31 registered environments built from over 1,500 Objaverse objects, 75 GraspNet-1B objects and 50 HSSD indoor scenes. Tasks span rigid pick-and-place, non-prehensile interaction and articulated-object manipulation.

List them at runtime:

import gymnasium as gym
import simple.envs as _  # register all envs

[k for k in gym.registry if k.startswith("simple/")]

Naming

simple/<Robot><Skill><Variant?><Pipeline>-v0

The pipeline suffix says how demonstrations are produced — MP (automated CuRobo motion planning, 18 envs) or Teleop / Sonic (VR teleoperation, 13 envs) — and selects the data-generation and evaluation path.

Environment families

Entry point

Envs

Scope

SonicLocoManipEnv

13

Whole-body teleoperation via the external SONIC controller

LocoManipulationEnv

11

G1 whole-body and fixed-base G1 motion planning

TabletopGraspEnv

7

Fixed-base tabletop arms (Franka, Aloha, Vega)

Whole-body task names compose a locomotion primitive with a manipulation skill — XMove, Locomotion for the base; Pick, PickAndPlace, Bend, Handover for the body — for example simple/G1WholebodyLocomotionPickBetweenTablesTeleop-v0.