Bimanual Fold Glasses

Task key: bimanual_fold_glasses.

bimanual_fold_glasses is a two-arm rigid-object task. A pair of glasses rests on a tray and the agent must fold the hinges.

Goal

• Fold the glasses hinges into the folded range.

Quickstart

import dexsuite as ds

env = ds.make(
    "bimanual_fold_glasses",
    manipulator=("franka", "franka"),
    gripper="robotiq",
    arm_control="osc_pose",
    gripper_control="joint_position",
    render_mode="human",
)

obs, info = env.reset()
obs, reward, terminated, truncated, info = env.step(env.action_space.sample())

env.close()

Task Entities

• tray: fixed tray mesh.

• glasses: glasses URDF with hinge joints.

Extra Observations

Bimanual fold-glasses adds (under obs["state"]["other"]):

• tray_top_z (shape (1,) in the current implementation)

• hinge_angles (shape (2,) in the current implementation)

• glasses_pos (shape (n_envs, 3))

Termination

This task is still being finalized. In the current code, reward and success/failure checks are not enabled, so episodes typically end by horizon truncation.

Simulation Settings

• SIM_DT = 0.01

• SUBSTEPS = 2

• HORIZON = 40000 control steps

Source

• Dexsuite/dexsuite/environments/bimanual/rigid/bimanual_fold_glasses.py