v0.2.0

Added

• Support new observation modes: rgbd_robot_seg and pointcloud_robot_seg
• Support enable_gt_seg option for environments.
• Add two new rigid-body environments: AssemblingKits-v0 and PandaAvoidObstacles-v0

Breaking Changes

• TurnFaucet-v0: Add target_link_pos to observations
• PickSingleEGAD-v0: Reduce the density of EGAD objects and update EGAD object information
• Remove tcp_goal_pos in PickCube, LiftCube, PickSingle
• Update TurnFaucet assets. Need to re-download assets
• Change segmentation images from 2-dim to 3-dim
• Replace xyz with xyzw in obs["pointcloud"]. We use the homogeneous representation to handle infinite points (beyond the far of camera).

Fixed

• TurnFaucet-v0: Cache the initial joint positions so that they will not be affected by previous episodes
• Pour-v0: Fix agent initialization typo
• Excavate-v0: Fix hand camera position and max number of particles

What's Changed

• Update README.md by @StoneT2000 in https://github.com/haosulab/ManiSkill2/pull/1
• Update Dockerfile by @xuanlinli17 in https://github.com/haosulab/ManiSkill2/pull/2
• Improved functionality of a few utility scripts by @xiqiangliu in https://github.com/haosulab/ManiSkill2/pull/4
• Update base_env.py by @xuanlinli17 in https://github.com/haosulab/ManiSkill2/pull/5
• Update README.md by @xuanlinli17 in https://github.com/haosulab/ManiSkill2/pull/6
• Update README.md by @xuanlinli17 in https://github.com/haosulab/ManiSkill2/pull/9
• Soft body patch by @fbxiang in https://github.com/haosulab/ManiSkill2/pull/12

New Contributors

• @StoneT2000 made their first contribution in https://github.com/haosulab/ManiSkill2/pull/1
• @xuanlinli17 made their first contribution in https://github.com/haosulab/ManiSkill2/pull/2
• @xiqiangliu made their first contribution in https://github.com/haosulab/ManiSkill2/pull/4
• @fbxiang made their first contribution in https://github.com/haosulab/ManiSkill2/pull/12

Full Changelog: https://github.com/haosulab/ManiSkill2/commits/v0.2.0