RoboTwin 2.0 — RLinf Support Status

Contributed by RLinf team

RoboTwin 2.0 now includes initial RLinf integration, enabling reinforcement learning training across a wide range of robot manipulation tasks.

At this stage, RLinf-style reward functions have been implemented for a subset of RoboTwin tasks.
Support for the remaining tasks is actively under development.

✅ Currently Supported Tasks (with RLinf Rewards)

The following tasks are fully supported with RLinf-compatible reward functions and can be directly used for reinforcement learning training:

Manipulation & Placement

• adjust_bottle
• place_a2b_left
• place_a2b_right
• place_bread_basket
• place_bread_skillet
• place_burger_fries
• place_can_basket
• place_cans_plasticbox
• place_container_plate
• place_dual_shoes
• place_empty_cup
• place_fan
• place_mouse_pad
• place_object_basket
• place_object_stand
• place_phone_stand
• place_shoe

Pick, Move & Transport

• pick_diverse_bottles
• pick_dual_bottles
• move_can_pot
• move_pillbottle_pad
• move_playingcard_away
• move_stapler_pad
• grab_roller
• lift_pot
• put_bottles_dustbin

Stacking & Ranking

• stack_blocks_two
• stack_blocks_three
• stack_bowls_two
• stack_bowls_three
• blocks_ranking_rgb
• blocks_ranking_size

Tool Use & Interaction

• beat_block_hammer
• press_stapler
• stamp_seal
• turn_switch
• click_bell
• click_alarmclock
• open_microwave

Handover & Human-Object Interaction

• handover_block
• handover_mic

Pouring, Dumping & Shaking

• shake_bottle
• shake_bottle_horizontally
• dump_bin_bigbin

Hanging & Special Tasks

• hanging_mug
• scan_object
• rotate_qrcode

❌ Currently Unsupported Tasks (RLinf Rewards Not Yet Available)

• open_laptop
• place_object_scale
• put_object_cabinet

🚧 Work in Progress

We are continuously extending RLinf reward support to all RoboTwin 2.0 tasks, including more complex long-horizon manipulation and multi-object interaction scenarios.

New task support and reward refinements will be released incrementally.

📢 Contributions & Feedback

Community feedback and contributions are highly welcome.
If you encounter issues, missing reward definitions, or would like to contribute improvements, please feel free to submit:

• GitHub issues
• Pull requests
• Feature requests

📅 Latest Update

RoboTwin 2.0 RLinf compatibility confirmed —
All tasks listed above are currently supported with RLinf-style rewards.

Thank you for using RoboTwin 2.0 with RLinf!