Pick, place & grasping
3Grasping, pick-and-place, and object transfer with robot arms or hands.
Research topic · Robot manipulation
Every record here is a published, source-linked manipulation corpus. Compare what actually decides transfer: episodes, embodiment, camera views, action signals, and usage terms.
A manipulation dataset is only comparable on the fields the source actually reports. These are the axes this topic keeps visible on every record; anything the source does not report stays unreported instead of being estimated.
01
The record's own headline stat (episodes, trajectories, hours). Scale claims stay source-linked; missing scale stays missing.
02
Which arm, hand, or humanoid the trajectories were recorded on — the field a cross-embodiment transfer depends on most.
03
Wrist, external, and depth views reported by the source. View placement drives policy observability.
04
Joint states, end-effector poses, proprioception, or teleoperation traces — what a policy can actually learn to output.
05
Open access is not a usage license. Each record links the official source for current terms.
Grouped by task
Records enter this topic through the published Catalog's manipulation classification; every card links the official source for access and terms. Groups are navigation aids derived from each record's own task labels and description — the record remains authoritative, and unmatched records stay under “General manipulation”.
Records can also appear in the VLA training data topic, which groups the same corpora by training stage instead of task.
Grasping, pick-and-place, and object transfer with robot arms or hands.
Dexterous hands, in-hand reorientation, and fine fingertip control.
Two-arm coordination and bimanual platforms such as ALOHA-class rigs.
Multi-robot corpora used to pretrain or transfer manipulation policies.
Failures, interventions, corrections, and recovery during manipulation.
Manipulation records whose own labels do not match a task group yet.
Task groups are derived from each record's own labels and description; they do not replace the record. Use All tasks to see every record.
8 shown · 24 total
AgiBot (Zhiyuan Robotics) · 2026
AgiBot World 2026 is the latest release of the AgiBot World embodied-AI dataset, succeeding the earlier Alpha/Beta releases. It is collected in real-world commercial, home, and general-purpose environments using the AGIBOT G2 wheeled humanoid with dual 7-DOF force-controlled arms, capturing synchronized RGB(D), LiDAR point clouds, tactile, IMU, and…
2,976.4 hours · humanoid_other
tonyzhaozh · 2023
Learning Fine-Grained Bimanual Manipulation with Low-Cost Hardware
300 trajectories Source-reported scale
Berkeley · 2023
A large dataset of WidowX manipulation in diverse kitchen and tabletop scenes.
60.1K episodes · widowx
University of Freiburg · 2022
CALVIN (Composing Actions from Language and Vision) is an open-source PyBullet-simulated benchmark for learning long-horizon, language-conditioned continuous-control manipulation policies with a Franka Panda arm. It spans four tabletop environments (A, B, C, D), 34 manipulation tasks, and ~1000 crowd-sourced language annotations, with evaluation rollouts…
24 hours · franka_panda
NVIDIA Research (with UT Austin and UC San Diego) · 2024
DexMimicGen is an automated data-generation system that synthesizes large-scale trajectory datasets for bimanual dexterous manipulation from a handful of human teleoperation demonstrations. It generated over 21,000 demonstrations from 60 source human demos across 9 simulation tasks in robosuite/MuJoCo, spanning three embodiments (bimanual Panda arms with…
21K episodes · franka_panda · humanoid_other
DROID Collaboration (Stanford, UC Berkeley, et al. — 13 institutions) · 2024
DROID (Distributed Robot Interaction Dataset) is a large-scale in-the-wild robot manipulation dataset of 76,000 teleoperated demonstration trajectories (~350 hours), collected across 564 scenes and 84 tasks by 50 collectors over 12 months in North America, Asia, and Europe (13 institutions). Every episode uses a standardized Franka Panda 7-DoF arm with two…
76K episodes · franka_panda
UC Berkeley (RAIL / Robotic AI & Learning Lab) · 2024
FMB is a real-world robotic manipulation benchmark from UC Berkeley's RAIL lab for studying functional manipulation, where a Franka Panda arm composes skills (grasping, repositioning, assembly/insertion) using 3D-printed objects on assembly boards. It comprises 22,550 expert demonstration trajectories across single-object and multi-object multi-stage tasks,…
22.6K episodes · franka_panda
CLVR Lab, KAIST · 2023
FurnitureBench is a reproducible real-world benchmark from CLVR Lab (KAIST) for long-horizon, complex robotic manipulation framed as furniture assembly with a Franka Panda arm. It releases 5100 successful teleoperation demonstrations totaling 219.6 hours across 9 furniture models and three task-initialization randomness levels (low/medium/high). The…
5,100 episodes · franka_panda
Suggest a dataset, correct a grouping, request data, or discuss sourcing, custom collection, and synthetic data with a human.
Submit a dataset
Maintainers: add a corpus or update access, license, and scale facts at the official source.
contact@openbot.aiRequest specific data
Tell us the task, embodiment, and signals you need; we point you at published records or gaps.
contact@openbot.aiDiscuss a data partnership
For data sourcing, custom collection, and synthetic data inquiries from teams and enterprises.
contact@openbot.ai