metadata
license: mit
task_categories:
- robotics
language:
- en
tags:
- robotics
- simulation
- pick-and-place
- niryo-ned
pretty_name: Niryo Ned Dataset – Blue Cube Pick & Place
Niryo Ned Dataset – Blue Cube Pick & Place
Description:
This dataset contains 4000 recorded trajectories(episodes) of a Niryo Ned/Ned2 robotic arm performing a pick-and-place task in simulation.
The robot picks up a blue cube and places it to the right of itself.
Both the robot’s initial joint configuration and the cube’s initial position are randomized per episode.
Task Instruction
Each episode corresponds to the following instruction:
"Place the blue cube to the right of the robot."
Dataset Structure
Each episode contains:
- Observation
observation.state
: current joint positions + gripper state (0 = closed, 1 = open)observation.images.env1
: 256×256 RGB image from the first camera viewobservation.images.env2
: 256×256 RGB image from the second camera view
- Action
action
: target joint positions + gripper state
- Recording Rate
- One step recorded every 0.333 seconds (≈ 3 steps per second).
- Each step contains joint state, next action, and images.
Highlights
- Trajectories: 4000 simulation episodes
- Task: Pick & Place of a blue cube (grasp → place to the right)
- Instruction: Uniform language instruction for all episodes
- Robot system: Niryo Ned/Ned2 (6-DOF arm)
- Data source: Recorded in simulation
- Interactive Exploration: Episodes can be explored in detail with the
LeRobot Dataset Visualizer