DEEL-AI/Hackathon_Team06_lvl3

This dataset was created using LeRobot.

Task description

The objective of the task is to grasp and place colored cubes in a 2x2 grid, using a SO-101 robot arm. Two cameras were used to visualize the scene.

The task to achieve is decomposed in three sub-tasks ordered by difficulty:

• level 1: pick a single black cube and place it in a cell of the 2x2 grid. The cell is indicated by a black cross and is always at the same position (top-left cell).
• level 2: same as level 1 but the target cell can be any cell in the 2x2 grid.
• level 3: pick two cubes (one black and one red) and place them on the cells marked with the corresponding colored crosses in the 2x2 grid.

Each level was the occasion to record a new dataset specific to the level to complete.

Dataset Description

This dataset contains a set of teleoperated demonstrations collected during a two-day hackathon using the LeRobot library and SO-101 robot arms in a leader–follower setup. Each episode shows the follower arm picking two colored cubes and placing them onto the matching colored crosses inside a 2×2 grid.

Two RGB cameras were used at +45° and -45° facing the follower robot arm.

The background was masked with cardboard and wood panes. Ambient lighting varied throughout the day.

Data Collection (level 3)

Data were already collected for levels 1 and 2, which involve picking up a single black cube. For level 3, we decided to record data using both cubes, always starting with the red cube. The reason is that the data for levels 1 and 2 only focus on picking up a single black cube. We wanted to encourage picking up the red cube first when both cubes are present. After that, picking up the black cube once the red cube is already placed closely resembles the behavior in levels 1 and 2. We expected that keeping this order would be simpler and require less data than starting with the black cube and then moving to the red one.

Episode Protocol

1- Start with the red cube: grasp and place it on the red cross

2- Go back to the starting position.

3- Repeat with the black cube.

4- Go back to the starting position.

Variability

Data were recorded with 4 different persons that manipulates the leader arm, each one with its own "style". Shades were also added randomly.