- Welcome to ARCLab, the astrodynamics, robotics, and controls lab in the AeroAstro department at MIT
Please upload datasets with the following convention
{robot}_{datatype}_{name}_dataset
where:
robot is the embodiment / simulation you are using (e.g. koch, xarm, rx1)
datatype is included if the dataset is augmented or modified in any way (e.g. a dataset with masking applied: masked)
name is the activity or data you are collecting (e.g. for pick and place data name: pickandplace)
- Always end with an appended _dataset
A full example for a dataset collected using the Koch arms, that applies binary masking, and performs pick and place tasks:
koch_masked_pickandplace_dataset
Likewise, for trained models follow a similar convention
{robot}_{modeltype}_{datatype}_{name}_model
where:
modeltype is the model used to train (e.g. action chunking with transformers: act)
name is the activity the model performs (e.g. for pick and place data name: pickandplace)
- Always end with an appended _model