What's the cartesian_position in DROID dataset?

[darthandvader]

FeaturesDict({
'episode_metadata': FeaturesDict({
'file_path': string,
'recording_folderpath': string,
}),
'steps': Dataset({
'action': Tensor(shape=(7,), dtype=float64),
'action_dict': FeaturesDict({
'cartesian_position': Tensor(shape=(6,), dtype=float64),
'cartesian_velocity': Tensor(shape=(6,), dtype=float64),
'gripper_position': Tensor(shape=(1,), dtype=float64),
'gripper_velocity': Tensor(shape=(1,), dtype=float64),
'joint_position': Tensor(shape=(7,), dtype=float64),
'joint_velocity': Tensor(shape=(7,), dtype=float64),
}),
'discount': Scalar(shape=(), dtype=float32),
'is_first': bool,
'is_last': bool,
'is_terminal': bool,
'language_instruction': string,
'language_instruction_2': string,
'language_instruction_3': string,
'observation': FeaturesDict({
'cartesian_position': Tensor(shape=(6,), dtype=float64),
'exterior_image_1_left': Image(shape=(180, 320, 3), dtype=uint8),
'exterior_image_2_left': Image(shape=(180, 320, 3), dtype=uint8),
'gripper_position': Tensor(shape=(1,), dtype=float64),
'joint_position': Tensor(shape=(7,), dtype=float64),
'wrist_image_left': Image(shape=(180, 320, 3), dtype=uint8),
}),
'reward': Scalar(shape=(), dtype=float32),
}),
})

Does the cartesian_position mean the gripper pose? In which coordinate frame?
Also, what does the gripper_position represent?