The code is shared to illustrate the research presented in our proposed paper
Embodied Reasoning for Discovering Object Properties via Manipulation. The
code is experimental and not tested thoroughly.
This folder contains part of the data used for the experiments: a scene description and 50 questions with the resulting action sequences (questions).
The script task_execution.py loads for a question the scene and action sequence and controls the execution as well as the logging.
Python libabries used by the task_execution.py. action_classes.py holds classes
and utility functions to represent scenes, actions, and action sequences.
The code generation for the action sequences is done here as well.
Skills related to the gripper are defined in gripper_actions.py.
Skills related to navigating the robotic arm are defined in manipulator_actions.py.
This work was supported by the project Interactive Perception-Action-Learning for Modelling Objects (IPALM, https://sites.google.com/view/ipalm) (H2020 - FET - ERA-NET Cofund - CHIST-ERA III / Technology Agency of the Czech Republic, EPSILON, no. TH05020001) and by the European Regional Development Fund under project Robotics for Industry 4.0 (reg. no.
