The general goal of the COCOBOTS project ANR-21-FAI2-0005 was to contribute to the development of models for situation conversation that would allow a human to effectively collaborate through conversation with a conversational cobot to perform a task. An important aspect of this objective is to be able to use situated dialogue to teach the robot new concepts or actions when necessary without requiring recourse to a roboticist, programmer or any specialized training methods. The method targeted is one in which a robot is taught simple programs through conversation and then learns to combine these programs to build more and more complex programs.
To this end, we designed an assembly task matching a use case provided by the German company Synergeticon, a partner of Airbus. In this use case, robots arms are used to fill peg boards with pre-assembled hardware parts (specific combinations of, say, metal plates, nuts and bolts) so that they are easily exploited by workers building more complex parts down the line. The goal of the COCOBOTS assembly task is thus for a builder to assemble specific combinations of hardware pieces on a grid board, following instructions given by another player. Only the player giving the instructions has visual access to the target board; the builder and instructor must collaborate through conversation until the builder understands how to assemble the board correctly.
The data in this repo result from small-scale experiments conducted via the data collection platform Prolific. They are suitable as the base of a small test or development corpus for other researchers interested in studying task-based conversations between a human instructor and cobot builder.
Below are the instructions shown to participants in the COCOBOTS assembly task as can be seen on the Slurk portal developed by the University of Potsdam.
Please read the following information carefully before accepting the HIT. Clicking the button on the landing page will take you to a waiting room. Once another player arrives, you will both be moved to the task room where you can begin the task.
- One player teaches another to build a construction on a Target Board by providing instructions in English in a chat window.
- The second player follows the instructions by picking objects from the Source Board and placing them on the Working Board.
- Each session includes 3 boards (~20 minutes).
- Participants who complete all three boards successfully will receive a monitary bonus.
- Participants who remain idle or disconnect for more than 5 minutes, causing the session to end, and participants who fail to understand the task after practice will not be allowed to participate in further sessions.
- A session will be shut down after the maximum allotted time, 5 minutes of idleness from one player, or 5 minutes in the waiting room (in the event that no partner is found).
By entering into the chatroom you confirm your permission for your data to be logged and used for research purposes. The data that is logged is stored on a server at the University of Potsdam. Any experiment-relevant information, such as what you type in the chat is logged with timestamp information. No information that can identify you personally is logged. Please refrain from entering personal details, as well as experiment-irrelevant information in the chat and respect other users. Please be aware that you will be paired with another anonymous participant -- as we are looking for dialogues-- and we have no control over what they might do. Please behave courteously, and demand this from your partner as well. Responsable party: Department of Linguistics University of Potsdam
If you use or refer to the data in this repository, please cite the following paper:
Paetzel-Prüsmann, M., Hunter, J., Chalamalasetti, K., Thompson, K., Nicolaou, A., Güngör, O., Shlangen, D. & Asher, N. (2022, May). Conversational Programming for Collaborative Robots. In ICRA Workshop on Collaborative Robots and Work of the Future (ICRA 2022 CoR-WotF) (pp. 1-5).
@inproceedings{paetzel2022conversational,
title={Conversational Programming for Collaborative Robots},
author={Paetzel-Pr{\"u}smann, Maike and Hunter, Julie and Chalamalasetti, Kranti and Thompson, Kate and Nicolaou, Alexandros and G{\"u}ng{\"o}r, Ozan and Schlangen, David and Asher, Nicholas},
booktitle={ICRA Workshop on Collaborative Robots and Work of the Future (ICRA 2022 CoR-WotF)},
pages={1--5},
year={2022}
}