Teaching Robots Interactively (TERI)

Project type

ERC Starting Grant; 2019-2023

Abstract

Programming and re-programming robots is extremely time-consuming and expensive, which presents a major bottleneck for new industrial, agricultural, care, and household robot applications. My goal is to realize a scientific breakthrough in enabling robots to learn how to perform manipulation tasks from few human demonstrations, based on novel interactive machine learning techniques.

Current robot learning approaches focus either on imitation learning (mimicking the teacher’s movement) or on reinforcement learning (self-improvement by trial and error). Learning even moderately complex tasks in this way still requires infeasibly many iterations or task-specific prior knowledge that needs to be programmed in the robot. To render robot learning fast, effective, and efficient, I propose to incorporate intermittent robot-teacher interaction, which so far has been largely ignored in robot learning although it is a prominent feature in human learning. This project will deliver a completely new and better approach: robot learning will no longer rely on initial demonstrations only, but it will effectively use additional user feedback to continuously optimize the task performance. It will enable the user to directly perceive and correct undesirable behavior and to quickly guide the robot toward the target behavior. In my previous research I have made ground-breaking contributions to the existing learning paradigms and I am therefore ideally prepared to tackle the three-fold challenge of this project: developing theoretically sound techniques which are at the same time intuitive for the user and efficient for real-world applications.

The novel framework will be validated with generic real-world robotic force-interaction tasks related to handling and (dis)assembly. The potential of the newly developed teaching framework will be demonstrated with challenging bi-manual tasks and a final study evaluating how well novice human operators can teach novel tasks to a robot.

Project members

dr. Zlatan Ajanović, dr. Carlos E. Celemin Paez, dr. Marta Ferraz, ing. Giovanni Franzese, Dr.-Ing. Jens Kober, dr. Ravi Prakash, dr. Leandro de Souza Rosa,

Publications


Anna Mészáros, Giovanni Franzese, and Jens Kober. Learning to Pick at Non-Zero-Velocity From Interactive Demonstrations. IEEE Robotics and Automation Letters, 7(3):6052-6059, 2022. [bibtex] [pdf] [url] [doi] [video]


Giovanni Franzese, Anna Mészáros, Luka Peternel, and Jens Kober. ILoSA: Interactive Learning of Stiffness and Attractors. In IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), pp. 7778–7785, 2021. [bibtex] [pdf] [url] [doi] [code] [video]


Snehal Jauhri, Carlos E. Celemin, and Jens Kober. Interactive Imitation Learning in State-Space. In 2020 Conference on Robot Learning (CoRL) (Jens Kober, Fabio Ramos, Claire Tomlin, eds.), PMLR, vol. 155 of Proceedings of Machine Learning Research, pp. 682–692, 2021. [bibtex] [pdf] [html] [code] [video]


Giovanni Franzese, Carlos E. Celemin, and Jens Kober. Learning Interactively to Resolve Ambiguity in Reference Frame Selection. In 2020 Conference on Robot Learning (CoRL) (Jens Kober, Fabio Ramos, Claire Tomlin, eds.), PMLR, vol. 155 of Proceedings of Machine Learning Research, pp. 1298–1311, 2021. [bibtex] [pdf] [html] [code] [video]


Rodrigo Pérez-Dattari, Carlos E. Celemin, Giovanni Franzese, Javier Ruiz-del-Solar, and Jens Kober. Interactive Learning of Temporal Features for Control: Shaping Policies and State Representations From Human Feedback. IEEE Robotics & Automation Magazine, 27(2):46–54, 2020. [bibtex] [pdf] [doi] [code] [video]

Linda van der Spaa, Giovanni Franzese, Jens Kober, and Michael Gienger. Disagreement-Aware Variable Impedance Control for Online Learning of Physical Human-Robot Cooperation Tasks. In ICRA 2022 full day workshop - Shared Autonomy in Physical Human-Robot Interaction: Adaptability and Trust, 2022. [bibtex] [pdf] [code] [video]

Carlos Celemin and Jens Kober. Uncertainties Based Queries for Interactive Policy Learning with Evaluations and Corrections. In Companion Publication of the 2021 International Conference on Multimodal Interaction, pp. 192–193, 2021. [bibtex] [pdf] [doi]

Bart Bootsma, Giovanni Franzese, and Jens Kober. Interactive Learning of Sensor Policy Fusion. In 30th IEEE International Conference on Robot & Human Interactive Communication (RO-MAN), pp. 665-670, 2021. [bibtex] [pdf] [doi]

Jan Scholten, Daan Wout, Carlos E. Celemin, and Jens Kober. Deep Reinforcement Learning with Feedback-based Exploration. In IEEE Conference on Decision and Control (CDC), pp. 803–808, 2019. [bibtex] [pdf] [doi] [code]

Daan Wout, Jan Scholten, Carlos E. Celemin, and Jens Kober. Learning Gaussian Policies from Corrective Human Feedback. arXiv:1903.05216 [cs.LG], 2019. [bibtex] [pdf] [url]