
IKIDA is a BMBF funded interdisciplinary research group researching in interactive AI algorithms and cognitive models for human-AI interaction.
IKIDA is an interdisciplinary junior research group and consists out of 6 young scientists from the research fields of machine learning, robotics and cognitive science.
The overall technical and scientific goal of IKIDA is the development of interactive, probabilistic AI algorithms, which benefit from direct interaction with users. Enabling such direct interaction between AI algorithms and users is also intended to lower application hurdles of AI solutions and can potentially increase their sustainability and acceptance. For the development of the AI algorithms we particularly plan to incorporate and develop new cognitive models for human-AI and human-robot interaction.
The associated industry partners Energy Robotics (Darmstadt), Franka Emika (Munich) and Porsche Motorsport (Weissach) will help to find and evaluate practical relevant use cases for the developed interactive AI methods, such as automated data classification and robot skill learning.
IKIDA is also supported by the AI initiative and the AI:DA . Centre for Cognitive Science
A short demonstration on Interactive Reinforcement Learning With Bayesian Fusion of Multimodal Advice (see: Trick, S., Herbert, F., Rothkopf, C., & Koert, D. (2022). Interactive Reinforcement Learning With Bayesian Fusion of Multimodal Advice. IEEE Robotics and Automation Letters, 7(3), 7558-7565. DOI (opens in new tab)) PDF
A short demonstration on Adapting Object-Centric Probabilistic Movement Primitives with Residual Reinforcement Learning (see Carvalho, J., Koert, D., Daniv, M., Peters, J. (2022). Adapting Object-Centric Probabilistic Movement Primitives with Residual Reinforcement Learning. Proceedings of the International Conference on Humanoid Robots (HUMANOIDS) 2022 (accepted) (opens in new tab)) PDF
A short demonstration on Learning from Unreliable Human Action Advice in Interactive Reinforcement Learning (see Scherf, L.; Turan, C., Koert, D. (2022). Learning from Unreliable Human Action Advice in Interactive Reinforcement Learning. 2022 IEEE-RAS 21st International Conference on Humanoid Robots (Humanoids). (opens in new tab) ) PDF
Project Details
Project: | IKIDA – Interaktive KI für Domänenexperten und Alltagsnutzer |
Project partners: | Technical University of Darmstadt (TU Darmstadt) |
Project duration: | October 2020 – September 2024 |
Project funding: | 1.95 Mio EUR |
Funded by: | German Federal Ministry of Education and Research (BMBF) |
Grant no.: | 01IS20045 |