Search our Publications
Latest Publications
Christian Internó, Jumpei Yamaguchi, Markus Olhofer, David Klindt, Barbara Hammer , "Self-Supervised Learning Can Encode Physics, Finetuning Can Corrupt It", World Modeling Workshop (Mila - Quebec AI Institute, Montréal, Canada), 2026.
AbstractNeural simulators are increasingly used to model complex physical systems with high predictive accuracy. However, their internal "world models" are opaque, making it difficult to verify if they learn genuine physical principles or rely on brittle heuristics. In this work, we investigate whether mechanistic interpretability tools can reverse-engineer the physical laws from a model's internal representations. Our method involves attaching a linear ...
Download Bibtex file Per Mail Request
Thomas H Weisswange , "Modelling Intended Impact of Assistive Interactions", Transparent and Interpretable Robots Conference (TRAIL 2026), 2026.
AbstractExplainability is first and foremost grounded in social interaction. While it is important to research transparent algorithms, understand causal attributions and design expressive interfaces when creating explainable agents, the target will always be to understand how to achieve a certain effect on a human perceiver. The need for an explanation only arises when part of the human's world model is flawed. For deciding when, what and how to communi...
Download Bibtex file Per Mail Request
Radu Stoican, Christian Goerick, Angelo Cangelosi, Thomas H Weisswange , "Task-Specific Exploration in Meta-Reinforcement Learning via Task Reconstruction", Transactions on Machine Learning Research, 2026.
AbstractReinforcement learning trains policies specialized for a single task. Meta-reinforcement learning (meta-RL) improves upon this by leveraging prior experience to train policies for few-shot adaptation to new tasks. However, existing meta-RL approaches often struggle to explore and learn tasks effectively. We introduce a novel meta-RL algorithm for learning to learn task-specific, sample-efficient exploration policies. We achieve this through t...
Download Bibtex file
Daniel Tanneberg , "Embodied Intelligence: Artificial and Biological Neural Systems for Autonomous and Assistive Agents", ETH Zürich, 2026.
AbstractTalk at ETH Zürich, organized by our new collaboration partner Prof. Vörös, in combination with project Kickoff meeting. Topics will cover general HRI overview, published work on SMILE, previous personal work, and results from collaboration with CL from approved publication (pub-6649)....
Download Bibtex file Per Mail Request
Beate Stattkus-Fortange, Christiane Attig, Christiane Wiebel, Thomas Franke , "Understanding the charging behavior of electric vehicle drivers on long-distance trips – The roles of range regulation and human-automation cooperation", Transportation Research Part F: Psychology and Behaviour, vol. 120, pp. 103610, 2026.
AbstractAdvances in battery technology and charging infrastructure have improved long-distance electric vehicle (EV) travel. However, effective trip planning can be challenging. EV trip planners (EVTs) can support drivers in range regulation, yet their effectiveness and acceptance depend on how drivers experience the interaction. In this context, we introduce Driver Electric Vehicle Trip Planner Interaction Style (DEVTIS), which captures individual diffe...
Download Bibtex file
Pietro Fanti, Leon Williams, Ondrej Dvorak, Marcus Märtens, Tat-Jun Chin, Hongbo Ji, Bofei Chen, Dongyu Xie, Kaifan Qiao, Bohao Li, Nils Einecke, Subramanian Arumugam, Amulya Ratna Padhy, Sumeet Kumar Rath, Swati Sonal Mahapatra, Dario Izzo , "Event-based lunar optical flow egomotion estimation challenge: design and results of the ELOPE competition", npj Space Exploration, 2026.
AbstractEvent-based vision is a promising technology with incredible potential for future space exploration. The Event-based Lunar OPtical flow Egomotion estimation (ELOPE) Challenge aims at evaluating and comparing approaches for lunar landing egomotion estimation using data from a single event-based camera. This work is based on the ELOPE Dataset, which is the first publicly available event-based camera dataset for lunar landing. Over 44 teams particip...
Download Bibtex file Per Mail Request
Steffen Limmer, Christiane Attig, Tobias Rodemann , "A Hierarchical Controller as an Alternative Multi-Objective EV Charging Manager", 16th International Conference on Power, Energy, and Electrical Engineering (CPEEE), 2026.
AbstractSmart electric vehicle (EV) charging typically requires the consideration of a multitude of objectives like (electricity) costs, carbon emissions, user request satisfaction, grid stabilization and other factors. These objectives are typically combined as a weighted sum of all individual objectives. While this is conceptually easy, finding proper weights and confirming that any specific set of weights correctly implements the decision maker’s pref...
Download Bibtex file
Phillip Richter, Heiko Wersing, Anna-Lisa Vollmer , "A Taxonomy of Mental Model Mismatches for Human-Robot Interaction", Errors, Mistakes, and Failures in Humans and Robots (EMF) Workshop at HRI Conference, 2026.
AbstractMental model mismatches (MMM) in human-robot interaction occur when humans' internal representations of robots systematically diverge from robots' actual properties or behaviors, leading to communication breakdowns, task failures, and reduced trust. The field lacks shared terminology to identify specific types of mismatches. We present a theoretically grounded taxonomy of thirteen MMM types organized into three cognitive clusters. These are cogni...
Download Bibtex file Per Mail Request
Johannes Varga, Guenther Raidl, Tobias Rodemann , "Timing of Queries in Interactive Job Scheduling", The 24th Conference of the International Federation of Operational Research Societies, 2026.
AbstractTraditionally, it is assumed that for solving scheduling problems the information about the availability of users is readily available. However, this often implies a burdensome process of collecting the information upfront or to operate on incomplete information. In interactive scheduling, the scheduler actively involves users by asking them a small number of targeted questions to create an effective schedule. We present a heuristic inter...
Download Bibtex file Per Mail Request
Jörg Deigmöller, Nakul Agarwal, Stephan Hasler, Daniel Tanneberg, Chao Wang, Reza Ghoddoosian, Felix Ocker, Anna Belardinelli, Fan Zhang, Behzad Dariush, Michael Gienger , "MERGE: Guided Vision-Language Models for Multi-Actor Event Reasoning and Grounding in Human–Robot Interaction", arXiv, 2026.
AbstractWe introduce MERGE, a system for situational grounding of actors, objects, and events in dynamic hu- man–robot group interactions. Effective collaboration in such settings requires consistent situational awareness, built on persistent representations of people and objects and an episodic abstraction of events. MERGE achieves this by uniquely iden- tifying physical instances of actors (humans or robots) and objects and structuring them into...
Download Bibtex file
1 2 3 4 ... 183
Follow us
