This project studies how autonomous systems can interact reliably and adaptively with other agents — including humans — in dynamic and uncertain environments. Building on our work on permissive strategy templates, we develop methods that combine strong formal guarantees with the flexibility needed for real-world interaction. Our frameworks allow robots and autonomous agents to continuously adapt their behavior online, exploit opportunities for cooperation, and request assistance only when necessary to guarantee long-term objectives. This enables rich and often emergent cooperative behaviors while preserving safety, progress, and human autonomy. We demonstrate these ideas in autonomous driving, robotic navigation, collaborative manipulation with a Franka robotic arm, and human-AI interaction benchmarks such as Overcooked-AI. Short video overviews of the projects can be found here and here.
(Past) Group Members Involved:
- Kilian Schweppe (Intern)
- Oz Gitelson (Intern)
- Satya Prakash Nayak
- Ritam Raha
- Anne-Kathrin Schmuck
Publications:
- * O. Gitelson, S.P. Nayak, R. Raha, A.-K. Schmuck. Maximal Adaptation, Minimal Guidance: Permissive Reactive Robot Task Planning with Humans in the Loop. (under review, preprint, video)
- K. Schweppe and A.-K. Schmuck. Context-Triggered Contingency Games for Strategic Multi-Agent Interaction. 2026. ICRA’26. (preprint, video)