Safe Guaranteed Exploration for Non-linear Systems
Event details
| Date | 04.12.2026 |
| Hour | 14:00 › 15:00 |
| Speaker | Dr Manish Prajapat Ph.D. reinforcement learning and control from ETH Zurich Winner of the IEEE CSS Young Author Best Journal Paper Award 2026 |
| Location | Online |
| Category | Conferences - Seminars |
| Event Language | English |
Abstract:
Safely exploring environments with a-priori unknown constraints is a fundamental challenge that restricts the autonomy of robots. While safety is paramount, guarantees on sufficient exploration are also crucial for ensuring autonomous task completion. To address these challenges, we propose a novel safe guaranteed exploration framework using optimal control, which achieves first-of-its-kind results: guaranteed exploration for non-linear systems with finite time sample complexity bounds, while being provably safe with arbitrarily high probability. The framework is general and applicable to many real-world scenarios with complex non-linear dynamics and unknown domains. For efficient implementation, we exploit goal-directed exploration, and receding-horizon replanning while preserving the framework’s guarantees, and demonstrate safe, efficient exploration in challenging unknown environments using a car model.
Biography:
Manish Prajapat earned his Ph.D. in reinforcement learning and control from ETH Zurich in May 2026. He was a Doctoral Fellow at the ETH AI Center working with Prof. Melanie Zeilinger and Prof. Andreas Krause. Previously, he earned his master’s degree in Robotics, Systems, and Control from ETH Zurich and was a visiting scholar at Caltech. He received his bachelor’s degree from the Indian Institute of Technology (IIT) Madras in 2017. At IIT Madras, he was honored as the Best Graduating Student (Co-curricular) in 2017 and received the Sivasailam Merit Prize for the best thesis. His research interests are sequential decision-making under complex scenarios, e.g., non-Markovian objectives, unknown constraints or unknown dynamics of non-linear systems.
Safely exploring environments with a-priori unknown constraints is a fundamental challenge that restricts the autonomy of robots. While safety is paramount, guarantees on sufficient exploration are also crucial for ensuring autonomous task completion. To address these challenges, we propose a novel safe guaranteed exploration framework using optimal control, which achieves first-of-its-kind results: guaranteed exploration for non-linear systems with finite time sample complexity bounds, while being provably safe with arbitrarily high probability. The framework is general and applicable to many real-world scenarios with complex non-linear dynamics and unknown domains. For efficient implementation, we exploit goal-directed exploration, and receding-horizon replanning while preserving the framework’s guarantees, and demonstrate safe, efficient exploration in challenging unknown environments using a car model.
Biography:
Manish Prajapat earned his Ph.D. in reinforcement learning and control from ETH Zurich in May 2026. He was a Doctoral Fellow at the ETH AI Center working with Prof. Melanie Zeilinger and Prof. Andreas Krause. Previously, he earned his master’s degree in Robotics, Systems, and Control from ETH Zurich and was a visiting scholar at Caltech. He received his bachelor’s degree from the Indian Institute of Technology (IIT) Madras in 2017. At IIT Madras, he was honored as the Best Graduating Student (Co-curricular) in 2017 and received the Sivasailam Merit Prize for the best thesis. His research interests are sequential decision-making under complex scenarios, e.g., non-Markovian objectives, unknown constraints or unknown dynamics of non-linear systems.
Practical information
- General public
- Free
Organizer
- Prof Giancarlo Ferrari Trecate The seminar is sponsored by the Swiss chapter of the IEEE-CSS