On April 20 at 5:00 PM, the 9th seminar of the CSIE Seminar Series was held, focusing on robust decision-making and control under uncertainty.

The seminar was delivered by CSIE research intern Amaras Nazarians, under the supervision of CSIE leading researcher Astghik Hakobyan.

The work addresses a key challenge in autonomous systems: how to keep robots, vehicles, or other intelligent systems safe when both the model of the system and the surrounding environment are uncertain.

The proposed framework combines:

◾Distributionally Robust Model Predictive Control — for planning safely under uncertain environments,

◾L1 Adaptive Control — for fast real-time correction when the system behaves differently than expected,

◾and mathematical safety guarantees based on uncertainty bounds.

In other words, the method helps a system plan ahead while also reacting quickly when reality does not match the model — making autonomous operation safer and more reliable in complex, changing environments.

Key results highlighted that the proposed DRP-L1AC approach ensures certifiable safety, improves robustness to distribution shifts, and provides real-time performance guarantees through Wasserstein-based uncertainty bounds.

Full paper: https://arxiv.org/abs/2603.28758