Andrea Manzoni

HypEMBER achieves superior robustness in policy learning, outperforming traditional RL methods even in the face of significant measurement noise and parameter uncertainties.

SHRED-ROM can achieve real-time optimal control in high-dimensional systems with minimal computational overhead, revolutionizing how we approach adaptive control.

Agents that leverage localized sensory cues can outperform those relying on global flow information, revealing a surprising trade-off in navigation strategies.