M. Ornik

Seemingly innocuous prompts can covertly hijack robotic actions, steering them toward adversarial outcomes while maintaining the facade of intended commands.

Real-world cyber-physical system resilience demands a holistic approach integrating hardware, software, human factors, proactive measures, and adaptive recovery, moving beyond traditional fault models.

Overcome the curse of dimensionality in motion planning for unknown dynamics by adaptively refining state-space partitions only where needed, enabling efficient online learning and control.