Yusuf Umut Ciftci

Viterbo University, University of Southern California

Papers

Total Citations

H-Index

About

Yusuf Umut Ciftci is a rising researcher at the forefront of safe and robust imitation learning for robotics. His work directly tackles the critical challenge of covariate shift in behavior cloning (BC), where a robot's learned policy inevitably encounters states unseen during expert demonstrations. In his highly-cited paper "Stable-BC: Controlling Covariate Shift With Stable Behavior Cloning," Ciftci introduces a novel framework that ensures the robot's actions remain stable and effective even when drifting from the training distribution, a fundamental problem in deploying learned policies in the real world. Building on this, his work "SAFE-GIL: SAFEty Guided Imitation Learning for Robotic Systems" pioneers a safety-critical layer for BC, guiding the robot to avoid catastrophic errors in high-stakes environments. With his recent publications already garnering significant early attention, Ciftci is establishing himself as a key voice in making imitation learning not just effective, but also reliable and safe for real-world robotic applications.