Shahab Kaynama

University of California, Berkeley, Apple (United States)

Papers

Total Citations

311

H-Index

About

Shahab Kaynama is a researcher specializing in safe learning, reachability analysis, and safety-critical control systems, with a particular focus on their application to robotics and autonomous systems. His most influential contribution, "Reachability-based Safe Learning with Gaussian Processes" (2014), has garnered over 260 citations and represents a landmark effort in bridging reinforcement learning with formal safety guarantees — a challenge that had long impeded the deployment of learning algorithms in real-world, safety-critical environments. By leveraging reachability analysis to define provably safe regions of the state space, Kaynama's work opened new pathways for applying machine learning to systems where constraint violations carry serious consequences. His 2018 paper, "A General Safety Framework for Learning-Based Control in Uncertain Robotic Systems," further extended these ideas, offering a broader theoretical foundation for safe operation during the learning process itself. His practical contributions are equally notable, including a collision avoidance algorithm for sampled-data systems demonstrated on Pioneer ground robots. Collectively, Kaynama's research addresses one of the most pressing challenges in modern robotics: ensuring that intelligent, adaptive systems can learn and operate reliably without compromising safety.