Michael Koval

Papers

Total Citations

H-Index

About

Michael Koval is a roboticist whose research focuses on enabling robots to perform robust, dexterous manipulation in the presence of uncertainty. His core contributions lie at the intersection of contact-rich manipulation, decision-making under uncertainty, and hybrid control. Koval’s most cited work, "Pre- and Post-Contact Policy Decomposition for Planar Contact Manipulation Under Uncertainty" (23 citations), addresses the challenge of using real-time contact sensor feedback to generate closed-loop pushing actions. By formulating this as a partially observable Markov decision process (POMDP) with a physics-based transition model, he provided a principled framework for robots to reason about uncertain contact interactions. He further advanced the field with his work on "Hybrid Control Trajectory Optimization Under Uncertainty," which tackles the computationally demanding problem of optimizing sequences of both discrete and continuous control actions—a critical capability for tasks like grasping and assembly. Through these contributions, Koval has helped bridge the gap between theoretical planning under uncertainty and practical robot manipulation, demonstrating how to systematically handle the noise and partial observability inherent in real-world physical interaction.