Duncan McKay

Papers

Total Citations

H-Index

About

Duncan McKay is a roboticist whose work sits at the intersection of computer vision and manipulation, with a focus on enabling robots to grasp objects with semantic and directional precision. His key research areas include deep reinforcement learning, sim-to-real transfer, and semantic grasping. In his most notable work, "Toward Sim-to-Real Directional Semantic Grasping," McKay tackles the challenge of instructing a robot not only to pick up a specific object but to grasp it from a specific direction—a critical capability for tasks like tool use or assembly. He approaches this using a double deep Q-network (DDQN) that learns from downsampled RGB images captured by a wrist-mounted camera, training entirely in simulation before transferring the policy to a real robot. While his citation count is still growing, his contribution lies in framing grasping as a directional, semantic problem rather than a purely geometric one, paving the way for more intuitive human-robot interaction. McKay’s work represents an important step toward robots that can follow natural language commands like “hand me the screwdriver by the handle.”