Terry Mosier

Papers

Total Citations

H-Index

About

Terry Mosier is a robotics researcher specializing in vision-based pose estimation, with a focus on enabling robots to perceive their environment using minimal, cost-effective hardware. His work centers on the critical challenge of determining the spatial relationship between cameras, robots, and objects from single monocular RGB images—a problem fundamental to accessible robotic manipulation. Mosier’s major contributions include developing a deep learning approach for **camera-to-robot pose estimation** from a single image, where a neural network detects 2D projections of robot keypoints (like joints) trained entirely in simulation. He further advanced the field with an **indirect object-to-robot pose estimation** system, combining two neural networks to compute the relative pose between an object and a robot via a shared camera view. While his most-cited papers (with 5 and 3 citations respectively) are early-career works, they demonstrate a clear, methodical approach to solving practical perception problems. Mosier’s research is particularly notable for its emphasis on using only a single external monocular RGB camera, moving away from expensive multi-sensor setups and toward more deployable, real-world robotic systems.