Ben Goodrich

Papers

Total Citations

H-Index

About

Ben Goodrich’s research lies at the intersection of robot learning, computer vision, and autonomous manipulation, with a focus on enabling robots to perceive and act in complex, real-world environments. His work tackles fundamental challenges in learning from demonstration (LfD) and depth estimation. In his 2018 paper “Imitation Learning from Visual Data with Multiple Intentions,” Goodrich advanced LfD by developing methods that allow robots to learn from high-dimensional visual inputs while disambiguating multiple task intentions—a critical step toward flexible, multi-skill robots. His 2020 paper “Depth by Poking: Learning to Estimate Depth from Self-Supervised Grasping” introduced a novel self-supervised approach that uses physical interaction (poking) to train neural networks for depth estimation, overcoming the limitations of traditional sensors on reflective or transparent surfaces. Though early in his career, with papers accumulating 6 and 5 citations respectively, Goodrich’s work is notable for its creative, embodied approach to perception—using the robot’s own actions to generate training data. His contributions are particularly relevant for robotic manipulation in unstructured settings, where robust depth sensing and versatile imitation learning are essential.