Herbie Wright

I am a CIS/Robotics PhD student at the University of Pennsylvania (GRASP). I am a member of the DAIR lab under Dr. Michael Posa. I am lucky to be funded by the NSF-GRFP fellowship. My undergrad was spent at the University of Utah, where I got my B.S. in Computer Science and participated in research under Dr. Tucker Hermans (also advised by Dr. Weiming Zhi). My formidable h-index is currently 1.

My research interests cluster around robotic manipulation. I have previously done research in perception for robotics in tabletop scenarios, but I would like to branch out to more control/planning, uncertainty-aware methods, and theoretical work in addition to perception. I am a fan of probabilistic approaches as well as quirky math. Occasionally, I write my partially-informed opinions on robotics in my blog →

Email: herbiew [at] seas [dot] upenn [dot] edu

Research Highlight - V-PRISM

Github → | Arxiv → | Website →

"V-PRISM: Probabilistic Mapping of Unknown Tabletop Scenes" By H. Wright, W. Zhi, M. Johnson-Roberson, T. Hermans.

Volumetric, Probabilistic, and Robust Instance Segmentation Maps (V-PRISM) is a method that probabilistically maps a tabletop scene of multiple objects. It takes in a segmented point cloud and builds a distribution of possible instance segmentation maps. This means that it can capture principled uncertainty about occluded areas of the scene.

I presented this work in 2024 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS) in Abu Dhabi. The paper is also up on Arxiv here.

Papers

  1. H. Wright, W. Zhi, M. Johnson-Roberson, T. Hermans. "Robust Bayesian Scene Reconstruction by Leveraging Retrieval-Augmented Priors". Arxiv, 2024. Learn More →
  2. H. Wright, W. Zhi, M. Johnson-Roberson, T. Hermans. "V-PRISM: Probabilistic Mapping of Unknown Tabletop Scenes". In 2024 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), 2024. Learn More →

You can also view my work from Google Scholar

You can see the source code for this site here