I am a Research Scientist at Epic Games, Inc. where I am working on the next generation of photorealistic capture systems. In October of 2024, I graduated with a PhD in Robotics (Computer Science) from The Robotics Institute at Carnegie Mellon University. I was advised by Prof. Chris Atkeson. For my PhD, I worked on systems and algorithms for better perception of small scenes.
I am very interested in 3D capture systems and curious about new ideas and hardware that capture interesting aspects of the 3D world around us. If this is interesting to you as well, I'd love to hear from you.
Thesis title: Moving Lights and Cameras for Better 3D Perception of Indoor Scenes
Executive Summary: We investigate the effects of moving lights and cameras for accurately measuring small 3D scenes. We first look at the effects of moving cameras through a robot mouunted ensembeld of cameras and camera-based tactile sensors for visual servoing and contact Localization. Following this, we look at the effects of moving lights through a robot workspace scaled photometric stereo system and finally, we investigate the effect of jointly moving lights and cameras through a robot mounted multi-flash stereo camera system for ohotorealistic capture of small assets.
We demonstrate a robot mounted multi-flash stereo camera for generating photorealistic, portable representations of small scenes. We also investigate the incorporation of dense metric depth from stereo into 3D reconstruction of small scenes.
In this work we demonstrate a robot workspace scaled controlled illumination setup that generates high quality information for table top scale objects for robotic manipulation. With our low angle of incidence directional illumination setup we can precisely capture surface normals and depth discontinuities of Lambertian objects. We demonstrate three use cases of our setup for robotic manipulation. We show that 1) by using the captured information we can perform general purpose grasping with a single point vacuum gripper, 2) we can visually measure the deformation of known objects and, 3) we can estimate pose of known objects and track unknown objects in the robot's workspace.
In this work we make the point that coordinating tactile sensing with vision by a co-located camera instead of only consideringthe tactile sensor inputs alone can a) provide useful informationin advance of contact and b) simplify the contact point andpose estimation problem.
Previous Research
In the past I have worked on medical robotics, design of mechanical components and parallel robots. I have also done some research on modelling the human three fingered grasp and on optimal motion planning inside constrained spaces.
In this short paper, we describe a new method to generate stiffness maps for soft tissues by fusing a stand alone optical tracking system with an off-the-shelf force sensor.
In this paper, we describe an optimization based approach for motion planning of hyper-redundant robots, which results in a natural motion of the links through ducts and onfined spaces. We demonstrate 3 scenarios: (video 1) snake-like robots inspecting an industrial pipeline, (video 2) motion of an endoscopic robot through gastro-intestinal (GI) tract and (video 3) motion of snake-like robots in search and rescue operations.
In this work, we use the Monte Carlo method in conjunction with gradient based optimization algorithms to optimally design multi-degree-of-freedom parallel manipulators and closed-loop mechanisms.
Inspired by a multi-jointed human finger and the hand, we propose a six-degree-of-freedom model of a three-fingered robotic hand as a parallel manipulator. We simulate the contact between the fingertips and objects and model the workspace of the human hand. We compare the model with real motion-capture data and show that our method provides a conservative upper bound on the actual workspace.
This paper presents a methodology for designing prismatic springs of non-circular coil shape and non-prismatic springs of circular coil shape using analytical and numerical methods.