Tony Tao | 陶龙

I am a second year M.S. Robotics Student at the Robotics Institute at Carnegie Mellon University, advised by Deepak Pathak. I completed my Bachelor's degree at Carnegie Mellon in 2024.

Previously, I've had the pleasure of being advised by Guanya Shi and have spent time working with Zeynep Temel and Oliver Kroemer

Email: tonytao [at] cmu.edu

Email  |  Google Scholar  |  LinkedIn  |  X (Twitter) 

Research

I am interested in building robots that can learn through interaction to complete useful tasks reliably and generally in the real world.

You can also find my papers on my Google Scholar

DexWild: Dexterous Human Interactions for In-the-Wild Robot Policies

DexWild introduces a dexterous human data collection system that works in diverse environments, without robots. We collect 9,290 human demos across 93 environments and show that cotraining with robot data unlocks generalization. In unseen scenes, Dexwild performs 3.8x better than training with robot data only.

FACTR: Force-Attending Curriculum Training for Contact-Rich Policy Learning

We present a low-cost bilateral teleoperation setup that allows you to feel what the robot is feeling. To leverage this force information, we present FACTR—a state-of-the-art policy learning framework designed to guide policies in attending to force information during training

AnyCar to Anywhere: Learning Universal Dynamics Model for Agile and Adaptive Mobility

AnyCar is a transformer-based dynamics model that can adapt to various vehicles, environments, state estimators, and tasks. We train a generalist model that exceeds specialist model capabilites across various different settings.

Agile Mobility with Rapid Online Adaptation via Meta-learning and Uncertainty-aware MPPI

This paper presents a meta-learning-based model adaptation approach for agile robotic mobility, integrating uncertainty-aware Model Predictive Path Integral (MPPI) control to enable rapid online adaptation to dynamic environments across various wheeled robot platforms.

Linear Delta Arrays for Compliant Dexterous Distributed Manipulation

Delta Arrays demonstrate dexterous manipulation capabilities of a new type of distributed dexterous manipulator. We show a wide range of capablities for performing coordinated distributed manipulation including translation, alignment, prehensile squeezing, lifting, and grasping.

Design and source code from Jon Barron's website