Daphne Chen

I am a PhD student at the Paul G. Allen School at the University of Washington in the Human-Centered Robotics Lab. Currently, I am also a Research Intern at Microsoft Research (MSR) in the Physically Embodied AI and Robotics group, working on human-in-the-loop robot learning.

I received my Masters of Science in Robotics from the Robotics Institute at Carnegie Mellon University. Previously, I graduated from Georgia Tech, where I got my start in robotics research at the RAIL Lab.

Email  /  GitHub  /  Google Scholar  /  Twitter

Updates

• Summer 2026 internship at Microsoft Research
• Started my PhD at the University of Washington!

Research

I'm interested in machine learning, robotics, and human-centered AI.

A Few Words Go a Long Way: Language Guided Robot Policy Synthesis

We present ARCHITECT, an agentic orchestration framework that synthesizes robot manipulation policies from natural language instructions and corrections, building a persistent skill library that transfers across tasks.

FlowDAgger: Human-in-the-Loop Adaptation of Generative Robot Policies in Latent Space

We present FlowDAgger, a sample- and compute-efficient method for adapting frozen generative robot policies from human interventions in latent space.

Improving Human-AI Coordination through Adversarial Training and Generative Models

We use generative models in an online adversarial training loop to train the learning agent against difficult coordination scenarios while maintaining realistic behavior.

Learning to Cooperate with Humans using Generative Agents

We use generative models to sample infinite human-like partner agents to train a coordinator agent. These agents cooperate well with real human players, achieving better performance compared to baselines FCP, CoMeDi and MEP.

DROID: A Large-Scale In-the-Wild Robot Manipulation Dataset

A large-scale, in-the-wild robot manipulation dataset spanning diverse scenes, tasks, and environments to support generalizable robot learning.

Open X-Embodiment: Robotic Learning Datasets and RT-X Models

A collaborative effort standardizing robot learning datasets across many institutions and embodiments, and training generalist RT-X policies that transfer across robots.

Anticipatory Human-Robot Collaboration via Multi-Objective Trajectory Optimization

We address the problem of adapting robot trajectories to improve safety, comfort, and efficiency in human-robot collaborative tasks. To this end, we propose CoMOTO, a trajectory optimization framework that utilizes stochastic motion prediction to anticipate the human’s motion and adapt the robot’s joint trajectory accordingly.

Benchmark for Skill Learning from Demonstration: Impact of User Experience, Task Complexity, and Start Configuration on Performance

We contribute a study benchmarking the performance of multiple motion-based learning from demonstration approaches.

Design and source code from Jon Barron's website