Research

Affordance Learning Affordance-Guided Reinforcement Learning via Visual Prompting [paper] [site]
Olivia Y. Lee, Annie Xie, Kuan Fang, Karl Pertsch, and Chelsea Finn
Robotics Science and Systems (RSS), 2024: Task Specification & Lifelong Robot Learning
We present an approach that extracts affordance representations from VLMs via keypoints to define dense shaping rewards for online reinforcement learning. We pretrain policies on Bridge data and a modest number of in-domain demonstrations of the task, and finetune policies online using VLM-generated dense rewards. On a real-world manipulation task, our VLM-generated rewards improve the sample efficiency of autonomous RL, enabling task completion in 20K online finetuning steps. Our approach is robust to reducing in-domain demonstrations used for pretraining, reaching comparable performance in 35K online finetuning steps.

Play it by Ear Play it by Ear: Learning Skills amidst Occlusion through Audio-Visual Imitation Learning [paper] [site] [publication]
Maximilian Du*, Olivia Y. Lee*, Suraj Nair, and Chelsea Finn
Robotics Science and Systems (RSS), 2022 [recording]
We present a system that learns to complete partially-observed manipulation tasks by reasoning over vision, audio, and memory. We combine offline imitation learning from a modest number of tele-operated demonstrations and online finetuning with human interventions. In simulation, our system benefits from using audio and online interventions, which improve success rates of offline imitation learning by ~20%. On a Franka Emika Panda robot, our system completes manipulation tasks under occlusion with a 70% success rate, 50% higher than a policy that does not use audio.

Satellite-QKD
An updated analysis of satellite quantum-key distribution missions [paper]
Olivia Y. Lee, Tom Vergoossen
We provide a meta-analysis of research and development efforts towards the realization of satellite quantum-key distribution (satellite-QKD). We analyze technical parameters of protocols and infrastructure for performing satellite-based QKD. Following a timeline of key milestones, we summarize globally completed and proposed quantum satellite missions, categorized by specific advancements thus far in satellite-QKD. We conclude by discussing the current technical in satellite-QKD and future research directions to address them.