Papers
9
Total Citations
431
H-Index
6
About
Adrian Li is a leading researcher in robotics and artificial intelligence, specializing in human-robot collaboration, reinforcement learning, and real-world robotic manipulation. His most impactful work, "Asking for Help Using Inverse Semantics" (167 citations), pioneered a novel approach enabling robots to autonomously detect failures and communicate specific needs to human partners using natural language—a critical step toward robust, collaborative robots. Li further advanced the field with "Collective Robot Reinforcement Learning with Distributed Asynchronous Guided Policy Search" (139 citations), which demonstrated how multiple robots can efficiently learn complex, generalizable skills across diverse real-world conditions. His more recent contributions include deploying deep reinforcement learning at scale for waste sorting in office buildings using a fleet of mobile manipulators, showcasing the practical impact of his methods. Li has also developed probabilistic multi-modal actor models for vision-based grasping and explored predictive information to improve multi-task robotic learning. With over 400 total citations, his work bridges the gap between theoretical RL advances and deployable robotic systems, making him a key figure in creating robots that can learn, fail gracefully, and collaborate effectively with humans.
Research Focus
Key Achievements
Top Papers
- 1Asking for Help Using Inverse Semantics167 citations · 2014
- 2
- 3Recovering from failure by asking for help79 citations · 2015
- 4
- 5
- 6
- 7
- 8
- 9