Papers
15
Total Citations
2,287
H-Index
10
About
Shixiang Gu is a prominent researcher at the intersection of deep reinforcement learning (RL) and robotics, whose work has fundamentally advanced how autonomous systems acquire complex behavioral skills. His most influential contribution, "Deep Reinforcement Learning for Robotic Manipulation with Asynchronous Off-Policy Updates" (2017), garnered over 1,450 citations and demonstrated how asynchronous, sample-efficient RL methods could enable robots to learn manipulation tasks with minimal human intervention. Gu has consistently tackled one of the field's central challenges — sample inefficiency — through innovations like model-based acceleration in continuous deep Q-learning (337 citations) and deployment-efficient offline optimization. His work on hierarchical reinforcement learning (265 citations) extended RL's reach to more complex, real-world tasks, while his "Leave No Trace" framework introduced principled approaches for safe, autonomous environment resetting during training. More recently, Gu has explored unsupervised off-policy learning to develop reward-free robotic skills and applied data-driven methods to household robot challenges. Across his career, his research has collectively shaped modern robotic learning pipelines, emphasizing autonomy, safety, and practical deployability — making him a highly influential figure for students and practitioners in AI-driven robotics.
Research Focus
Key Achievements
Top Papers
- 1
- 2Continuous Deep Q-Learning with Model-based Acceleration337 citations · 2016
- 3Data-Efficient Hierarchical Reinforcement Learning265 citations · 2018
- 4
- 5
- 6
- 7
- 8
- 9
- 10