Aurick Zhou

About

Aurick Zhou is a leading researcher in deep reinforcement learning (RL), whose work has fundamentally advanced the sample efficiency and practical deployment of model-free algorithms. He is best known as a co-author of the seminal "Soft Actor-Critic Algorithms and Applications" (2018), which has amassed over 1,950 citations. This work introduced a maximum-entropy framework that dramatically improves the stability and robustness of RL training, directly addressing the field's long-standing challenges of high sample complexity and hyperparameter sensitivity. Zhou has also made pivotal contributions to robotic locomotion and manipulation. His highly cited paper "Learning to Walk Via Deep Reinforcement Learning" (2019, over 430 citations) demonstrated how deep RL can automate the acquisition of complex walking controllers directly from sensory inputs, minimizing the need for explicit engineering. Further extending this work, his research on "Composable Deep Reinforcement Learning for Robotic Manipulation" (2018) explored modular skill learning for real-world tasks. Most recently, in "MURAL" (2021), Zhou has tackled the exploration problem by introducing meta-learning for uncertainty-aware rewards, pushing the boundaries of outcome-driven RL. His body of work is essential reading for anyone seeking to understand modern, practical deep RL.

Research Focus

Key Achievements

Top Papers

1. 1
   Soft Actor-Critic Algorithms and Applications

   1,952 citations · 2018
2. 2
   Learning to Walk Via Deep Reinforcement Learning

   434 citations · 2019
3. 3
   Learning to Walk via Deep Reinforcement Learning

   42 citations · 2018
4. 4
   Composable Deep Reinforcement Learning for Robotic Manipulation

   38 citations · 2018
5. 5
   MURAL: Meta-Learning Uncertainty-Aware Rewards for Outcome-Driven Reinforcement Learning

   6 citations · 2021

Key Collaborators

Contact & Links