Papers
17
Total Citations
3,254
H-Index
10
About
Tuomas Haarnoja is a pioneering researcher in deep reinforcement learning (RL) and robotics, best known for his transformative contributions to maximum entropy RL frameworks. His most influential work, the Soft Actor-Critic (SAC) algorithm, has amassed nearly 2,000 citations and fundamentally addressed two persistent challenges in model-free deep RL: sample inefficiency and hyperparameter sensitivity. Building on earlier foundational work in energy-based policies and soft Q-learning (434 citations), Haarnoja established a coherent theoretical and practical framework for learning stochastic, entropy-maximizing policies that generalize robustly across complex control tasks. Beyond algorithmic development, Haarnoja has demonstrated remarkable success translating these methods to real-world robotics. His research spans robotic locomotion, humanoid soccer playing, and vision-guided bipedal motion, including a highly cited study on teaching robots to walk via deep RL and groundbreaking work on agile soccer skills for humanoid robots (147 citations). His more recent contributions incorporate sim-to-real transfer using Neural Radiance Fields and motion imitation from human and animal behavior, reflecting a broad vision for generalizable robot intelligence. With over 3,200 cumulative citations, Haarnoja's work has profoundly shaped modern RL research and its applications in autonomous robotics.
Research Focus
Key Achievements
Top Papers
- 1Soft Actor-Critic Algorithms and Applications1,952 citations · 2018
- 2Learning to Walk Via Deep Reinforcement Learning434 citations · 2019
- 3Reinforcement Learning with Deep Energy-Based Policies434 citations · 2017
- 4
- 5Backprop KF: Learning Discriminative Deterministic State Estimators102 citations · 2016
- 6
- 7Learning to Walk via Deep Reinforcement Learning42 citations · 2018
- 8Composable Deep Reinforcement Learning for Robotic Manipulation38 citations · 2018
- 9
- 10