Papers on RL Applications (Hall of Fame)

These papers are presented in chronological order.

• Temporal Difference Learning and TD-Gammon (Tesauro 1995)
  • Q-Learning with a Multi-Layer Perceptron to learn backgammon playing entirely by reinforcement learning and self-play
• Playing Atari with Deep Reinforcement Learning (Mnih et al. 2013)
  • Deep Q-Learning applied to Atari 2600 games
• Human-level control through deep reinforcement learning (Mnih et al. 2015)
  • DQN applied to Atari 2600 games
• Mastering the game of Go with deep neural networks and tree search (Silver et al. 2016)
  • Supervised learning with MCTS applied to the game of Go (AlphaGo)
• Learning Complex Dexterous Manipulation with Deep Reinforcement Learning and Demonstrations (Rajesweran et al. 2018)
  • Natural policy gradients (NPG) bootstrapped with imitation learning for dexterous manipulation with a 24 DoF hand
• Dota 2 with Large Scale Deep Reinforcement Learning (OpenAI 2019)
  • OpenAI Five became the first AI system to defeat the world champions at an esports game, Dota 2, in 2019
• Grandmaster level in StarCraft II using multi-agent reinforcement learning (Vinyals et al. 2019)
  • AlphaStar for StarCraft II, from the AlphaGo team - outperforms above 99.8% of officially ranked human players
• Outracing champion Gran Turismo drivers with deep reinforcement learning (Wurman et al. 2022)
  • Quantile regression soft-actor critic (SAC) applied to PlayStation 4 (PS4) game Gran Turismo (GT) Sport
• Champion-level drone racing using deep reinforcement learning (Kaufmann et al. 2023)
  • Model-free on-policy RL applied to autonomous first-person view (FPV) drone racing