Continuous control with deep reinforcement learning

Code

• github.com/DLR-RM/stable-baselines3
  pytorch★ 12,962
• github.com/facebookresearch/Horizon
  pytorch★ 3,686
• github.com/toni-sm/skrl
  jax★ 1,014
• github.com/tensorlayer/RLzoo
  tf★ 644
• github.com/stevenpjg/ddpg-aigym
  tf★ 276
• github.com/baturaysaglam/RIS-MISO-Deep-Reinforcement-Learning
  pytorch★ 218
• github.com/massquantity/DBRL
  pytorch★ 154
• github.com/Brook1711/RIS_components
  tf★ 77
• github.com/shahin-01/vqa-ad
  pytorch★ 19
• github.com/s-sd/task-amenability
  tf★ 18

Abstract

We adapt the ideas underlying the success of Deep Q-Learning to the continuous action domain. We present an actor-critic, model-free algorithm based on the deterministic policy gradient that can operate over continuous action spaces. Using the same learning algorithm, network architecture and hyper-parameters, our algorithm robustly solves more than 20 simulated physics tasks, including classic problems such as cartpole swing-up, dexterous manipulation, legged locomotion and car driving. Our algorithm is able to find policies whose performance is competitive with those found by a planning algorithm with full access to the dynamics of the domain and its derivatives. We further demonstrate that for many of the tasks the algorithm can learn policies end-to-end: directly from raw pixel inputs.

Tasks

Benchmark Results

Reproductions