All about Deep Reinforcement Learning

I work in TensorFlow2 Framework

Environments

• Snake_v1
  • Observation space: (RGB array or Extracted features)
  • Action space: Discrete(3), [Go Straight / Turn Left / Turn Right]
  • Reward scheme: (Dense or Sparse)

• Snake_v2
  • Observation space: (RGB array)
  • Action space: Box(2), [Speed(0.0 ~ 1.0), Angle(-1.0 ~ 1.0)]
  • Reward scheme: (Dense or Sparse)

RL Agents

• Randomly
• TODOs
  • PPO (Proximal Policy Optimization)
  • DDPG (Deep Deterministic Policy Gradient)
  • A3C (Asynchronous Advantage Actor Critic)
  • DQN (Deep Q-Network)

IL Agents

• TODOs
  • BC (Behavior Cloning)
  • GAIL (Generative Adversarial Imitation Learning)
  • VAIL (Variational GAIL)
  • DI-GAIL (Directed-Info GAIL)