AI-Learning

Learning and Artificial Intelligence in Robotics

Contents:

• Fixed Topology Neural Network Search.
• Q-Learning.
• Deep RL Policy Network.
• Deep Q-Learning (DQN).
• Deep Q-Learning (DQN) +[target network, reward clipping, frame skipping].
• Deep Deterministic Policy Gradient (DDPG).
• Deep Neuroevolution. [working, no results to show!]

---

Results:

Random Neural Network

MountainCar-v0:

• Using random weight search for fixed topology neural network.

  

---

Q-Learning

CartPole-v1:

• Tabular Q-Learning.

  

---

Deep Q-Learning

LunarLander-v2:

• Deep Q Learning with frame skipping(repeat same action for 3 frames), target network updated at (epsiode%2==0) & reward clipping(-1,1). landing at epsiode 720:

  

  paremeters, for below: - refresh target net every 10 episodes. - skip 3 frames. - minibatch size 32. - at episode 460.

  

---

Deep RL Policy

LunarLander-v2:

---

Deep Deterministic Policy Gradient (DDPG).

InvertedPendulum-v2:

Pendulum-v0:

InvertedDoublePendulum-v2:

Link: https://youtu.be/fXbqDDaJDvg

Reacher-v2:

--------

Interesting Resouces:

• Welcoming the Era of Deep Neuroevolution: https://eng.uber.com/deep-neuroevolution/?lipi=urn%3Ali%3Apage%3Ad_flagship3_feed%3BzyxkMF5OTd%2BI48jAyJJ%2B2A%3D%3D
• MIT Deep-RL self-driving cars: https://selfdrivingcars.mit.edu
• Deep RL lecture by David Silver UCL: http://www0.cs.ucl.ac.uk/staff/d.silver/web/Resources_files/deep_rl.pdf
• DQN Nature paper, 'Human-level control through deep reinforcement learning' (2015)
• DDPG Paper, 'Continuous control with deep reinforcement learning' https://arxiv.org/abs/1509.02971
• Deep Neuroevolution @ UberAI Labs, https://arxiv.org/abs/1712.06567