Reinforcement Learning DDPG agent

Hey,

What is the function that does the epsilon gridy operation in dqn in DDPG agent?

My agent stuck in the same results.

Thanks.

Answer 1

[Florian Rosner]

Hey,

Epsilon-greedy is used for discrete action spaces. For continuous action spaces, exploration is controlled by adding noise to the action itself.

For setting the noise options for DDPG just check https://de.mathworks.com/help/reinforcement-learning/ref/rlddpgagentoptions.html . For deeper insight you can check the Noise Model itself on https://de.mathworks.com/help/reinforcement-learning/ref/rlddpgagentoptions.html#mw_2875b71d-bfb0-4be4-b0d3-a44592c3cb30_head or the DDPG Paper https://arxiv.org/pdf/1509.02971.pdf .