The deep deterministic policy gradient (DDPG) algorithm is a model-free, online, off-policy reinforcement learning method. A DDPG agent is an actor-critic reinforcement learning agent that computes an optimal policy that maximizes the long-term reward.
For more information on the different types of reinforcement learning agents, see Reinforcement Learning Agents.
DDPG agents can be trained in environments with the following observation and action spaces.
|Observation Space||Action Space|
|Continuous or discrete||Continuous|
During training, a DDPG agent:
Updates the actor and critic properties at each time step during learning.
Stores past experience using a circular experience buffer. The agent updates the actor and critic using a mini-batch of experiences randomly sampled from the buffer.
Perturbs the action chosen by the policy using a stochastic noise model at each training step.
To estimate the policy and value function, a DDPG agent maintains four function approximators:
Actor μ(S) — The actor takes observation S and outputs the corresponding action that maximizes the long-term reward.
Target actor μ'(S) — To improve the stability of the optimization, the agent periodically updates the target actor based on the latest actor parameter values.
Critic Q(S,A) — The critic takes observation S and action A as inputs and outputs the corresponding expectation of the long-term reward.
Target critic Q'(S,A) — To improve the stability of the optimization, the agent periodically updates the target critic based on the latest critic parameter values.
Both Q(S,A) and Q'(S,A) have the same structure and parameterization, and both μ(S) and μ'(S) have the same structure and parameterization.
When training is complete, the trained optimal policy is stored in actor μ(S).
For more information on creating actors and critics for function approximation, see Create Policy and Value Function Representations.
To create a DDPG agent:
DDPG agents use the following training algorithm, in which they update their actor and
critic models at each time step. To configure the training algorithm, specify options using
Initialize the critic Q(S,A) with random parameter values θQ, and initialize the target critic with the same random parameter values: .
Initialize the actor μ(S) with random parameter values θμ, and initialize the target actor with the same parameter values: .
For each training time step:
For the current observation S, select action A = μ(S) +
N, where N is stochastic noise from the noise
model. To configure the noise model, use the
Execute action A. Observe the reward R and next observation S'.
Store the experience (S,A,R,S') in the experience buffer.
Sample a random mini-batch of M experiences
from the experience buffer. To specify M, use the
If S'i is a terminal state, set the value function target yi to Ri. Otherwise set it to:
The value function target is the sum of the experience reward
Ri and the discounted future reward.
To specify the discount factor γ, use the
To compute the cumulative reward, the agent first computes a next action by passing the next observation Si' from the sampled experience to the target actor. The agent finds the cumulative reward by passing the next action to the target critic.
Update the critic parameters by minimizing the loss L across all sampled experiences.
Update the actor parameters using the following sampled policy gradient to maximize the expected discounted reward.
Here, Gai is the gradient of the critic output with respect to the action computed by the actor network, and Gμi is the gradient of the actor output with respect to the actor parameters. Both gradients are evaluated for observation Si.
Update the target actor and critic parameters depending on the target update method. For more information see Target Update Methods.
For simplicity, the actor and critic updates in this algorithm show a gradient update
using basic stochastic gradient descent. The actual gradient update method depends on the
optimizer specified using
DDPG agents update their target actor and critic parameters using one of the following target update methods.
Smoothing — Update the target parameters at every
time step using smoothing factor τ. To specify the smoothing factor,
Periodic — Update the target parameters
periodically without smoothing (
TargetSmoothFactor = 1). To specify
the update period, use the
Periodic Smoothing — Update the target parameters periodically with smoothing.
To configure the target update method, create a
object, and set the
TargetSmoothFactor parameters as shown in the following table.
|Smoothing (default)||Less than |
|Periodic||Greater than |
|Periodic smoothing||Greater than ||Less than |
 T. P. Lillicrap, J. J. Hunt, A. Pritzel, N. Heess, T. Erez, Y. Tassa, D. Silver, and D. Wierstra. “Continuous control with deep reinforcement learning,” International Conference on Learning Representations, 2016.