In a reinforcement learning scenario, the environment models the dynamics with which the agent interacts. The environment:
Receives actions from the agent
Outputs observations resulting from the dynamic behavior of the environment model
Generates a reward measuring how well the action contributes to achieving the task
You can create predefined and custom environments in MATLAB. For more information, see Create MATLAB Reinforcement Learning Environments.
|Create a predefined reinforcement learning environment|
|Specify custom reinforcement learning environment dynamics using functions|
|Create custom reinforcement learning environment template|
|Create Markov decision process environment for reinforcement learning|
|Create Markov decision process model|
|Create a two-dimensional grid world for reinforcement learning|
|Validate custom reinforcement learning environment|
|Generate a reward function from control specifications to train a reinforcement learning agent|
|Exterior penalty value for a point with respect to a bounded region|
|Hyperbolic penalty value for a point with respect to a bounded region|
|Logarithmic barrier penalty value for a point with respect to a bounded region|
Neural Network Environment
|Environment model with deep neural network transition models|
|Deterministic transition function approximator object for neural network-based environment|
|Stochastic Gaussian transition function approximator object for neural network-based environment|
|Deterministic reward function approximator object for neural network-based environment|
|Stochastic Gaussian reward function approximator object for neural network-based environment|
|Is-done function approximator object for neural network-based environment|
|Predict next observation, next reward, or episode termination given observation and action input data|
|Evaluate function approximator object given observation (or observation-action) input data|
|Option to accelerate computation of gradient for approximator object based on neural network|
|Create discrete action or observation data specifications for reinforcement learning environments|
|Create continuous action or observation data specifications for reinforcement learning environments|
|Obtain action data specifications from reinforcement learning environment, agent, or experience buffer|
|Obtain observation data specifications from reinforcement learning environment, agent, or experience buffer|
- Create MATLAB Reinforcement Learning Environments
Model environment dynamics using a MATLAB object that interacts with the agent, generating rewards and observations in response to agent actions.
- Create or Import MATLAB Environments in Reinforcement Learning Designer
Import a custom MATLAB environment or create a predefined MATLAB environment.
- Define Reward Signals
Create a reward signal that measures how successful the agent is at achieving its goal.
- Load Predefined Control System Environments
Load predefined MATLAB control system environments.
- Load Predefined Grid World Environments
Train agents in predefined MATLAB grid world environments for which the actions, observations, and rewards are already defined.
- Create Custom Grid World Environments
Create custom MATLAB grid world environments by defining your own size, rewards and obstacles.
- Create MATLAB Environment Using Custom Functions
Create a reinforcement learning environment by supplying custom dynamic functions.
- Create Custom MATLAB Environment from Template
Define a custom reinforcement learning environment by creating and modifying a template environment object.