rlTRPOAgentOptions

Options for TRPO agent

Since R2021b

Description

Use an rlTRPOAgentOptions object to specify options for trust region policy optimization (TRPO) agents. To create a TRPO agent, use rlTRPOAgent.

For more information on TRPO agents, see Trust Region Policy Optimization (TRPO) Agents.

For more information on the different types of reinforcement learning agents, see Reinforcement Learning Agents.

Creation

Syntax

opt = rlTRPOAgentOptions

opt = rlTRPOAgentOptions(Name=Value)

Description

opt = rlTRPOAgentOptions creates an rlPPOAgentOptions object for use as an argument when creating a TRPO agent using all default settings. You can modify the object properties using dot notation.

opt = rlTRPOAgentOptions(Name=Value) creates the options set opt and sets its properties using one or more name-value arguments. For example, rlTRPOAgentOptions(DiscountFactor=0.95) creates an option set with a discount factor of 0.95. You can specify multiple name-value arguments.

Properties

ExperienceHorizon — Number of steps the agent interacts with the environment before learning

512 (default) | positive integer

Number of steps the agent interacts with the environment before learning from its experience, specified as a positive integer. When the agent is trained in parallel, ExperienceHorizon is ignored, and the whole episode is used to compute the gradients.

The ExperienceHorizon value must be greater than or equal to the MiniBatchSize value.

Example: ExperienceHorizon=512

MiniBatchSize — Mini-batch size

128 (default) | positive integer

Mini-batch size used for each learning epoch, specified as a positive integer. When the agent uses a recurrent neural network, MiniBatchSize is treated as the training trajectory length.

The MiniBatchSize value must be less than or equal to the ExperienceHorizon value.

Example: MiniBatchSize=256

EntropyLossWeight — Entropy loss weight

0.01 (default) | scalar value between 0 and 1

Entropy loss weight, specified as a scalar value between 0 and 1. A higher entropy loss weight value promotes agent exploration by applying a penalty for being too certain about which action to take. Doing so can help the agent move out of local optima.

When gradients are computed during training, an additional gradient component is computed for minimizing the entropy loss. For more information, see Entropy Loss.

Example: EntropyLossWeight=0.02

NumEpoch — Number of epochs

1 (default) | positive integer

Number of epochs for which the actor and critic networks learn from the current experience set, specified as a positive integer.

Example: NumEpoch=2

AdvantageEstimateMethod — Method for estimating advantage values

"gae" (default) | "finite-horizon"

Method for estimating advantage values, specified as one of the following:

• "gae" — Generalized advantage estimator

• "finite-horizon" — Finite horizon estimation

For more information on these methods, see the training algorithm information in Proximal Policy Optimization (PPO) Agents.

Example: AdvantageEstimateMethod="finite-horizon"

GAEFactor — Smoothing factor for generalized advantage estimator

0.95 (default) | scalar value between 0 and 1

Smoothing factor for generalized advantage estimator, specified as a scalar value between 0 and 1, inclusive. This option applies only when the AdvantageEstimateMethod option is "gae"

Example: GAEFactor=0.97

ConjugateGradientDamping — Conjugate gradient damping factor

0.01 (default) | nonnegative scalar

Conjugate gradient damping factor for numerical stability, specified as a nonnegative scalar.

Example: ConjugateGradientDamping=0.05

KLDivergenceLimit — Upper limit for KL divergence

0.01 (default) | positive scalar

Upper limit for the Kullback-Leibler (KL) divergence between the old policy and the current policy, specified as a positive scalar.

Example: KLDivergenceLimit=0.02

NumIterationsConjugateGradient — Maximum number of iterations for conjugate gradient descent

10 (default) | positive integer

Maximum number of iterations for conjugate gradient descent, specified as positive integer.

Example: NumIterationsConjugateGradient=30

NumIterationsLineSearch — Number of iterations for line search

10 (default) | positive integer

Number of iterations for line search, specified as a positive integer.

Typically, the default value works well for most cases.

Example: NumIterationsLineSearch=5

ConjugateGradientResidualTolerance — Conjugate gradient residual tolerance factor

1e-8 (default) | positive scalar

Conjugate gradient residual tolerance, specified as a positive scalar. Once the residual for the conjugate gradient algorithm is below this tolerance, the algorithm stops.

Typically, the default value works well for most cases.

Example: ConjugateGradientResidualTolerance=2e-8

NormalizedAdvantageMethod — Method for normalizing advantage function

"none" (default) | "current | "moving"

Method for normalizing advantage function values, specified as one of the following:

• "none" — Do not normalize advantage values

• "current" — Normalize the advantage function using the mean and standard deviation for the current mini-batch of experiences.

• "moving" — Normalize the advantage function using the mean and standard deviation for a moving window of recent experiences. To specify the window size, set the AdvantageNormalizingWindow option.

In some environments, you can improve agent performance by normalizing the advantage function during training. The agent normalizes the advantage function by subtracting the mean advantage value and scaling by the standard deviation.

Example: NormalizedAdvantageMethod="moving"

AdvantageNormalizingWindow — Window size for normalizing advantage function

1e6 (default) | positive integer

Window size for normalizing advantage function values, specified as a positive integer. Use this option when the NormalizedAdvantageMethod option is "moving".

Example: AdvantageNormalizingWindow=1e5

CriticOptimizerOptions — Critic optimizer options

rlOptimizerOptions object

Critic optimizer options, specified as an rlOptimizerOptions object. It allows you to specify training parameters of the critic approximator such as learning rate, gradient threshold, as well as the optimizer algorithm and its parameters. For more information, see rlOptimizerOptions and rlOptimizer.

Example: CriticOptimizerOptions = rlOptimizerOptions(LearnRate=5e-3)

SampleTime — Sample time of agent

1 (default) | positive scalar | -1

Sample time of agent, specified as a positive scalar or as -1. Setting this parameter to -1 allows for event-based simulations.

Within a Simulink^® environment, the RL Agent block in which the agent is specified to execute every SampleTime seconds of simulation time. If SampleTime is -1, the block inherits the sample time from its parent subsystem.

Within a MATLAB^® environment, the agent is executed every time the environment advances. In this case, SampleTime is the time interval between consecutive elements in the output experience returned by sim or train. If SampleTime is -1, the time interval between consecutive elements in the returned output experience reflects the timing of the event that triggers the agent execution.

Example: SampleTime=-1

DiscountFactor — Discount factor

0.99 (default) | positive scalar less than or equal to 1

Discount factor applied to future rewards during training, specified as a positive scalar less than or equal to 1.

Example: DiscountFactor=0.9

Object Functions

rlTRPOAgent

Trust region policy optimization (TRPO) reinforcement learning agent

Examples

Create TRPO Agent Options Object

Create a TRPO agent options object, specifying the discount factor.

opt = rlTRPOAgentOptions(DiscountFactor=0.9)

opt = 
  rlTRPOAgentOptions with properties:

                            SampleTime: 1
                        DiscountFactor: 0.9000
                     EntropyLossWeight: 0.0100
                     ExperienceHorizon: 512
                         MiniBatchSize: 128
                              NumEpoch: 3
                CriticOptimizerOptions: [1x1 rl.option.rlOptimizerOptions]
              ConjugateGradientDamping: 0.1000
                     KLDivergenceLimit: 0.0100
        NumIterationsConjugateGradient: 10
               NumIterationsLineSearch: 10
    ConjugateGradientResidualTolerance: 1.0000e-08
               AdvantageEstimateMethod: "gae"
                             GAEFactor: 0.9500
             NormalizedAdvantageMethod: "none"
            AdvantageNormalizingWindow: 1000000
                            InfoToSave: [1x1 struct]

You can modify options using dot notation. For example, set the agent sample time to 0.1.

opt.SampleTime = 0.1;

Version History

Introduced in R2021b

R2022a: Simulation and deployment: UseDeterministicExploitation will be removed

The property UseDeterministicExploitation of the rlTRPOAgentOptions object will be removed in a future release. Use the UseExplorationPolicy property of rlTRPOAgent instead.

Previously, you set UseDeterministicExploitation as follows.

• Force the agent to always select the action with maximum likelihood, thereby using a greedy deterministic policy for simulation and deployment.

  agent.AgentOptions.UseDeterministicExploitation = true;

• Allow the agent to select its action by sampling its probability distribution for simulation and policy deployment, thereby using a stochastic policy that explores the observation space.

  agent.AgentOptions.UseDeterministicExploitation = false;

Starting in R2022a, set UseExplorationPolicy as follows.

• Force the agent to always select the action with maximum likelihood, thereby using a greedy deterministic policy for simulation and deployment.

  agent.UseExplorationPolicy = false;

• Allow the agent to select its action by sampling its probability distribution for simulation and policy deployment, thereby using a stochastic policy that explores the observation space.

  agent.UseExplorationPolicy = true;

Similarly to UseDeterministicExploitation, UseExplorationPolicy affects only simulation and deployment; it does not affect training.