interact: Interaction between agent and environment.

In markdumke/reinforcelearn: Reinforcement Learning

Interaction between agent and environment.

Description

Run interaction between agent and environment for specified number of steps or episodes.

Usage

interact(env, agent, n.steps = Inf, n.episodes = Inf,
  max.steps.per.episode = Inf, learn = TRUE, visualize = FALSE)

Arguments

env	[Environment] Reinforcement learning environment created by makeEnvironment.
agent	[Agent] Agent created by makeAgent.
n.steps	[integer(1)] Number of steps to run.
n.episodes	[integer(1)] Number of episodes to run.
max.steps.per.episode	[integer(1)] Maximal number of steps allowed per episode.
learn	[logical(1)] Should the agent learn?
visualize	[logical(1)] Visualize the interaction between agent and environment?

Value

[list] Return and number of steps per episode.

Examples

env = makeEnvironment("windy.gridworld")
agent = makeAgent("softmax", "table", "qlearning")
interact(env, agent, n.episodes = 10L)

markdumke/reinforcelearn documentation built on Nov. 17, 2022, 12:53 a.m.