search

markusdumke / reinforcelearn

R Package for Reinforcement Learning
https://markusdumke.github.io/reinforcelearn/
Other
33 stars 6 forks source link

what happens in eg mountaincar if the car is not pushed to the top? #21

Closed berndbischl closed 6 years ago

berndbischl commented 6 years ago

so the question is: what if the episode does not end in a terminal state? is there a max length? what is the reward then?

markusdumke commented 6 years ago

You can set a maximal number of allowed steps per episode in interact.

library(reinforcelearn)
env = makeEnvironment("mountain.car")
agent = makeAgent("random")
interact(env, agent, n.episodes = 10L, max.steps.per.episode = 50L)