In order to make my own environment and use the some codes of github, I need to see what does happen inside gym.make('env') for instance gym.make('carpole0')
Where inside the gym github, I am able to find it? I found https://github.com/openai/gym/blob/master/gym/envs/classic_control/cartpole.py but it does not have make?
How to write the update section of defining an environment (env) for DQN which is not in gym library?
I am looking for an environment definition "env" example on github or another resource that is not designed for Atari games. I saw several models but most of them use OpenAI's gym library and are written for playing the Atari games which have relatively simple environments. I am looking for a game environment with more complicated states.
I want to write an update function (step function of environment) for the state t+1 based on state t. What is my problem is that if the state depends on more than one state before how do I implement that? I am looking for an example to demonstrate this. It seems to have an obligation to send the time t in environment.
It would be more helpful for me if an example is defined for an adaptive control problem.
Store all states of the environment that occur in an array or dictionary.
If your environment needs access to previous states before t to determine the next state t + 1, those states will be in the array.
# array that maintains list of all states the agents experiences
states_experienced = []
# each time a new state is encountered, add it to the array
states_experienced.append(current_state)
If order doesn't matter or you'd like to index the states by keys, you can use a dictionary instead.