gameBot1.py

# imports two libraries: gym, which is from open ai and supports a bunch of
# machine learning stuff, and universe, which im not too sure does
# ah there it is, univers let syou run s many environmetns as you want in parallel
import gym
import universe

# using neon racers as the test environment; this just configures the env based on gym syntax
env = gym.make('flashgames.NeonRace-v0')
env.configure(remotes=1) # creates a local docker container
# i guess creating this container makes it portable. i dont really know lol
# also id guess you HAVE TO HAVE docker running while this is running as well

# env.reset() generates an observation that is specific to the environment (number of pixels,
# score, etc.) - this initiates theenvironment and gets a list of the observatiosn of its initial state
observation_n = env.reset()

# just an infinite loop where up arrow is pressed for all observations
while True:
    action_n = [[('KeyEvent', 'ArrowUp', True)] for ob in observation_n] # this is the agent
    observation_n, reward_n, done_n, info = env.step(action_n) # reinforcement learning action by agent
    print('observation: ', observation_n) # prints out the observation information for debugging purposes
    # and so we can understand what exactly the observation_n is itself
    print('reward: ', reward_n) # see if the reward was +/-1
    env.render() # run agent on environment - i suppose that this needs to be run every time, because
    # the agent must be run every time its in the loop (thats how the agent keeps running)