basic_gamebot.py

import gym
import universe # register the universe environments
import random

env = gym.make('flashgames.NeonRace-v0')
env.configure(remotes=1) # automatically creates a local docker container
observation_n = env.reset()

# Create variable for moving the car
goleft = [('KeyEvent', 'ArrowUp', True), ('KeyEvent', 'ArrowLeft', True),
        ('KeyEvent', 'ArrowRight', False)]
goright = [('KeyEvent', 'ArrowUp', True), ('KeyEvent', 'ArrowLeft', False),
         ('KeyEvent', 'ArrowRight', True)]

# Move the car forward with nitroboost
boostforward = [('KeyEvent', 'ArrowUp', True), ('KeyEvent', 'ArrowRight', False),
       ('KeyEvent', 'ArrowLeft', False), ('KeyEvent', 'n', True)]


sum_reward = 0
turn = 0
rewards = []
buffer_size = 100
action = boostforward

while True:

    turn -= 1
    if turn <= 0:
        action = boostforward
        turn = 0
    # choose action based on speed
    action_n = [action for ob in observation_n]

    # perform action
    observation_n, reward_n, done_n, info = env.step(action_n)

    sum_reward += reward_n[0]

    rewards += [reward_n[0]]
    #if stuck, try going one side for some time
    if len(rewards) >= buffer_size:
        mean = sum(rewards)/len(rewards)

        if mean == 0:
            turn = 25
            if random.random() < 0.5:
                action = goleft
            else:
                action = goright

        rewards = []

    env.render()