gym-coders-strike-back's Introduction

gym-coders-strike-back

This environment is an implementation of the Coders Strike Back contest, hosted by the Codin Game website.

Only single game mode is implemented, i.e. you can control only one pod through a random set of checkpoints, and the game finish when you reach the final checkpoint.

Installation

git clone https://github.com/alenic/gym-coders-strike-back
cd gym-coders-strike-back
python setup.py install

Usage

This piece of code simulates a single episode with this simple policy: go to the next checkpoint with a constant thrust

import gym
from gym_coders_strike_back import *
import time
import numpy as np

env = gym.make('CodersStrikeBack-v0')
fps =  env.metadata.get('video.frames_per_second')

totalReward = 0.0
# Set pseudorandom seed for the getting the same game
env.seed(1234)
state = env.reset()
for i in range(1,10000):
    # display the game
    render = env.render()
    if not render:
        break
    # Take action (simple policy)
    targetX, targetY = state[5:7]
    thrust = 60
    action = np.array([targetX, targetY, thrust], dtype=np.float32)
    # Do a game step
    state, reward, done, _ = env.step(action)

    # Print the state
    
    print('---------- Tick %d' % i)
    print('Pod angle ', state[0])
    print('Pod (x,y): (%d, %d)' % (state[1], state[3]))
    print('Pod velocity (v_x,v_y): (%d, %d)' % (state[2], state[4]))
    print('First Checkpoint (x,y): (%d, %d)' % (state[5],state[6]))
    print('Second Checkpoint (x,y): (%d, %d)' % (state[7],state[8]))
    print('Reward ', reward)
    print('done ', done)

    totalReward += reward
    # Slow down the cycle for a realistic simulation
    time.sleep(1.0/fps)

    # The game end if the flag done is True
    if done:
        break

print("Total reward: ", totalReward)
env.close()

Recommend Projects