adamprice97

cartpole = CartpoleFitness(200)
timeTotal = 0
iterationTotal = 0
for i in range(1, 11): #Solve Cartpole 10 times and average results.
    start = time.time()
    #weightSpaceBounds numberOfParticles, inertia, initialVelocityBounds, localWeight, globalWeight
    solver = ParticleSwarmOptimisation(cartpole.evaluate, (-8,8), 15, 0.5, 0.2, 2, 2)
    iterationTotal += solver.maxamise(25, 195)
    end = time.time() - start
    print('Trail ' + str(i) + ' solved in ' + str(k) + ' iterations and ' + "{:.1f}".format(end) + ' seconds.')

adamprice97

class CartpoleFitness(object):

    def __init__(self, terminationStep):
        self.env = gym.make('CartPole-v1')
        self.terminationStep = terminationStep

    def probabilities(self, state, weights):
	    z = state.dot(weights)
	    exp = np.exp(z)
	    return exp/np.sum(exp)

adamprice97

class ParticleSwarmOptimisation(object):
    
    def __init__(self, fitnessFunction, bounds, numParticles, omega, v, phiL, phiG):
        self.bestGlobalFitness = -np.inf; self.bestGlobalPos = []; self.fitnessFunction = fitnessFunction; self.bounds = bounds 
        self.swarm = [] #Create swarm
        for i in range(numParticles):
            self.swarm.append(Particle(bounds, omega, v, phiL, phiG))

    def maxamise(self, maxIterations, target):
        bestGlobalFitness = -np.inf; bestGlobalPos = []                

adamprice97

for i in resample:
    avg = 0
    for k in range(100):
        state = env.reset()
        policy = createPolicy(i)
        step = 0
        while True:
            step += 1
            state = observationSpace.Discretise(state[2:])
            action = policy[state]

adamprice97

env = gym.make('CartPole-v1')
thetaHigh = 10 * 2 * np.pi / 360
high = np.array([thetaHigh, np.radians(15)])
observationSpace = DiscreteBox(-high, high, (3,3))
resample = []
for i in range(512):
    state = env.reset()
    policy = createPolicy(i)
    step = 0
    while True:

adamprice97

def createPolicy(id):
    binary = unpackbits(np.array([id]), 9) #generate 9 long binary string
    return np.reshape(binary, (3,3)) #reshape to form policy

#Credit for this function https://stackoverflow.com/a/51509307
def unpackbits(x, num_bits):
          xshape = list(x.shape)
          x = x.reshape([-1,1])
          to_and = 2**np.arange(num_bits).reshape([1,num_bits])
          return (x & to_and).astype(bool).astype(int).reshape(xshape + [num_bits])

adamprice97

class DiscreteBox(object):
    def __init__(self, low, high, shape):
        self.low, self.high, self.shape = low, high, shape

    def Discretise(self, state):   
        discreteState = [int(np.floor((state[i] - self.low[i])/(self.high[i]-self.low[i])*(self.shape[i]-1))) for i in range(len(state))]
        return tuple([np.min([self.shape[i]-1, np.max([discreteState[i], 0])]) for i in range(len(state))])

adamprice97

import networkx as nx
import matplotlib.pyplot as plt 
import colorsys
import numpy as np
import queue

#Input - G: networkX graph
#Output - cSh: array of ShapelyBetweeness of each node
def ShapelyBetweeness(self, G):
    #Distance between nodes

adamprice97

import pygame, sys
from pygame.locals import *
from Paddle import *
from SimpleBall import *
from Scorer import *
from random import randint
import numpy as np

### Credit to Will Baumbach for the code that   ###
### formed the starting point of this game      ###