inputs.py

import numpy as np
np.set_printoptions(linewidth = 200)
from snake import Snake

from cprofiler import profile 

'''
The class for the inputs being fed into the Neural Network
Each instance of the NN requires an instance of Inputs class
'''

class Inputs:
    def __init__(self, settings, snake):
        self.settings = settings
        self.snake = snake
        self.food = self.snake.food
        self.inputVector = []
    
    def update(self, snake):
        self.snake = snake
        self.food = self.snake.food

    
    def generateInputs(self):
        
        inputVector = np.array([0]*24, dtype=float)
        #distances: temp[0] is food, temp[1] is tail, temp[2] is wall
        #look left
        temp = self.look([-10, 0])
        inputVector[0] = temp[0]
        inputVector[1] = temp[1]
        inputVector[2] = temp[2]
        #look left/up  
        temp2 = self.look([-10, -10])
        inputVector[3] = temp2[0]
        inputVector[4] = temp2[1]
        inputVector[5] = temp2[2]
        #look up
        temp3 = self.look([0, -10])
        inputVector[6] = temp3[0]
        inputVector[7] = temp3[1]
        inputVector[8] = temp3[2]
        #look up/right
        temp4 = self.look([10, -10])
        inputVector[9] = temp4[0]
        inputVector[10] = temp4[1]
        inputVector[11] = temp4[2]
        #look right
        temp5 = self.look([10, 0])
        inputVector[12] = temp5[0]
        inputVector[13] = temp5[1]
        inputVector[14] = temp5[2]
        #look right/down
        temp6 = self.look([10, 10])
        inputVector[15] = temp6[0]
        inputVector[16] = temp6[1]
        inputVector[17] = temp6[2]
        #look down
        temp7 = self.look([0, 10])
        inputVector[18] = temp7[0]
        inputVector[19] = temp7[1]
        inputVector[20] = temp7[2]
        #look down/left
        temp8 = self.look([-10, 10])
        inputVector[21] = temp8[0]
        inputVector[22] = temp8[1]
        inputVector[23] = temp8[2]
        self.inputVector = inputVector

    
    def look(self, dir): #dir is a vector
        pos = np.array([self.snake.x, self.snake.y], dtype=float)
        dir = np.array(dir, dtype=float)
        vision = np.array([0]*3, dtype=float)
        distance = 1
        pos += dir
        
        foodFound = False
        tailFound = False

        while(not(pos[0] < 0 or pos[1] < 0 or pos[0] >= self.settings.xyMaxValue or pos[1] >= self.settings.xyMaxValue)):            
            #check for food
            if(not foodFound and pos[0] == self.snake.food.x and pos[1] == self.snake.food.y):
                vision[0] = 1
                foodFound = True
            #check for tail
            if(not tailFound and self.isOnTail(pos)):
                vision[1] = 1/distance
                tailFound = True
            #reach wall
            pos += dir
            distance += 1
        vision[2] = 1 / distance
        return(vision) 

    def isOnTail(self, pos):
        for i in range(len(self.snake.tail)):
            if(pos[0] == self.snake.tail[i][0] and pos[1] == self.snake.tail[i][1]):
                return True
        return False