-
-
Notifications
You must be signed in to change notification settings - Fork 1
/
environment.py
158 lines (126 loc) · 6.16 KB
/
environment.py
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
import os
import pickle as pkl
import random
import math
import pygame
from car import Car
from utils import generate_border_walls
import numpy as np
from gym import Env
from gym.spaces import Box, Discrete
class RacerEnvironment(Env):
def __init__(self, render=False, evaluate=False):
super(RacerEnvironment, self).__init__()
self.action_space = Discrete(3)
self.observation_space = Box(low=0, high=1000, shape=(12,), dtype=np.float32)
self.reward = 0
self.render_flag = render
self.evaluate_flag = evaluate
if self.evaluate_flag:
self.tracks = os.listdir('tracks/testing')
self.metadata = pkl.load(open('tracks/testing/'+random.choice(self.tracks),'rb'))
else:
self.tracks = os.listdir('tracks/training')
self.metadata = pkl.load(open('tracks/training/'+random.choice(self.tracks),'rb'))
self.ppu = self.metadata['ppu']
self.timesteps = 0
pygame.init()
width = 1280
height = 720
if self.render_flag:
pygame.display.set_caption("Racer_AI")
self.screen = pygame.display.set_mode((width, height))#, pygame.NOFRAME)
self.screen_dims = (width, height)
self.clock = pygame.time.Clock()
self.ticks = 60
self.done = False
# self.dt = self.clock.get_time() / 100
self.walls = []
self.walls.extend(generate_border_walls(self.screen_dims))
self.walls.extend(self.metadata['walls'])
self.checkpoints = []
self.checkpoints.extend(self.metadata['checkpoints'])
self.car = Car(x=self.metadata['car_x']/self.metadata['ppu'], y=self.metadata['car_y']/self.metadata['ppu'], ppu=self.ppu, angle=self.metadata['car_angle'], screen_width=width, screen_height=height)
self.track_counter = 1
def reset(self):
if self.evaluate_flag:
self.metadata = pkl.load(open('tracks/testing/'+'metadata_{}.pkl'.format(self.track_counter),'rb'))
else:
self.metadata = pkl.load(open('tracks/training/'+'metadata_{}.pkl'.format(self.track_counter),'rb'))
self.walls = []
self.walls.extend(generate_border_walls(self.screen_dims))
self.walls.extend(self.metadata['walls'])
self.checkpoints = []
self.checkpoints.extend(self.metadata['checkpoints'])
self.reward = 0
self.car = Car(x=self.metadata['car_x']/self.metadata['ppu'], y=self.metadata['car_y']/self.metadata['ppu'], ppu=self.ppu, angle=self.metadata['car_angle'])
state = []
state.extend(self.car.state())
state.extend([self.checkpoints[0].state()[0], self.checkpoints[0].state()[1]])
state.extend([self.car.position.x * self.ppu, self.car.position.y * self.ppu])
# nearest_checkpoint_normalized_distance = math.sqrt(((self.car.position.x * self.ppu) - self.checkpoints[0].state()[0]) ** 2 + ((self.car.position.y * self.ppu) - self.checkpoints[0].state()[1]) ** 2) / math.sqrt(self.screen_dims[0]**2 + self.screen_dims[1]**2)
# state.extend([nearest_checkpoint_normalized_distance])
state = np.array(state).astype(np.float32)
self.done = False
self.track_counter += 1
if self.track_counter > len(self.tracks):
self.track_counter = 1
return state
def get_state(self):
state = []
state.extend(self.car.state())
state.extend([self.checkpoints[0].state()[0]/self.screen_dims[0], self.checkpoints[0].state()[1]/self.screen_dims[1]])
state.extend([(self.car.position.x * self.ppu)/self.screen_dims[0], (self.car.position.y * self.ppu)/self.screen_dims[1]])
# nearest_checkpoint_normalized_distance = math.sqrt(((self.car.position.x * self.ppu) - self.checkpoints[0].state()[0]) ** 2 + ((self.car.position.y * self.ppu) - self.checkpoints[0].state()[1]) ** 2) / math.sqrt(self.screen_dims[0]**2 + self.screen_dims[1]**2)
# state.extend([nearest_checkpoint_normalized_distance])
state = np.array(state).astype(float)
return state
def step(self, action):
self.timesteps += 1
info = {}
self.dt = self.clock.get_time() / 200
self.done, reward = self.car.step(action=action, walls=self.walls, dt=self.dt)
if self.timesteps % (1024//2) == 0:
self.done = True
reward -= 5
nearest_checkpoint_normalized_distance = math.sqrt(((self.car.position.x * self.ppu) - self.checkpoints[0].state()[0]) ** 2 + ((self.car.position.y * self.ppu) - self.checkpoints[0].state()[1]) ** 2) / math.sqrt(self.screen_dims[0]**2 + self.screen_dims[1]**2)
checkpoint_distance_reward = (1-nearest_checkpoint_normalized_distance) * 0.5
# print('check point reward: {}'.format(checkpoint_distance_reward))
reward += checkpoint_distance_reward
if self.done:
self.reward = reward
state = self.get_state()
self.render()
self.clock.tick(self.ticks)
return state, self.reward, self.done, {}
checkpoint_collision_flag = self.car.checkpoint_collision(self.checkpoints[0])
if checkpoint_collision_flag:
# print("Checkpoint!")
self.checkpoints.pop(0)
reward = 10
self.timesteps = 1
if len(self.checkpoints) == 0:
print("You win!")
reward = 20
self.done = True
info['done'] = 'You win'
if not self.done:
self.render()
self.clock.tick(self.ticks)
next_state = self.get_state()
else:
next_state = None
# else:
# self.reset()
return next_state, reward, self.done, info#{}
def render(self):
if self.render_flag:
pygame.event.get()
self.screen.fill((0, 0, 0))
for wall in self.walls:
wall.draw(self.screen)
if not self.evaluate_flag:
for checkpoint in self.checkpoints:
checkpoint.draw(self.screen)
self.car.draw(self.screen, self.evaluate_flag)
pygame.display.flip()