forked from WarwickMasson/aaai-platformer
-
Notifications
You must be signed in to change notification settings - Fork 0
/
simulator.py
339 lines (301 loc) · 11.9 KB
/
simulator.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
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
'''
This file implements the platformer rules.
'''
import numpy as np
from numpy.random import uniform
from util import vector
def bound(value, lower, upper):
''' Clips off a value which exceeds the lower or upper bounds. '''
if value < lower:
return lower
elif value > upper:
return upper
else:
return value
def bound_vector(vect, xmax, ymax):
''' Bounds a vector between a negative and positive maximum range. '''
xval = bound(vect[0], -xmax, xmax)
yval = bound(vect[1], -ymax, ymax)
return vector(xval, yval)
WIDTH1 = 250
WIDTH2 = 275
WIDTH3 = 50
GAP1 = 225
GAP2 = 235
HEIGHT1 = 0.0
HEIGHT2 = 0.0
HEIGHT3 = 0.0
MAX_HEIGHT = max(1.0, HEIGHT1, HEIGHT2, HEIGHT3)
MAX_PLATWIDTH = max(WIDTH1, WIDTH2, WIDTH3)
PLATHEIGHT = 40.0
MAX_WIDTH = WIDTH1 + WIDTH2 + WIDTH3 + GAP1 + GAP2
MAX_GAP = max(GAP1, GAP2)
DT = 0.05
MAX_DX = 100.0
MAX_DY = 200.0
MAX_DX_ON = 70.0
MAX_DDX = (MAX_DX - MAX_DX_ON) / DT
MAX_DDY = MAX_DY / DT
ENEMY_SPEED = 30.0
LEAP_DEV = 1.0
HOP_DEV = 1.0
ENEMY_NOISE = 0.5
CHECK_SCALE = False
GRAVITY = 9.8
def scale_state(state):
''' Scale state variables between 0 and 1. '''
new_state = np.copy(state)
scaled = (new_state + SHIFT_VECTOR) / SCALE_VECTOR
if CHECK_SCALE:
for i in range(scaled.size):
if not 0 <= scaled[i] <= 1:
print i, scaled[i], new_state[i]
assert 0 <= scaled[i] <= 1
return scaled
def platform_features(state):
''' Compute the implicit features of the platforms. '''
xpos = state[0]
if xpos < WIDTH1 + GAP1:
pos = 0.0
wd1 = WIDTH1
wd2 = WIDTH2
gap = GAP1
diff = HEIGHT2 - HEIGHT1
elif xpos < WIDTH1 + GAP1 + WIDTH2 + GAP2:
pos = WIDTH1 + GAP1
wd1 = WIDTH2
wd2 = WIDTH3
gap = GAP2
diff = HEIGHT3 - HEIGHT2
else:
pos = WIDTH1 + GAP1 + WIDTH2 + GAP2
wd1 = WIDTH3
wd2 = 0.0
gap = 0.0
diff = 0.0
return [wd1 / MAX_PLATWIDTH, wd2 / MAX_PLATWIDTH, gap / MAX_GAP, pos / MAX_WIDTH, diff / MAX_HEIGHT]
class Platform:
''' Represents a fixed platform. '''
def __init__(self, xpos, ypos, width):
self.position = vector(xpos, ypos)
self.size = vector(width, PLATHEIGHT)
class Simulator:
''' This class represents the environment. '''
def __init__(self):
''' The entities are set up and added to a space. '''
self.xpos = 0.0
self.player = Player()
self.platform1 = Platform(0.0, HEIGHT1, WIDTH1)
self.platform2 = Platform(GAP1 + self.platform1.size[0], HEIGHT2, WIDTH2)
self.platform3 = Platform(self.platform2.position[0] +
GAP2 + self.platform2.size[0], HEIGHT3, WIDTH3)
self.enemy1 = Enemy(self.platform1)
self.enemy2 = Enemy(self.platform2)
self.states = []
def get_state(self):
''' Returns the representation of the current state. '''
if self.player.position[0] > self.platform2.position[0]:
enemy = self.enemy2
else:
enemy = self.enemy1
state = np.array([
self.player.position[0], #0
self.player.velocity[0], #1
enemy.position[0], #2
enemy.dx]) #3
return state
def on_platforms(self):
''' Checks if the player is on any of the platforms. '''
for platform in [self.platform1, self.platform2, self.platform3]:
if self.player.on_platform(platform):
return True
return False
def perform_action(self, action, dt=DT):
''' Applies for selected action for the given agent. '''
if self.on_platforms():
if action:
act, parameters = action
if act == 'jump':
self.player.jump(parameters)
elif act == 'run':
self.player.run(parameters, dt)
elif act == 'leap':
self.player.leap_to(parameters)
elif act == 'hop':
self.player.hop_to(parameters)
else:
self.player.fall()
def lower_bound(self):
''' Returns the lowest height of the platforms. '''
lower = min(self.platform1.position[1], self.platform2.position[1], self.platform3.position[1])
return lower
def right_bound(self):
''' Returns the edge of the game. '''
return self.platform3.position[0] + self.platform3.size[0]
def terminal_check(self, reward=0.0):
''' Determines if the episode is ended, and the reward. '''
end_episode = self.player.position[1] < self.lower_bound() + PLATHEIGHT
right = self.player.position[0] >= self.right_bound()
for entity in [self.enemy1, self.enemy2]:
if self.player.colliding(entity):
end_episode = True
if right:
reward = (self.right_bound() - self.xpos) / self.right_bound()
end_episode = True
return reward, end_episode
def update(self, action, dt=DT, interface = False):
''' Performs a single transition with the given action,
then returns the new state and a reward. '''
if interface:
self.xpos = self.player.position[0]
self.states.append([self.player.position.copy(),
self.enemy1.position.copy(),
self.enemy2.position.copy()])
self.perform_action(action, dt)
if self.on_platforms():
self.player.ground_bound()
if self.player.position[0] > self.platform2.position[0]:
enemy = self.enemy2
else:
enemy = self.enemy1
for entity in [self.player, enemy]:
entity.update(dt)
for platform in [self.platform1, self.platform2, self.platform3]:
if self.player.colliding(platform):
self.player.decollide(platform)
self.player.velocity[0] = 0.0
reward = (self.player.position[0] - self.xpos) / self.right_bound()
return self.terminal_check(reward)
def take_action(self, action):
''' Take a full, stabilised update. '''
end_episode = False
run = True
act, params = action
self.xpos = self.player.position[0]
step = 0
difft = 1.0
while run:
if act == "run":
reward, end_episode = self.update(('run', abs(params)), DT)
difft -= DT
run = difft > 0
elif act in ['jump', 'hop', 'leap']:
reward, end_episode = self.update(action)
run = not self.on_platforms()
action = None
if end_episode:
run = False
step += 1
state = self.get_state()
return state, reward, end_episode, step
class Enemy:
''' Defines the enemy. '''
size = vector(20.0, 30.0)
def __init__(self, platform):
''' Initializes the enemy on the platform. '''
self.dx = -ENEMY_SPEED
self.platform = platform
self.position = self.platform.size + self.platform.position
self.position[0] -= self.size[0]
def update(self, dt):
''' Shift the enemy along the platform. '''
right = self.platform.position[0] + self.platform.size[0] - self.size[0]
if not self.platform.position[0] < self.position[0] < right:
self.dx *= -1
self.dx += np.random.normal(0.0, ENEMY_NOISE*dt)
self.dx = bound(self.dx, -ENEMY_SPEED, ENEMY_SPEED)
self.position[0] += self.dx * dt
self.position[0] = bound(self.position[0], self.platform.position[0], right)
class Player(Enemy):
''' Represents the player character. '''
decay = 0.99
def __init__(self):
''' Initialize the position to the starting platform. '''
self.position = vector(0, PLATHEIGHT)
self.velocity = vector(0.0, 0.0)
def update(self, dt):
''' Update the position and velocity. '''
self.position += self.velocity * dt
self.position[0] = bound(self.position[0], 0.0, MAX_WIDTH)
self.velocity[0] *= self.decay
def accelerate(self, accel, dt=DT):
''' Applies a power to the entity in direction theta. '''
accel = bound_vector(accel, MAX_DDX, MAX_DDY)
self.velocity += accel * dt
self.velocity[0] -= abs(np.random.normal(0.0, ENEMY_NOISE*dt))
self.velocity = bound_vector(self.velocity, MAX_DX, MAX_DY)
self.velocity[0] = max(self.velocity[0], 0.0)
def ground_bound(self):
''' Bound dx while on the ground. '''
self.velocity[0] = bound(self.velocity[0], 0.0, MAX_DX_ON)
def run(self, power, dt):
''' Run for a given power and time. '''
if dt > 0:
self.accelerate(vector(power / dt, 0.0), dt)
def jump(self, power):
''' Jump up for a single step. '''
self.accelerate(vector(0.0, power / DT))
def jump_to(self, diffx, dy0, dev):
''' Jump to a specific position. '''
time = 2.0 * dy0 / GRAVITY + 1.0
dx0 = diffx / time - self.velocity[0]
dx0 = bound(dx0, -MAX_DDX, MAX_DY - dy0)
if dev > 0:
noise = -abs(np.random.normal(0.0, dev, 2))
else:
noise = np.zeros((2,))
accel = vector(dx0, dy0) + noise
self.accelerate(accel / DT)
def hop_to(self, diffx):
''' Jump high to a position. '''
self.jump_to(diffx, 35.0, HOP_DEV)
def leap_to(self, diffx):
''' Jump over a gap. '''
self.jump_to(diffx, 25.0, LEAP_DEV)
def fall(self):
''' Apply gravity. '''
self.accelerate(vector(0.0, -GRAVITY))
def decollide(self, other):
''' Shift overlapping entities apart. '''
precorner = other.position - self.size
postcorner = other.position + other.size
newx, newy = self.position[0], self.position[1]
if self.position[0] < other.position[0]:
newx = precorner[0]
elif self.position[0] > postcorner[0] - self.size[0]:
newx = postcorner[0]
if self.position[1] < other.position[1]:
newy = precorner[1]
elif self.position[1] > postcorner[1] - self.size[1]:
newy = postcorner[1]
if newx == self.position[0]:
self.velocity[1] = 0.0
self.position[1] = newy
elif newy == self.position[1]:
self.velocity[0] = 0.0
self.position[0] = newx
elif abs(self.position[0] - newx) < abs(self.position[1] - newy):
self.velocity[0] = 0.0
self.position[0] = newx
else:
self.velocity[1] = 0.0
self.position[1] = newy
def above_platform(self, platform):
''' Checks the player is above the platform. '''
return -self.size[0] <= self.position[0] - platform.position[0] <= platform.size[0]
def on_platform(self, platform):
''' Checks the player is standing on the platform. '''
ony = self.position[1] - platform.position[1] == platform.size[1]
return self.above_platform(platform) and ony
def colliding(self, other):
''' Check if two entities are overlapping. '''
precorner = other.position - self.size
postcorner = other.position + other.size
collide = (precorner < self.position).all()
collide = collide and (self.position < postcorner).all()
return collide
SHIFT_VECTOR = np.array([Player.size[0], 0.0, 0.0,
ENEMY_SPEED])
SCALE_VECTOR = np.array([MAX_WIDTH + Player.size[0], MAX_DX,
MAX_WIDTH, 2*ENEMY_SPEED])
STATE_DIM = Simulator().get_state().size