-
Notifications
You must be signed in to change notification settings - Fork 0
/
yaw_planner.py
255 lines (194 loc) · 9.5 KB
/
yaw_planner.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
import numpy as np
import math
import matplotlib.pyplot as plt
import random
from utils import Drone2D
from math import cos, sin, radians, atan2, degrees
from numpy.linalg import norm
from utils import *
class NoControl(object):
def __init__(self, params):
self.params = params
def plan(self, state):
# self.v_yaw_space = np.arange(-self.params.drone_max_yaw_speed, self.params.drone_max_yaw_speed, self.params.drone_max_yaw_speed/3)
return 0
class LookAhead(object):
"""Make the drone look at the direction of its velocity
Args:
object (_type_): _description_
"""
def __init__(self, params):
self.dt = params.dt
self.params = params
def plan(self, state):
if state['drone'].velocity[1]==0 and state['drone'].velocity[0]==0:
return 0
target_yaw = math.degrees(math.atan2(-state['drone'].velocity[1], state['drone'].velocity[0])) % 360
# print(target_yaw)
if abs(target_yaw - state['drone'].yaw) < 180:
yaw_vel = max(min((target_yaw - state['drone'].yaw) / self.dt, self.params.drone_max_yaw_speed), -self.params.drone_max_yaw_speed)
else:
yaw_vel = -max(min((target_yaw - state['drone'].yaw) / self.dt, self.params.drone_max_yaw_speed), -self.params.drone_max_yaw_speed)
return yaw_vel / self.params.drone_max_yaw_speed
class Oxford(object):
"""Oxford method to plan gaze
Args:
object (_type_): _description_
"""
def __init__(self, params):
self.params = params
self.last_time_observed_map = 5 * np.ones((params.map_size[0]//params.map_scale, params.map_size[1]//params.map_scale))
self.swep_map = np.zeros((params.map_size[0]//params.map_scale, params.map_size[1]//params.map_scale))
self.dim = params.map_size
self.dt = params.dt
# Farthest step in trajectory that considered as priority
self.tau_s = 3
# Safe last time observed
self.tau_c = 0.5
self.c1 = 1000000
self.c2 = 1000
self.c3 = 1
# Primitive time step
self.v_yaw_space = np.arange(-self.params.drone_max_yaw_speed, self.params.drone_max_yaw_speed, self.params.drone_max_yaw_speed/3)
def get_view_map(self, drone):
dim = self.dim
x = np.arange(int(dim[0]//self.params.map_scale)).reshape(-1, 1) * self.params.map_scale
y = np.arange(int(dim[1]//self.params.map_scale)).reshape(1, -1) * self.params.map_scale
vec_yaw = np.array([math.cos(math.radians(drone.yaw)), -math.sin(math.radians(drone.yaw))])
view_angle = math.radians(drone.yaw_range / 2)
#((drone.x - x)**2 + (drone.y - y)**2 <= drone.yaw_depth ** 2) and
# math.acos(np.array([math.cos(math.radians(drone.yaw)), -math.sin(math.radians(drone.yaw))]).dot(np.array([x - drone.x, y - drone.y]))/np.norm(np.array([x - drone.x, y - drone.y]))) <= math.radians(drone.yaw_range / 2)
np.seterr(divide='ignore', invalid='ignore')
view_map = np.where(np.logical_or((drone.x - x)**2 + (drone.y - y)**2 <= 0, np.logical_and(np.arccos(((x - drone.x)*vec_yaw[0] + (y - drone.y)*vec_yaw[1]) / np.sqrt((drone.x - x)**2 + (drone.y - y)**2)) <= view_angle, ((drone.x - x)**2 + (drone.y - y)**2 <= drone.yaw_depth ** 2))), 1, 0)
return view_map
def plan(self, observation):
drone = observation['drone']
trajectory = observation['trajectory']
# update v_i
self.swep_map = np.zeros_like(self.swep_map)
for i, pos in enumerate(trajectory.positions):
self.swep_map[int(pos[0]//self.params.map_scale), int(pos[1]//self.params.map_scale)] = i * self.dt
# update t_i
view_map = self.get_view_map(self, drone)
self.last_time_observed_map = np.where(view_map,
0,
self.last_time_observed_map + (1 - view_map) * self.dt)
# plt.subplot(1,2,1)
# plt.imshow(self.last_time_observed_map.T)
# plt.subplot(1,2,2)
# plt.imshow(self.swep_map.T)
# plt.show()
# plt.pause(0.001)
# plt.clf()
# calculate reward
reward_map = np.where((self.swep_map > 0) & (self.swep_map <= self.tau_s) & (self.last_time_observed_map >= self.tau_c), self.c1,
np.where((self.swep_map > self.tau_s) & (self.last_time_observed_map >= self.tau_c), self.c2,
np.clip(self.c3*self.last_time_observed_map, -np.inf, 1)))
# calculate primitive for yaw control
target_yaw = drone.yaw + self.v_yaw_space * self.dt
max_reward = 0
best_action = 0
if len(trajectory)==0:
return 0
for i, yaw in enumerate(target_yaw):
new_drone = Drone2D(trajectory.positions[0][0], trajectory.positions[0][1], yaw, self.dt, self.params)
new_view_map = self.get_view_map(self, new_drone)
if max_reward < np.sum(new_view_map*reward_map):
best_action = i
max_reward = np.sum(new_view_map*reward_map)
return self.v_yaw_space[best_action] / self.params.drone_max_yaw_speed
# plt.imshow(reward_map)
# plt.show()
# plt.pause(0.1)
# plt.clf()
class Rotating(object):
def __init__(self, params):
self.params = params
def plan(self, observation):
return 1
def angle_between(angle1, angle2):
angle1 = angle1 % 360
angle2 = angle2 % 360
diff = abs(angle1 - angle2)
diff = np.minimum(diff, 360 - diff)
return diff
class Owl(object):
def __init__(self, params):
self.params = params
self.dt = 0.8
self.u = []
# weights for different costs
self.lamb = np.array([0.2, 0.9, 1, 0.1, 0])
self.u_space = np.arange(-self.params.drone_max_yaw_speed, self.params.drone_max_yaw_speed, self.params.drone_max_yaw_speed/10)
self.theta_h = params.drone_view_range
self.l_hit = 0.4
self.l_miss = -0.05
self.beta = 1
self.U_list = np.zeros(36)
@classmethod
def G(self, theta):
if angle_between(theta, 0) <= self.theta_h / 2:
return 0
else:
return radians(angle_between(theta, self.theta_h / 2)) * radians(angle_between(theta, -self.theta_h / 2))
@classmethod
def update_U(self, drone, dt):
delta_p = drone.velocity * dt
for i, d_i in enumerate(np.arange(0, 360, 10)):
d_i_hat = np.array([cos(radians(d_i)), sin(radians(d_i))])
L_yt = - delta_p.dot(d_i_hat) / self.params.drone_view_depth
L_yt += self.l_hit if angle_between(d_i, -drone.yaw) < self.theta_h / 2 else self.l_miss
self.U_list[i] = max(min(self.U_list[i] + L_yt,1),0)
@classmethod
def U(self, theta):
idx = np.argmin(angle_between(np.arange(0, 360, 10), theta))
return self.U_list[idx]
def plan(self, observation):
if len(self.u) != 0:
u = self.u[-1]
self.u.pop()
return u / self.params.drone_max_yaw_speed
drone = observation['drone']
target = observation['target']
trackers = drone.trackers
self.update_U(drone, self.dt)
d_g = degrees(atan2(target[1]-drone.y, target[0]-drone.x))
d_v = degrees(atan2(*((drone.velocity / norm(drone.velocity))[::-1])))
d_o = [degrees(atan2(*((tracker.mu_upds[-1][:2,0] - np.array([drone.x, drone.y]))[::-1]))) for tracker in trackers if tracker.active is True]
yaws = -(drone.yaw + self.u_space * self.dt)
costs = np.ones_like(yaws)
f = np.zeros([yaws.shape[0], 5])
for i, yaw in enumerate(yaws):
f[i, 0] = self.G(yaw - d_g) * (1-self.U(d_g))
f[i, 1] = norm(drone.velocity/10)**2 * self.G(yaw - d_v)*(1-self.U(d_v))
for d_o_i, tracker in zip(d_o, trackers):
f[i, 2] += self.beta * norm(tracker.mu_upds[-1][2:,0]) / norm(tracker.mu_upds[-1][:2,0] - np.array([drone.x, drone.y])) * self.G(yaw - d_o_i)
f[i, 3] = self.U(yaw)
f[i ,4] = abs(radians(self.u_space[i] * self.dt))
costs[i] = np.sum(f[i, :].dot(self.lamb))
idx = np.argmin(costs)
# print(f[:,3])
for i in range(int(self.dt // self.params.dt) - 1):
self.u.append(self.u_space[idx])
return self.u_space[idx] / self.params.drone_max_yaw_speed
class LookGoal(object):
def __init__(self, params):
self.params = params
def plan(self, observation):
trajectory = observation['trajectory']
drone = observation['drone']
if len(trajectory) == 0:
return 0
x_look = trajectory.positions[-1][0]
y_look = trajectory.positions[-1][1]
for position in trajectory.positions:
if drone.map.get_grid(position[0], position[1])==grid_type['UNEXPLORED']:
x_look = position[0]
y_look = position[1]
break
target_yaw = math.degrees(math.atan2(-(y_look-drone.y), x_look-drone.x)) % 360
if abs(target_yaw - drone.yaw) < 180:
yaw_vel = max(min((target_yaw - drone.yaw) / self.params.dt, self.params.drone_max_yaw_speed), -self.params.drone_max_yaw_speed)
else:
yaw_vel = -max(min((target_yaw - drone.yaw) / self.params.dt, self.params.drone_max_yaw_speed), -self.params.drone_max_yaw_speed)
return yaw_vel / self.params.drone_max_yaw_speed