-
Notifications
You must be signed in to change notification settings - Fork 0
/
a_star.py
210 lines (171 loc) · 8.67 KB
/
a_star.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
from itertools import combinations, product
from copy import deepcopy
from time import time
from math import fabs
import pandas as pd
import random
INF = float("inf")
def print_obj_lst(lst):
print([str(obj) for obj in lst])
class AStar():
def __init__(self, env, agent_name, end_at_goal=True):
self.n_iter = 0
self.total_nodes = 0
self.agent_name = agent_name
self.start_state = env.agent_dict[agent_name]["start"]
self.goal_state = env.agent_dict[agent_name]["goal"]
self.end_at_goal = end_at_goal
self.get_neighbors = env.get_neighbors
self.check_goal = env.check_goal
self.CAT = env.current_solution_df.drop(agent_name, axis=1)
self.old_path = env.current_solution_df[agent_name]
self.closed_set, self.open_set = set(), set()
self.f_score, self.g_score, self.h_score, self.num_conflict, self.came_from = {}, {}, {}, {}, {}
self.step_cost = 1
def admissible_heuristic(self, state):
return fabs(state.location.x - self.goal_state.location.x) + fabs(state.location.y - self.goal_state.location.y)
def reconstruct_path(self, current):
path = [current]
while current in self.came_from:
current = self.came_from[current]
path.append(current)
return path[::-1]
def get_best_state(self):
f_benchmark = {state: self.f_score[state] for state in self.open_set}
min_f = min(f_benchmark.values())
min_f_states = [state for state in self.open_set if self.f_score[state] == min_f]
for state in min_f_states:
if self.end_at_goal:
if state.is_equal_except_time(self.goal_state):
#return True, state
conflict_t = self.check_goal(state)
if int(conflict_t) == -1:
return True, state
else:
current = self.wait(conflict_t, state)
return False, current
else:
if state.is_equal_except_time(self.goal_state):
if state.time == self.goal_state.time:
return True, state
else:
self.goal_state = self.old_path.iloc[-1]
self.end_at_goal = True
if len(min_f_states) > 1:
conflict_benchmark = {state: self.num_conflict[state] for state in min_f_states}
min_conflict = min(conflict_benchmark.values())
min_conflict_states = [state for state, n_conflict in conflict_benchmark.items() if n_conflict == min_conflict]
if len(min_conflict_states) > 1:
h_benchmark = {state: self.h_score[state] for state in min_conflict_states}
min_h = min(h_benchmark.values())
min_h_states = [state for state, h in h_benchmark.items() if h == min_h]
random.shuffle(min_h_states)
current = min_h_states[0]
if len(min_h_states) > 1 and self.old_path.shape[0] > current.time:
for state in min_h_states:
if state != self.old_path[state.time]:
current = state
break
else:
current = min_conflict_states[0]
else:
current = min_f_states[0]
return False, current
def wait(self, t, state):
conflicted_state = deepcopy(state)
total_time_diff = t-state.time
for time_diff in range(total_time_diff-1):
next_state = state.move(1)
self.came_from[next_state] = state
state = next_state
self.num_conflict[state] = self.num_conflict[conflicted_state]
self.g_score[state] = self.g_score[conflicted_state]
self.h_score[state] = 0
self.open_set.remove(conflicted_state)
self.closed_set.add(conflicted_state)
self.open_set.add(state)
return state
def count_conflicts(self, current_state, next_state):
n_vertex_conflicts = sum(self.CAT.loc[next_state.time] == next_state)
filt = list(map(current_state.is_equal_except_time, self.CAT.loc[next_state.time]))
n_edge_conflicts = sum(list(map(next_state.is_equal_except_time, self.CAT.loc[current_state.time, filt])))
return n_vertex_conflicts + n_edge_conflicts
def search(self, from_=None, to_=None, find_all_opt=False):
self.start_state = from_ if from_ else self.start_state
self.goal_state = to_ if to_ else self.goal_state
self.open_set.add(self.start_state)
self.num_conflict[self.start_state] = 0
self.g_score[self.start_state] = 0
self.h_score[self.start_state] = self.admissible_heuristic(self.start_state)
self.f_score[self.start_state] = self.h_score[self.start_state]
while self.open_set:
self.n_iter += 1
is_goal, current = self.get_best_state()
if is_goal:
if find_all_opt:
return self.opt_search(agent_name, f_score[current], min_f_states)
else:
return self.reconstruct_path(current)
if current.time > max(self.h_score[self.start_state] * 2, 100):
print("Too long path for", self.agent_name)
break
neighbor_list = self.get_neighbors(current, self.agent_name)
for neighbor in neighbor_list:
tentative_g_score = self.g_score.setdefault(current, INF) + self.step_cost
if neighbor in self.open_set and tentative_g_score >= self.g_score.setdefault(neighbor, INF):
continue
elif neighbor in self.closed_set:
if tentative_g_score >= self.g_score.setdefault(neighbor, INF):
continue
else:
self.closed_set.remove(neighbor)
self.came_from[neighbor] = current
self.g_score[neighbor] = tentative_g_score
self.h_score[neighbor] = self.admissible_heuristic(neighbor)
self.f_score[neighbor] = self.g_score[neighbor] + self.h_score[neighbor]
if self.CAT.shape[0] == 0:
self.num_conflict[neighbor] = 0
elif neighbor.time >= self.CAT.shape[0]:
self.num_conflict[neighbor] = self.num_conflict[current]
else:
self.num_conflict[neighbor] = self.count_conflicts(current, neighbor) + self.num_conflict[current]
self.open_set |= {neighbor}
self.total_nodes += 1
self.open_set.remove(current)
self.closed_set.add(current)
return False
def opt_search(self, agent_name, opt_cost, candidate_states):
goal_state = self.agent_dict[agent_name]["goal"]
closed_set = set()
open_set = candidate_states
opt_paths = []
while len(open_set) > 0:
copy_open = deepcopy(open_set)
for state in copy_open:
if state.is_equal_except_time(goal_state):
opt_paths += [self.reconstruct_path(state)]
open_set.remove(state)
if len(open_set) == 0:
break
current = open_set.pop()
closed_set.add(current)
neighbor_list = self.get_neighbors(current, agent_name)
for neighbor in neighbor_list:
tentative_g_score = self.g_score.setdefault(current, INF) + self.step_cost
if neighbor in open_set and tentative_g_score >= self.g_score.setdefault(neighbor, INF):
continue
elif neighbor in closed_set:
if tentative_g_score >= self.g_score.setdefault(neighbor, INF):
continue
else:
closed_set.remove(neighbor)
self.came_from[neighbor] = current
self.g_score[neighbor] = tentative_g_score
self.h_score[neighbor] = self.admissible_heuristic(neighbor)
self.f_score[neighbor] = self.g_score[neighbor] + self.h_score[neighbor]
if f_score[neighbor] == opt_cost:
open_set.add(neighbor)
self.total_nodes += 1
else:
closed_set.add(neighbor)
return opt_paths