582 lines
24 KiB
Python
582 lines
24 KiB
Python
import copy
|
||
import random
|
||
from robowaiter.behavior_tree.obtea.BehaviorTree import Leaf,ControlBT
|
||
|
||
|
||
|
||
class CondActPair:
|
||
def __init__(self, cond_leaf,act_leaf):
|
||
self.cond_leaf = cond_leaf
|
||
self.act_leaf = act_leaf
|
||
|
||
#定义行动类，行动包括前提、增加和删除影响
|
||
class Action:
|
||
def __init__(self,name='anonymous action',pre=set(),add=set(),del_set=set(),cost=1):
|
||
self.pre=copy.deepcopy(pre)
|
||
self.add=copy.deepcopy(add)
|
||
self.del_set=copy.deepcopy(del_set)
|
||
self.name=name
|
||
self.cost=cost
|
||
|
||
def __str__(self):
|
||
return self.name
|
||
# 从状态随机生成一个行动
|
||
def generate_from_state(self,state,num):
|
||
for i in range(0,num):
|
||
if i in state:
|
||
if random.random() >0.5:
|
||
self.pre.add(i)
|
||
if random.random() >0.5:
|
||
self.del_set.add(i)
|
||
continue
|
||
if random.random() > 0.5:
|
||
self.add.add(i)
|
||
continue
|
||
if random.random() >0.5:
|
||
self.del_set.add(i)
|
||
def print_action(self):
|
||
print (self.pre)
|
||
print(self.add)
|
||
print(self.del_set)
|
||
|
||
#生成随机状态
|
||
def generate_random_state(num):
|
||
result = set()
|
||
for i in range(0,num):
|
||
if random.random()>0.5:
|
||
result.add(i)
|
||
return result
|
||
#从状态和行动生成后继状态
|
||
def state_transition(state,action):
|
||
if not action.pre <= state:
|
||
print ('error: action not applicable')
|
||
return state
|
||
new_state=(state | action.add) - action.del_set
|
||
return new_state
|
||
|
||
|
||
def conflict(c):
|
||
have_at = False
|
||
have_holding = False
|
||
for str in c:
|
||
if 'At' in str:
|
||
if not have_at:
|
||
have_at = True
|
||
else:
|
||
return True
|
||
|
||
if 'Holding' in str:
|
||
if not have_holding:
|
||
have_holding = True
|
||
else:
|
||
return True
|
||
return False
|
||
|
||
import heapq
|
||
#本文所提出的完备规划算法
|
||
class OptBTExpAlgorithm:
|
||
def __init__(self,verbose=False,lit_act_dic=None):
|
||
self.bt = None
|
||
self.nodes=[]
|
||
self.traversed=[]
|
||
self.mounted=[]
|
||
self.conditions=[]
|
||
self.conditions_index=[]
|
||
self.verbose=verbose
|
||
self.goal=None
|
||
self.bt_merge = True
|
||
self.lit_act_dic = lit_act_dic
|
||
|
||
def clear(self):
|
||
self.bt = None
|
||
self.nodes = []
|
||
self.traversed = [] #存cond
|
||
self.expanded = [] #存整个
|
||
self.conditions = []
|
||
self.conditions_index = []
|
||
self.traversed_state_num=0
|
||
self.fot_times = 0
|
||
self.expand_conds=0
|
||
self.tree_size=0
|
||
self.bt_without_merge = None
|
||
self.subtree_count=1
|
||
|
||
#运行规划算法，从初始状态、目标状态和可用行动，计算行为树self.bt
|
||
# def run_algorithm(self,goal,actions,scene):
|
||
#运行规划算法，从初始状态、目标状态和可用行动，计算行为树self.bt
|
||
# def run_algorithm(self,goal,actions,scene):
|
||
def run_algorithm_selTree(self, start, goal, actions,merge_time=99999999):
|
||
|
||
|
||
self.traversed_state_num=0
|
||
|
||
self.goal = goal
|
||
if self.verbose:
|
||
print("\n算法开始！")
|
||
bt = ControlBT(type='cond')
|
||
# 初始行为树只包含目标条件
|
||
gc_node = Leaf(type='cond', content=goal, mincost=0) # 为了统一，都成对出现
|
||
ga_node = Leaf(type='act', content=None, mincost=0)
|
||
subtree = ControlBT(type='?')
|
||
subtree.add_child([gc_node]) # 子树首先保留所扩展结
|
||
self.expand_conds+=1
|
||
bt.add_child([subtree])
|
||
cond_anc_pair = CondActPair(cond_leaf=gc_node,act_leaf=ga_node)
|
||
|
||
heapq.heappush(self.nodes, cond_anc_pair)
|
||
self.expanded.append(goal)
|
||
self.traversed_state_num += 1
|
||
|
||
self.traversed = [goal] # the set of expanded conditions
|
||
min_cost = float('inf')
|
||
|
||
# if goal <= start:
|
||
# self.bt_without_merge = bt
|
||
# return bt, 0
|
||
|
||
while len(self.nodes)!=0:
|
||
|
||
self.fot_times+=1
|
||
|
||
# Find the condition for the shortest cost path
|
||
# ======================== Next Goal ============================ #
|
||
min_cost = float ('inf')
|
||
pair_node = heapq.heappop(self.nodes)
|
||
|
||
if self.verbose:
|
||
print("选择扩展条件结点：",pair_node.cond_leaf.content)
|
||
# Update self.nodes and self.traversed
|
||
c = pair_node.cond_leaf.content # 子树所扩展结点对应的条件（一个文字的set）
|
||
# Mount the action node and extend BT. T = Eapand(T,c,A(c))
|
||
|
||
|
||
|
||
if c!=goal:
|
||
if c!=set():
|
||
sequence_structure = ControlBT(type='>')
|
||
sequence_structure.add_child(
|
||
[pair_node.cond_leaf, pair_node.act_leaf])
|
||
subtree.add_child([copy.deepcopy(sequence_structure)]) # subtree 是回不断变化的，它的父亲是self.bt
|
||
# self.expanded.append(copy.deepcopy(pair_node))
|
||
# self.expanded.append(pair_node.cond_leaf.content)
|
||
#
|
||
# if c <= start:
|
||
# if self.bt_merge:
|
||
# # bt = self.merge_adjacent_conditions_stack(bt)
|
||
# bt = self.merge_adjacent_conditions_stack_time(bt,merge_time=merge_time)
|
||
# return bt, min_cost
|
||
else:
|
||
subtree.add_child([copy.deepcopy(pair_node.act_leaf)])
|
||
self.expand_conds += 1
|
||
self.expanded.append(c)
|
||
if c <= start:
|
||
self.tree_size = self.bfs_cal_tree_size_subtree(bt)
|
||
self.bt_without_merge = bt
|
||
if self.bt_merge:
|
||
# bt = self.merge_adjacent_conditions_stack(bt)
|
||
bt = self.merge_adjacent_conditions_stack_time(bt,merge_time=merge_time)
|
||
return bt, min_cost
|
||
|
||
|
||
|
||
if self.verbose:
|
||
print("完成扩展 a_node= %s,对应的新条件 c_attr= %s,mincost=%d" \
|
||
% (cond_anc_pair.act_leaf.content.name, cond_anc_pair.cond_leaf.content,
|
||
cond_anc_pair.cond_leaf.mincost))
|
||
|
||
if self.verbose:
|
||
print("遍历所有动作, 寻找符合条件的动作")
|
||
# 遍历所有动作, 寻找符合条件的动作
|
||
current_mincost = pair_node.cond_leaf.mincost # 当前的最短路径是多少
|
||
# ======================== End Next Goal ============================ #
|
||
|
||
# ====================== Action Trasvers ============================ #
|
||
traversed_current = []
|
||
|
||
# act_tmp_set = set()
|
||
# for lit in c:
|
||
# act_tmp_set |= self.lit_act_dic[lit]
|
||
|
||
# for i in act_tmp_set:
|
||
for i in range(0, len(actions)):
|
||
|
||
# if c=={'RobotNear(Chips)', 'Holding(Nothing)'} and actions[i].name=='Clean(Chairs)0':
|
||
# xx=1
|
||
if not c & ((actions[i].pre | actions[i].add) - actions[i].del_set) <= set() :
|
||
if (c - actions[i].del_set) == c:
|
||
if self.verbose:
|
||
print("———— 满足条件可以扩展")
|
||
c_attr = (actions[i].pre | c) - actions[i].add
|
||
|
||
if conflict(c_attr):
|
||
if self.verbose:
|
||
print("———— 冲突：动作 %s， 条件 %s"% (actions[i].name,c_attr))
|
||
continue
|
||
|
||
# 剪枝操作,现在的条件是以前扩展过的条件的超集
|
||
valid = True
|
||
|
||
for j in self.expanded:
|
||
if j <= c_attr:
|
||
valid = False
|
||
break
|
||
|
||
# tmp_heap = list(self.nodes)
|
||
# # tmp_heap = copy.deepcopy(self.nodes)
|
||
# while tmp_heap: # 剪枝操作 self.expanded?
|
||
# cond_anc_pair = heapq.heappop(tmp_heap)
|
||
# j = cond_anc_pair.cond_leaf.content
|
||
# if j <= c_attr:
|
||
# if cond_anc_pair.cond_leaf.mincost < current_mincost + actions[i].cost:
|
||
# valid = False
|
||
# break
|
||
|
||
|
||
if valid:
|
||
# c_attr_string = "".join(sorted(list(c_attr)))
|
||
c_attr_node = Leaf(type='cond', content=c_attr, mincost=current_mincost + actions[i].cost)
|
||
a_attr_node = Leaf(type='act', content=actions[i], mincost=current_mincost + actions[i].cost)
|
||
cond_anc_pair = CondActPair(cond_leaf=c_attr_node, act_leaf=a_attr_node)
|
||
# heapq.heappush(self.nodes, copy.deepcopy(cond_anc_pair))
|
||
heapq.heappush(self.nodes, cond_anc_pair)
|
||
|
||
|
||
self.traversed_state_num+=1
|
||
traversed_current.append(c_attr)
|
||
# 把符合条件的动作节点都放到列表里
|
||
if self.verbose:
|
||
print("———— -- %s 符合条件放入列表,对应的c为 %s" % (actions[i].name,c_attr))
|
||
|
||
# print(len(traversed_current))
|
||
self.traversed.extend(traversed_current)
|
||
# ====================== End Action Trasvers ============================ #
|
||
# self.tree_size = self.bfs_cal_tree_size_subtree(bt)
|
||
self.bt_without_merge = bt
|
||
if self.bt_merge:
|
||
# bt = self.merge_adjacent_conditions_stack(bt)
|
||
bt = self.merge_adjacent_conditions_stack_time(bt,merge_time=merge_time)
|
||
if self.verbose:
|
||
print("算法结束！\n")
|
||
return bt,min_cost
|
||
|
||
def run_algorithm(self, start, goal, actions):
|
||
self.bt = ControlBT(type='cond')
|
||
subtree = ControlBT(type='?')
|
||
|
||
subtree_with_costs_ls=[]
|
||
|
||
if len(goal) > 1:
|
||
for g in goal:
|
||
bt_sel_tree,mincost = self.run_algorithm_selTree(start, g, actions)
|
||
subtree_with_costs_ls.append((bt_sel_tree,mincost))
|
||
# 要排个序再一次add
|
||
# subtree.add_child([copy.deepcopy(bt_sel_tree.children[0])])
|
||
# self.bt.add_child([subtree])
|
||
sorted_trees = sorted(subtree_with_costs_ls, key=lambda x: x[1])
|
||
for tree,cost in sorted_trees:
|
||
subtree.add_child([copy.deepcopy(tree.children[0])])
|
||
self.bt.add_child([subtree])
|
||
else:
|
||
self.bt,mincost = self.run_algorithm_selTree(start, goal[0], actions)
|
||
return True
|
||
|
||
|
||
|
||
def merge_adjacent_conditions_stack_time(self,bt_sel,merge_time=9999999):
|
||
|
||
merge_time = min(merge_time,500)
|
||
|
||
# 只针对第一层合并，之后要考虑层层递归合并
|
||
bt = ControlBT(type='cond')
|
||
sbtree = ControlBT(type='?')
|
||
# gc_node = Leaf(type='cond', content=self.goal, mincost=0) # 为了统一，都成对出现
|
||
# sbtree.add_child([copy.deepcopy(gc_node)]) # 子树首先保留所扩展结
|
||
bt.add_child([sbtree])
|
||
|
||
parnode = bt_sel.children[0]
|
||
stack=[]
|
||
time_stack=[]
|
||
for child in parnode.children:
|
||
if isinstance(child, ControlBT) and child.type == '>':
|
||
if stack==[]:
|
||
stack.append(child)
|
||
time_stack.append(0)
|
||
continue
|
||
# 检查合并的条件，前面一个的条件包含了后面的条件，把包含部分提取出来
|
||
last_child = stack[-1]
|
||
last_time = time_stack[-1]
|
||
|
||
|
||
|
||
if last_time<merge_time and isinstance(last_child, ControlBT) and last_child.type == '>':
|
||
set1 = last_child.children[0].content
|
||
set2 = child.children[0].content
|
||
inter = set1 & set2
|
||
|
||
# print("merge time:", last_time,set1,set2)
|
||
|
||
if inter!=set():
|
||
c1 = set1-set2
|
||
c2 = set2-set1
|
||
inter_node = Leaf(type='cond', content=inter)
|
||
c1_node = Leaf(type='cond', content=c1)
|
||
c2_node = Leaf(type='cond', content=c2)
|
||
a1_node = last_child.children[1]
|
||
a2_node = child.children[1]
|
||
|
||
|
||
# set1<=set2,此时set2对应的动作永远不会执行
|
||
if (c1==set() and isinstance(last_child.children[1], Leaf) and isinstance(child.children[1], Leaf) \
|
||
and isinstance(last_child.children[1].content, Action) and isinstance(child.children[1].content, Action)):
|
||
continue
|
||
|
||
# 再写一个特殊情况处理，三个结点动作last 遇到 两个结点 且动作相同
|
||
if len(last_child.children)==3 and \
|
||
isinstance(last_child.children[2], Leaf) and isinstance(child.children[1], Leaf) \
|
||
and isinstance(last_child.children[2].content, Action) and isinstance( child.children[1].content, Action) \
|
||
and last_child.children[2].content.name == child.children[1].content.name \
|
||
and c1==set() and c2!=set():
|
||
last_child.children[1].add_child([c2_node])
|
||
continue
|
||
elif len(last_child.children)==3:
|
||
stack.append(child)
|
||
time_stack.append(0)
|
||
continue
|
||
|
||
# 判断动作相不相同
|
||
if isinstance(last_child.children[1], Leaf) and isinstance(child.children[1], Leaf) \
|
||
and isinstance(last_child.children[1].content, Action) and isinstance(child.children[1].content, Action) \
|
||
and last_child.children[1].content.name == child.children[1].content.name:
|
||
|
||
if c2==set():
|
||
tmp_tree = ControlBT(type='>')
|
||
tmp_tree.add_child(
|
||
[inter_node, a1_node])
|
||
else:
|
||
_sel = ControlBT(type='?')
|
||
_sel.add_child([c1_node, c2_node])
|
||
tmp_tree = ControlBT(type='>')
|
||
tmp_tree.add_child(
|
||
[inter_node, _sel,a1_node])
|
||
else:
|
||
if c1 == set():
|
||
seq1 = last_child.children[1]
|
||
else:
|
||
seq1 = ControlBT(type='>')
|
||
seq1.add_child([c1_node, a1_node])
|
||
|
||
if c2 == set():
|
||
seq2 = child.children[1]
|
||
else:
|
||
seq2 = ControlBT(type='>')
|
||
seq2.add_child([c2_node, a2_node])
|
||
sel = ControlBT(type='?')
|
||
sel.add_child([seq1, seq2])
|
||
tmp_tree = ControlBT(type='>')
|
||
tmp_tree.add_child(
|
||
[inter_node,sel])
|
||
|
||
stack.pop()
|
||
time_stack.pop()
|
||
stack.append(tmp_tree)
|
||
time_stack.append(last_time+1)
|
||
|
||
else:
|
||
stack.append(child)
|
||
time_stack.append(0)
|
||
else:
|
||
stack.append(child)
|
||
time_stack.append(0)
|
||
else:
|
||
stack.append(child)
|
||
time_stack.append(0)
|
||
|
||
for tree in stack:
|
||
sbtree.add_child([tree])
|
||
bt_sel = bt
|
||
return bt_sel
|
||
|
||
|
||
|
||
def merge_adjacent_conditions_stack(self,bt_sel):
|
||
# 只针对第一层合并，之后要考虑层层递归合并
|
||
bt = ControlBT(type='cond')
|
||
sbtree = ControlBT(type='?')
|
||
# gc_node = Leaf(type='cond', content=self.goal, mincost=0) # 为了统一，都成对出现
|
||
# sbtree.add_child([copy.deepcopy(gc_node)]) # 子树首先保留所扩展结
|
||
bt.add_child([sbtree])
|
||
|
||
parnode = copy.deepcopy(bt_sel.children[0])
|
||
stack=[]
|
||
for child in parnode.children:
|
||
if isinstance(child, ControlBT) and child.type == '>':
|
||
if stack==[]:
|
||
stack.append(child)
|
||
continue
|
||
# 检查合并的条件，前面一个的条件包含了后面的条件，把包含部分提取出来
|
||
last_child = stack[-1]
|
||
if isinstance(last_child, ControlBT) and last_child.type == '>':
|
||
set1 = last_child.children[0].content
|
||
set2 = child.children[0].content
|
||
inter = set1 & set2
|
||
if inter!=set():
|
||
c1 = set1-set2
|
||
c2 = set2-set1
|
||
inter_node = Leaf(type='cond', content=inter)
|
||
c1_node = Leaf(type='cond', content=c1)
|
||
c2_node = Leaf(type='cond', content=c2)
|
||
a1_node = copy.deepcopy(last_child.children[1])
|
||
a2_node = copy.deepcopy(child.children[1])
|
||
|
||
|
||
# set1<=set2,此时set2对应的动作永远不会执行
|
||
if (c1==set() and isinstance(last_child.children[1], Leaf) and isinstance(child.children[1], Leaf) \
|
||
and isinstance(last_child.children[1].content, Action) and isinstance(child.children[1].content, Action)):
|
||
continue
|
||
|
||
# 再写一个特殊情况处理，三个结点动作last 遇到 两个结点 且动作相同
|
||
if len(last_child.children)==3 and \
|
||
isinstance(last_child.children[2], Leaf) and isinstance(child.children[1], Leaf) \
|
||
and isinstance(last_child.children[2].content, Action) and isinstance( child.children[1].content, Action) \
|
||
and last_child.children[2].content.name == child.children[1].content.name \
|
||
and c1==set() and c2!=set():
|
||
last_child.children[1].add_child([copy.deepcopy(c2_node)])
|
||
continue
|
||
elif len(last_child.children)==3:
|
||
stack.append(child)
|
||
continue
|
||
|
||
# 判断动作相不相同
|
||
if isinstance(last_child.children[1], Leaf) and isinstance(child.children[1], Leaf) \
|
||
and isinstance(last_child.children[1].content, Action) and isinstance(child.children[1].content, Action) \
|
||
and last_child.children[1].content.name == child.children[1].content.name:
|
||
|
||
if c2==set():
|
||
tmp_tree = ControlBT(type='>')
|
||
tmp_tree.add_child(
|
||
[copy.deepcopy(inter_node), copy.deepcopy(a1_node)])
|
||
else:
|
||
_sel = ControlBT(type='?')
|
||
_sel.add_child([copy.deepcopy(c1_node), copy.deepcopy(c2_node)])
|
||
tmp_tree = ControlBT(type='>')
|
||
tmp_tree.add_child(
|
||
[copy.deepcopy(inter_node), copy.deepcopy(_sel),copy.deepcopy(a1_node)])
|
||
else:
|
||
if c1 == set():
|
||
seq1 = copy.deepcopy(last_child.children[1])
|
||
else:
|
||
seq1 = ControlBT(type='>')
|
||
seq1.add_child([copy.deepcopy(c1_node), copy.deepcopy(a1_node)])
|
||
|
||
if c2 == set():
|
||
seq2 = copy.deepcopy(child.children[1])
|
||
else:
|
||
seq2 = ControlBT(type='>')
|
||
seq2.add_child([copy.deepcopy(c2_node), copy.deepcopy(a2_node)])
|
||
sel = ControlBT(type='?')
|
||
sel.add_child([copy.deepcopy(seq1), copy.deepcopy(seq2)])
|
||
tmp_tree = ControlBT(type='>')
|
||
tmp_tree.add_child(
|
||
[copy.deepcopy(inter_node), copy.deepcopy(sel)])
|
||
|
||
stack.pop()
|
||
stack.append(tmp_tree)
|
||
|
||
else:
|
||
stack.append(child)
|
||
else:
|
||
stack.append(child)
|
||
else:
|
||
stack.append(child)
|
||
|
||
for tree in stack:
|
||
sbtree.add_child([tree])
|
||
bt_sel = copy.deepcopy(bt)
|
||
return bt_sel
|
||
|
||
|
||
|
||
def print_solution(self):
|
||
print("========= BT ==========") # 树的bfs遍历
|
||
nodes_ls = []
|
||
nodes_ls.append(self.bt)
|
||
while len(nodes_ls) != 0:
|
||
parnode = nodes_ls[0]
|
||
print("Parrent:", parnode.type)
|
||
for child in parnode.children:
|
||
if isinstance(child, Leaf):
|
||
print("---- Leaf:", child.content)
|
||
elif isinstance(child, ControlBT):
|
||
print("---- ControlBT:", child.type)
|
||
nodes_ls.append(child)
|
||
print()
|
||
nodes_ls.pop(0)
|
||
print("========= BT ==========\n")
|
||
|
||
# 返回所有能到达目标状态的初始状态
|
||
def get_all_state_leafs(self):
|
||
state_leafs=[]
|
||
|
||
nodes_ls = []
|
||
nodes_ls.append(self.bt)
|
||
while len(nodes_ls) != 0:
|
||
parnode = nodes_ls[0]
|
||
for child in parnode.children:
|
||
if isinstance(child, Leaf):
|
||
if child.type == "cond":
|
||
state_leafs.append(child.content)
|
||
elif isinstance(child, ControlBT):
|
||
nodes_ls.append(child)
|
||
nodes_ls.pop(0)
|
||
|
||
return state_leafs
|
||
|
||
|
||
# 树的dfs
|
||
def dfs_ptml(self,parnode,is_root=False):
|
||
for child in parnode.children:
|
||
if isinstance(child, Leaf):
|
||
if child.type == 'cond':
|
||
|
||
if is_root and len(child.content) > 1:
|
||
# 把多个 cond 串起来
|
||
self.ptml_string += "sequence{\n"
|
||
self.ptml_string += "cond "
|
||
c_set_str = '\n cond '.join(map(str, child.content)) + "\n"
|
||
self.ptml_string += c_set_str
|
||
self.ptml_string += '}\n'
|
||
else:
|
||
self.ptml_string += "cond "
|
||
c_set_str = '\n cond '.join(map(str, child.content)) + "\n"
|
||
self.ptml_string += c_set_str
|
||
|
||
elif child.type == 'act':
|
||
if '(' not in child.content.name:
|
||
self.ptml_string += 'act ' + child.content.name + "()\n"
|
||
else:
|
||
self.ptml_string += 'act ' + child.content.name + "\n"
|
||
elif isinstance(child, ControlBT):
|
||
if child.type == '?':
|
||
self.ptml_string += "selector{\n"
|
||
self.dfs_ptml(parnode=child)
|
||
elif child.type == '>':
|
||
self.ptml_string += "sequence{\n"
|
||
self.dfs_ptml( parnode=child)
|
||
self.ptml_string += '}\n'
|
||
|
||
|
||
def get_ptml(self):
|
||
self.ptml_string = "selector{\n"
|
||
self.dfs_ptml(self.bt.children[0],is_root=True)
|
||
self.ptml_string += '}\n'
|
||
return self.ptml_string
|
||
|
||
|
||
def save_ptml_file(self,file_name):
|
||
self.ptml_string = "selector{\n"
|
||
self.dfs_ptml(self.bt.children[0])
|
||
self.ptml_string += '}\n'
|
||
with open(f'./{file_name}.ptml', 'w') as file:
|
||
file.write(self.ptml_string)
|
||
return self.ptml_string
|