346 lines
12 KiB
Python
346 lines
12 KiB
Python
|
import random
|
||
|
|
import numpy as np
|
||
|
|
import copy
|
||
|
|
import time
|
||
|
|
from OptimalBTExpansionAlgorithm import Action,generate_random_state,state_transition
|
||
|
|
from Examples import *
|
||
|
|
import os
|
||
|
|
output_path = os.path.join(os.path.dirname(__file__), "outputs")
|
||
|
|
from tools import print_action_data_table,BTTest_act_start_goal
|
||
|
|
|
||
|
|
|
||
|
|
def modify_set(s, obj):
|
||
|
|
return {f"{num}_{obj}" for num in s}
|
||
|
|
|
||
|
|
def get_act_start_goal(seed=1, max_copy_times=5,cond_pred=10, depth=10, act_pred=10, total_count=1000,obj_num=0):
|
||
|
|
max_copy_times=max_copy_times
|
||
|
|
|
||
|
|
literals_num_set = {i for i in range(cond_pred)}
|
||
|
|
all_obj_set = {i for i in range(obj_num)}
|
||
|
|
|
||
|
|
act_list = []
|
||
|
|
start_list = []
|
||
|
|
goal_list = []
|
||
|
|
|
||
|
|
total_literals_count=[]
|
||
|
|
total_states_num=[]
|
||
|
|
total_action_num = []
|
||
|
|
total_act_pred_num=[]
|
||
|
|
|
||
|
|
start_time_0=time.time()
|
||
|
|
for count in range(total_count):
|
||
|
|
|
||
|
|
actions_ls = []
|
||
|
|
literals_obj_set=set()
|
||
|
|
state_num = 0
|
||
|
|
action_num = 1
|
||
|
|
|
||
|
|
# 生成一个规划问题，包括随机的状态和行动，以及目标状态
|
||
|
|
states = []
|
||
|
|
actions = []
|
||
|
|
start = generate_random_state(cond_pred)
|
||
|
|
state = copy.deepcopy(start)
|
||
|
|
states.append(state)
|
||
|
|
obj_index = random.randint(0, obj_num-1) # 确定一个物体
|
||
|
|
|
||
|
|
aa_main = [] # 主路径上的动作谓词
|
||
|
|
|
||
|
|
for i in range(0, depth):
|
||
|
|
a = Action()
|
||
|
|
a.generate_from_state_local(state, literals_num_set,all_obj_set,obj_num,obj_index)
|
||
|
|
a.cost = random.randint(0, 100)
|
||
|
|
|
||
|
|
if not a in actions:
|
||
|
|
a.name = "a" + str(action_num)
|
||
|
|
action_num += 1
|
||
|
|
actions.append(a)
|
||
|
|
aa_main.append(a)
|
||
|
|
|
||
|
|
state = state_transition(state, a)
|
||
|
|
# 存入一个状态
|
||
|
|
if state in states:
|
||
|
|
pass
|
||
|
|
else:
|
||
|
|
states.append(state)
|
||
|
|
state_num+= a.vaild_num
|
||
|
|
|
||
|
|
|
||
|
|
goal = states[-1]
|
||
|
|
# state = copy.deepcopy(start)
|
||
|
|
for i in range (0,act_pred):
|
||
|
|
a = Action()
|
||
|
|
state = generate_random_state(cond_pred) # 随机选择一个状态
|
||
|
|
a.generate_from_state_local(state, literals_num_set, all_obj_set, obj_num)
|
||
|
|
a.cost = random.randint(0, 100)
|
||
|
|
if not a in actions:
|
||
|
|
a.name = "a"+str(action_num)
|
||
|
|
action_num+=1
|
||
|
|
actions.append(a)
|
||
|
|
|
||
|
|
# 存入一个状态
|
||
|
|
if state in states:
|
||
|
|
pass
|
||
|
|
else:
|
||
|
|
states.append(state)
|
||
|
|
state_num += a.vaild_num
|
||
|
|
state = state_transition(state,a)
|
||
|
|
# 存入一个状态
|
||
|
|
if state in states:
|
||
|
|
pass
|
||
|
|
else:
|
||
|
|
states.append(state)
|
||
|
|
state_num += a.vaild_num
|
||
|
|
|
||
|
|
for act in actions:
|
||
|
|
# if act.vild_args==set():
|
||
|
|
# print(a.name,a.pre,a.add,a.cost,a.vaild_num)
|
||
|
|
for obj in act.vild_args:
|
||
|
|
a = Action(
|
||
|
|
name=act.name + "_" + str(obj),
|
||
|
|
pre=modify_set(act.pre, obj),
|
||
|
|
add=modify_set(act.add, obj),
|
||
|
|
del_set=modify_set(act.del_set, obj),
|
||
|
|
cost=act.cost,
|
||
|
|
)
|
||
|
|
if a.pre != set():
|
||
|
|
literals_obj_set |= a.pre
|
||
|
|
if a.add != set():
|
||
|
|
literals_obj_set |= a.add
|
||
|
|
if a.del_set != set():
|
||
|
|
literals_obj_set |= a.del_set
|
||
|
|
actions_ls.append(a)
|
||
|
|
|
||
|
|
# actions_ls = [
|
||
|
|
# copy.deepcopy(act).update(
|
||
|
|
# name=act.name + "_" + str(obj),
|
||
|
|
# pre=modify_set(act.pre, obj),
|
||
|
|
# del_set=modify_set(act.del_set, obj),
|
||
|
|
# add=modify_set(act.add, obj)
|
||
|
|
# )
|
||
|
|
# for act in actions
|
||
|
|
# for obj in act.vild_args
|
||
|
|
# ]
|
||
|
|
|
||
|
|
if max_copy_times != 0:
|
||
|
|
for a in aa_main:
|
||
|
|
# if a.vild_args == set():
|
||
|
|
# print(a.name, a.pre, a.add, a.cost, a.vaild_num)
|
||
|
|
copy_times = random.randint(0, max_copy_times)
|
||
|
|
for ck in range(copy_times):
|
||
|
|
ca=copy.deepcopy(a)
|
||
|
|
ca.cost = random.randint(1, 100)
|
||
|
|
ca = ca.update(
|
||
|
|
name=ca.name + "_" + str(obj_index),
|
||
|
|
pre=modify_set(ca.pre, obj_index),
|
||
|
|
del_set=modify_set(ca.del_set, obj_index),
|
||
|
|
add=modify_set(ca.add, obj_index)
|
||
|
|
)
|
||
|
|
if ca.pre != set():
|
||
|
|
literals_obj_set |= ca.pre
|
||
|
|
if ca.add != set():
|
||
|
|
literals_obj_set |= ca.add
|
||
|
|
if ca.del_set != set():
|
||
|
|
literals_obj_set |= ca.del_set
|
||
|
|
actions_ls.append(ca)
|
||
|
|
|
||
|
|
# 计算所有 文字
|
||
|
|
|
||
|
|
|
||
|
|
|
||
|
|
start = modify_set(start, obj_index)
|
||
|
|
goal = modify_set(goal,obj_index)
|
||
|
|
|
||
|
|
|
||
|
|
act_list.append(actions_ls)
|
||
|
|
start_list.append(start)
|
||
|
|
goal_list.append(goal)
|
||
|
|
# print(literals_obj_set)
|
||
|
|
# print(len(literals_obj_set))
|
||
|
|
total_literals_count.append(len(literals_obj_set))
|
||
|
|
total_action_num.append(len(actions_ls))
|
||
|
|
total_act_pred_num.append(len(actions))
|
||
|
|
total_states_num.append(state_num)
|
||
|
|
|
||
|
|
end_time_0=time.time()
|
||
|
|
|
||
|
|
print("Total Time:", end_time_0-start_time_0)
|
||
|
|
# print("Total Time (start_to_goal):", total_time_dic["start_to_goal"])
|
||
|
|
# print("Total Time (random_act):",total_time_dic["random_act"])
|
||
|
|
print("Average Number of literals_obj", round(np.mean(total_literals_count), 3)) # 1000次问题的平均行动数
|
||
|
|
print("Average Number of States", round(np.mean(total_states_num), 3)) # 1000次问题的平均行动数
|
||
|
|
print("Average Number of Actions", round(np.mean(total_action_num),3)) # 1000次问题的平均行动数
|
||
|
|
print("Average Number of Act Pred", round(np.mean(total_act_pred_num), 3)) # 1000次问题的平均行动数
|
||
|
|
# print_action_data_table(goal, start, list(actions))
|
||
|
|
return act_list, start_list, goal_list, \
|
||
|
|
round(np.mean(total_literals_count), 3),\
|
||
|
|
round(np.mean(total_states_num),3),round(np.mean(total_action_num),3),round(np.mean(total_act_pred_num),3)
|
||
|
|
|
||
|
|
|
||
|
|
|
||
|
|
# # 设置生成规划问题集的超参数：文字数、解深度、迭代次数
|
||
|
|
seed = 1
|
||
|
|
random.seed(1)
|
||
|
|
np.random.seed(1)
|
||
|
|
|
||
|
|
# literals_num_ls=[10,10,10,100,100,10,10,10,100,100]
|
||
|
|
# depth_ls=[10,10,10,10,10,50,50,50,50,50]
|
||
|
|
# iters_ls=[10,100,1000,10,1000,10,100,1000,10,1000]
|
||
|
|
# literals_num_ls=[10,10,10,10, 100,100,100,100]
|
||
|
|
# depth_ls=[10,10,10,10, 10,10,10,10]
|
||
|
|
# iters_ls=[5,10,100,500, 5,10,100,500]
|
||
|
|
|
||
|
|
# literals_num_ls=[10,10,10,10,100,100,100,100] #+[100,100]
|
||
|
|
# depth_ls=[10,10,10,10,10,10,10,10] #+[10,10]
|
||
|
|
# iters_ls=[5,10,100,500,5,10,100,500] #+[int(500/2),int(500/4)]
|
||
|
|
# max_copy_times_ls=8*[5] #+[10,20]
|
||
|
|
|
||
|
|
# literals_num_ls=[10,10,10,10,100,100,100,100] + [100,100] #+[100,100]
|
||
|
|
# depth_ls=15 * [10]
|
||
|
|
# iters_ls=[5,10,100,300,5,10,100,300,175,105] #+[int(500/2),int(500/4)]
|
||
|
|
# max_copy_times_ls=8*[5] + [10,20] #+[10,20]
|
||
|
|
|
||
|
|
|
||
|
|
# obj_num_ls = [10,50,100,500]
|
||
|
|
# obj_num_ls = 3*[100]
|
||
|
|
# obj_num_ls = 4*[100]
|
||
|
|
# obj_num_ls=3*[100]
|
||
|
|
# obj_num_ls=[50]
|
||
|
|
|
||
|
|
|
||
|
|
# 随机动作数量
|
||
|
|
# iters_ls= 3*[100]
|
||
|
|
# iters_ls = 3* [10]
|
||
|
|
# iters_ls = [100,500,1000]
|
||
|
|
# iters_ls = 4*[100]
|
||
|
|
|
||
|
|
# depth_ls=[30,30,10] + 3*[30] +[50,50]
|
||
|
|
# obj_num_ls = [10,100] + 2*[50] + 4*[100]
|
||
|
|
# iters_ls=5*[100] + 3*[500]
|
||
|
|
# literals_num_ls= 7*[100] + [300]
|
||
|
|
# max_copy_times_ls=[0,0] + 6*[5]
|
||
|
|
|
||
|
|
# depth_ls=[30,30,10] + 3*[30] +[50]
|
||
|
|
# obj_num_ls = [10,100] + 2*[50] + 3*[100]
|
||
|
|
# iters_ls=5*[100] + 2*[500]
|
||
|
|
# literals_num_ls= 7*[100]
|
||
|
|
# max_copy_times_ls=[0,0] + 5*[5]
|
||
|
|
|
||
|
|
# depth_ls=[50]
|
||
|
|
# obj_num_ls = [100]
|
||
|
|
# iters_ls=[500]
|
||
|
|
# literals_num_ls= [300]
|
||
|
|
# max_copy_times_ls=[5]
|
||
|
|
|
||
|
|
|
||
|
|
|
||
|
|
# obj_num_ls = 9*[100,300,500] #9* [100,300,500]
|
||
|
|
# # Iterations Action Predicates
|
||
|
|
# act_pred_ls= 3*[0] + 3*[20] + 3*[40]
|
||
|
|
# act_pred_ls = 3*act_pred_ls
|
||
|
|
# #Condition Predicates
|
||
|
|
# cond_pred_ls= 9* [10] + 9* [30] + 9*[50]
|
||
|
|
|
||
|
|
obj_num_ls = [10] #9* [100,300,500]
|
||
|
|
# Iterations Action Predicates
|
||
|
|
act_pred_ls= [10]
|
||
|
|
act_pred_ls = 3*act_pred_ls
|
||
|
|
#Condition Predicates
|
||
|
|
cond_pred_ls= [10]
|
||
|
|
#
|
||
|
|
#
|
||
|
|
# max_copy_times_ls= [0,0] + 6*[5] #[5]*27
|
||
|
|
|
||
|
|
# obj_num_ls = [300]
|
||
|
|
# # Iterations Action Predicates
|
||
|
|
# act_pred_ls= [40]
|
||
|
|
# #Condition Predicates
|
||
|
|
# cond_pred_ls= [50]
|
||
|
|
|
||
|
|
max_copy_times_ls= 27*[0]
|
||
|
|
depth_ls=[10]*27
|
||
|
|
|
||
|
|
# obj_num_ls = [500]
|
||
|
|
# # Iterations Action Predicates
|
||
|
|
# act_pred_ls= [40]
|
||
|
|
# #Condition Predicates
|
||
|
|
# cond_pred_ls= [50]
|
||
|
|
# max_copy_times_ls= [5]
|
||
|
|
#
|
||
|
|
# depth_ls=[10]*8
|
||
|
|
|
||
|
|
|
||
|
|
# 文字数量 Cond Predicate
|
||
|
|
# literals_num_ls= 5*[10]
|
||
|
|
# literals_num_ls= [100,500,1000]
|
||
|
|
# literals_num_ls= 4*[100]
|
||
|
|
|
||
|
|
# depth_ls= 100 * [10]
|
||
|
|
|
||
|
|
# max_copy_times_ls=[5,5]
|
||
|
|
|
||
|
|
|
||
|
|
all_result=[]
|
||
|
|
|
||
|
|
for cond_pred,depth,act_pred,max_copy_times,obj_num in zip(cond_pred_ls, depth_ls, act_pred_ls,max_copy_times_ls,obj_num_ls):
|
||
|
|
|
||
|
|
# if obj_num!=500:
|
||
|
|
# continue
|
||
|
|
|
||
|
|
print(
|
||
|
|
f"\n------depth:{depth},obj: {obj_num},cond_pred:{cond_pred},act_pred:{act_pred},max_copy_times:{max_copy_times}-----------")
|
||
|
|
# print(f"\n------literals_num: {literals_num},depth:{depth},iters:{iters},obj_num:{obj_num},max_copy_times:{max_copy_times}-----------")
|
||
|
|
# 为 act建立 add映射
|
||
|
|
|
||
|
|
act_list, start_list, goal_list,literals_obj_count,state_avg,act_avg,act_pred_num = get_act_start_goal(seed=seed, max_copy_times=max_copy_times,\
|
||
|
|
cond_pred=cond_pred, depth=depth, act_pred=act_pred,\
|
||
|
|
total_count=1000,obj_num=obj_num)
|
||
|
|
|
||
|
|
# param_ls = [max_copy_times,depth, obj_num,iters,act_pred_num,literals_num,act_avg]
|
||
|
|
param_ls = [depth,obj_num, cond_pred, act_pred_num, max_copy_times, literals_obj_count, state_avg,act_avg]
|
||
|
|
|
||
|
|
literals_num = cond_pred
|
||
|
|
obt_result = BTTest_act_start_goal(bt_algo_opt=True, act_list=act_list, start_list=start_list, goal_list=goal_list,literals_num=literals_num)
|
||
|
|
|
||
|
|
|
||
|
|
# baseline_result = BTTest_act_start_goal(bt_algo_opt=False, act_list=act_list, start_list=start_list,
|
||
|
|
# goal_list=goal_list,literals_num=literals_num)
|
||
|
|
|
||
|
|
|
||
|
|
a_result=[]
|
||
|
|
a_result.extend(param_ls)
|
||
|
|
a_result.append("OBTEA")
|
||
|
|
a_result.extend(obt_result)
|
||
|
|
all_result.append(a_result)
|
||
|
|
|
||
|
|
a_result=[]
|
||
|
|
a_result.extend(param_ls)
|
||
|
|
a_result.append("Baseline")
|
||
|
|
a_result.extend(baseline_result)
|
||
|
|
all_result.append(a_result)
|
||
|
|
|
||
|
|
|
||
|
|
# import pandas as pd
|
||
|
|
# df = pd.DataFrame(all_result, columns=[
|
||
|
|
# # 'max_actcopy',
|
||
|
|
# # 'literals_num','depth','iters','state','act',
|
||
|
|
# # 'copy_act',
|
||
|
|
# # 'depth', 'obj_num','iters','act_pred_num','literals_num','act_avg',
|
||
|
|
# 'depth','obj_num', 'cond_pred', 'act_pred', 'max_copy_times', 'literals_obj_count', 'state_avg','act_avg',
|
||
|
|
# 'btalgorithm',
|
||
|
|
#
|
||
|
|
# 'tree_size_avg', 'tree_size_std',
|
||
|
|
# 'ticks_avg', 'ticks_std',
|
||
|
|
# 'wm_cond_tick_avg','wm_cond_tick_std',
|
||
|
|
# 'cond_tick_avg','cond_tick_std',
|
||
|
|
# 'cost_avg', 'cost_std',
|
||
|
|
# 'step_avg','step_std',
|
||
|
|
# # 'state_num_avg','state_num_std',
|
||
|
|
# 'expand_num_avg','expand_num_std',
|
||
|
|
# 'for_num_avg','for_num_std',
|
||
|
|
# 'plan_time_avg', 'plan_time_std', 'plan_time_total'])
|
||
|
|
#
|
||
|
|
# time_str = time.strftime('%Y-%m-%d %H:%M:%S', time.localtime()).replace("-","").replace(":","")
|
||
|
|
# csv_file_path = f'bt_randon_o={obj_num_ls[0]}_cp={cond_pred_ls[0]}_ap={act_pred_ls[0]+10}_MAE={max_copy_times_ls[0]}_time={time_str}.csv'
|
||
|
|
# # param_ls = [depth, obj_num, cond_pred, act_pred_num, max_copy_times, literals_obj_count, state_avg, act_avg]
|
||
|
|
# df.to_csv(csv_file_path, index=True)
|
||
|
|
# print("CSV文件已生成:", csv_file_path)
|