From 3e6887c6557afd90ad0b4bfe97b5cd674f638e39 Mon Sep 17 00:00:00 2001 From: weixin_46229132 Date: Sat, 29 Mar 2025 16:53:03 +0800 Subject: [PATCH] =?UTF-8?q?=E6=AF=8F=E4=B8=80=E4=B8=AA=E5=8A=A0=E4=B8=80?= =?UTF-8?q?=E4=B8=AA=E5=A5=96=E5=8A=B1?= MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit --- GA/ga.py | 1 + env_partion.py | 130 +++++++++++++++++++++++++----------------------- human_action.py | 2 +- 3 files changed, 70 insertions(+), 63 deletions(-) diff --git a/GA/ga.py b/GA/ga.py index 0c31d22..ff0bb73 100644 --- a/GA/ga.py +++ b/GA/ga.py @@ -3,6 +3,7 @@ import random # import matplotlib.pyplot as plt import numpy as np +# np.random.seed(42) class GA(object): diff --git a/env_partion.py b/env_partion.py index a9da83f..6629df9 100644 --- a/env_partion.py +++ b/env_partion.py @@ -108,82 +108,88 @@ class PartitionEnv(gym.Env): # 出现无效调整，直接结束 if not valid_adjust: return state, reward, True, False, {} + # 调整合理，计算当前时间 else: - if self.partition_step < self.CUT_NUM: - return state, 0.0, False, False, {} + rectangles = self.if_valid_partition() + + if not rectangles: + reward = -10 + return state, reward, True, False, {} else: - # 完成 4 步后，判断分区是否合理，并计算各个分区的任务卸载率ρ - valid_partition = True - for i in range(len(self.ori_row_cuts) - 1): - for j in range(len(self.ori_col_cuts) - 1): - d = (self.ori_col_cuts[j+1] - self.ori_col_cuts[j]) * self.W * \ - (self.ori_row_cuts[i+1] - - self.ori_row_cuts[i]) * self.H - rho_time_limit = (self.flight_time_factor - self.trans_time_factor) / \ - (self.comp_time_factor - self.trans_time_factor) - rho_energy_limit = (self.battery_energy_capacity - self.flight_energy_factor * d - self.trans_energy_factor * d) / \ - (self.comp_energy_factor * d - - self.trans_energy_factor * d) - if rho_energy_limit < 0: - valid_partition = False - break - rho = min(rho_time_limit, rho_energy_limit) + # 继续进行路径规划 + # 使用遗传算法解多旅行商 + best_time, best_path = self.ga_solver(rectangles) + # print(best_time) + # print(best_path) - flight_time = self.flight_time_factor * d - bs_time = self.bs_time_factor * (1 - rho) * d - - self.rectangles.append({ - 'center': ((self.ori_row_cuts[i] + self.ori_row_cuts[i+1]) * self.H / 2, (self.ori_col_cuts[j+1] + self.ori_col_cuts[j]) * self.W / 2), - 'flight_time': flight_time, - 'bs_time': bs_time, - }) - if not valid_partition: - break - - if not valid_partition: - reward = -10 - return state, reward, True, False, {} + reward = self.BASE_LINE / best_time + if self.partition_step < self.CUT_NUM: + done = False else: - # 继续进行路径规划 - # 使用q_learning解多旅行商 - # cities: [[x1, x2, x3...], [y1, y2, y3...]] 城市坐标 - # rec_center_lt = [rec_info['center'] - # for rec_info in self.rectangles] - # cities = np.column_stack(rec_center_lt) - # cities = np.column_stack((self.center, cities)) + done = True + reward = reward * 3 - # center_idx = [] - # for i in range(self.num_cars - 1): - # cities = np.column_stack((cities, self.center)) - # center_idx.append(cities.shape[1] - 1) + return state, reward, done, False, best_path - # tsp = mTSP(params=self.params, num_cities=cities.shape[1], cities=cities, num_cars=self.num_cars, - # center_idx=center_idx, rectangles=self.rectangles) + def if_valid_partition(self): + rectangles = [] + for i in range(len(self.ori_row_cuts) - 1): + for j in range(len(self.ori_col_cuts) - 1): + d = (self.ori_col_cuts[j+1] - self.ori_col_cuts[j]) * self.W * \ + (self.ori_row_cuts[i+1] - + self.ori_row_cuts[i]) * self.H + rho_time_limit = (self.flight_time_factor - self.trans_time_factor) / \ + (self.comp_time_factor - self.trans_time_factor) + rho_energy_limit = (self.battery_energy_capacity - self.flight_energy_factor * d - self.trans_energy_factor * d) / \ + (self.comp_energy_factor * d - + self.trans_energy_factor * d) + if rho_energy_limit < 0: + return [] + rho = min(rho_time_limit, rho_energy_limit) - # best_time, best_path = tsp.train(self.mTSP_STEPS) + flight_time = self.flight_time_factor * d + bs_time = self.bs_time_factor * (1 - rho) * d - # 使用遗传算法解多旅行商 - cities = [self.center] - for rec in self.rectangles: - cities.append(rec['center']) - cities = np.array(cities) + rectangles.append({ + 'center': ((self.ori_row_cuts[i] + self.ori_row_cuts[i+1]) * self.H / 2, (self.ori_col_cuts[j+1] + self.ori_col_cuts[j]) * self.W / 2), + 'flight_time': flight_time, + 'bs_time': bs_time, + }) + return rectangles + + # def q_learning_solver(self): + # 使用q_learning解多旅行商 + # cities: [[x1, x2, x3...], [y1, y2, y3...]] 城市坐标 + # rec_center_lt = [rec_info['center'] + # for rec_info in rectangles] + # cities = np.column_stack(rec_center_lt) + # cities = np.column_stack((self.center, cities)) - center_idx = [0] - for i in range(self.num_cars - 1): - cities = np.row_stack((cities, self.center)) - center_idx.append(cities.shape[0] - 1) + # center_idx = [] + # for i in range(self.num_cars - 1): + # cities = np.column_stack((cities, self.center)) + # center_idx.append(cities.shape[1] - 1) - ga = GA(num_drones=self.num_cars, num_city=cities.shape[0], num_total=20, - data=cities, to_process_idx=center_idx, rectangles=self.rectangles) + # tsp = mTSP(params=self.params, num_cities=cities.shape[1], cities=cities, num_cars=self.num_cars, + # center_idx=center_idx, rectangles=rectangles) - best_path, best_time = ga.run() + # best_time, best_path = tsp.train(self.mTSP_STEPS) - # print(best_time) - # print(best_path) + def ga_solver(self, rectangles): + cities = [self.center] + for rec in rectangles: + cities.append(rec['center']) + cities = np.array(cities) - reward = self.BASE_LINE / best_time + center_idx = [0] + for i in range(self.num_cars - 1): + cities = np.row_stack((cities, self.center)) + center_idx.append(cities.shape[0] - 1) - return state, reward, True, False, best_path + ga = GA(num_drones=self.num_cars, num_city=cities.shape[0], num_total=20, + data=cities, to_process_idx=center_idx, rectangles=rectangles) + best_path, best_time = ga.run() + return best_time, best_path def render(self): if self.phase == 1: diff --git a/human_action.py b/human_action.py index f8d6006..007eff1 100644 --- a/human_action.py +++ b/human_action.py @@ -10,7 +10,7 @@ print('state:', state) # action_series = [[0.67], [0], [0], [0], [0.7]] # action_series = [0, 0, 3, 0, 10] -action_series = [[0.2], [0.4], [0.7], [0.5]] +# action_series = [[0.2], [0.4], [0.7], [0.5]] action_series = [[-0.1], [0], [0], [0]] for i in range(100):