172 lines
9.0 KiB
Python
172 lines
9.0 KiB
Python
import random
|
||
import numpy as np
|
||
# 导入数据结构和工具类（根据实际项目结构调整导入路径）
|
||
from data_structures import OrderData, RiskEnterpriseData, SupplierData, Config
|
||
from chromosome_utils import ChromosomeUtils
|
||
from objective_calculator import ObjectiveCalculator
|
||
from encoder import Encoder
|
||
from genetic_operators import GeneticOperator
|
||
from nsga2 import NSGA2
|
||
from visualizer import ResultVisualizer
|
||
|
||
|
||
def main():
|
||
"""主函数：执行NSGA-II算法求解多目标优化问题"""
|
||
try:
|
||
# 1. 初始化随机种子（确保结果可复现）
|
||
random.seed(42)
|
||
np.random.seed(42)
|
||
|
||
# 2. 初始化数据（订单、风险企业、供应商、算法配置）
|
||
print("初始化数据结构...")
|
||
order_data = OrderData() # 订单数据（需求、交货期等）
|
||
risk_data = RiskEnterpriseData() # 风险企业数据
|
||
supplier_data = SupplierData() # 供应商数据
|
||
config = Config() # 算法参数配置
|
||
|
||
# 3. 初始化工具类和算法组件
|
||
print("初始化算法组件...")
|
||
utils = ChromosomeUtils(order_data, risk_data, supplier_data) # 染色体工具
|
||
calculator = ObjectiveCalculator(order_data, risk_data, supplier_data, utils, config) # 目标函数计算器
|
||
encoder = Encoder(config, utils) # 种群初始化编码器
|
||
genetic_op = GeneticOperator(config, utils) # 遗传操作器（交叉、变异）
|
||
nsga2 = NSGA2(config.pop_size, config.objective_num) # NSGA-II算法实例
|
||
visualizer = ResultVisualizer(utils) # 结果可视化工具
|
||
|
||
# 4. 初始化种群
|
||
print("初始化种群...")
|
||
population = encoder.initialize_population()
|
||
print(f"初始化种群完成，种群大小: {population.shape if population.size > 0 else '空'}")
|
||
# 若种群初始化失败（为空），直接退出
|
||
if population.size == 0:
|
||
print("错误：种群初始化失败，无法继续进化")
|
||
return
|
||
|
||
# 5. 记录进化过程中的历史数据
|
||
all_objectives = [] # 所有代的目标函数值
|
||
convergence_history = [] # 收敛趋势（每代最优前沿的平均目标值）
|
||
best_front = [] # 最终帕累托前沿解
|
||
best_front_objs = [] # 最终帕累托前沿的目标值
|
||
no_improve_count = 0 # 无改进计数器（用于早停）
|
||
prev_best_avg = float('inf') # 上一代的最优平均目标值
|
||
|
||
# 6. 进化主循环
|
||
print(f"开始进化（最大代数：{config.max_generations}，早停耐心：{config.early_stop_patience}）...")
|
||
for generation in range(config.max_generations):
|
||
try:
|
||
# 计算当前种群的目标函数值
|
||
objectives = [calculator.calculate_objectives(chrom) for chrom in population]
|
||
all_objectives.extend(objectives) # 记录所有目标值
|
||
|
||
# 非支配排序，获取当前代的帕累托前沿
|
||
ranks, fronts = nsga2.fast_non_dominated_sort(objectives)
|
||
current_front = fronts[0] if fronts else [] # 第0层为最优前沿
|
||
current_front_objs = [objectives[i] for i in current_front] if current_front else []
|
||
best_front = population[current_front] if current_front else [] # 更新当前最优前沿解
|
||
best_front_objs = current_front_objs # 更新当前最优前沿目标值
|
||
|
||
# 记录收敛趋势（基于最优前沿的平均目标值）
|
||
if len(current_front_objs) > 0:
|
||
avg_cost = sum(obj[0] for obj in current_front_objs) / len(current_front_objs)
|
||
avg_tardiness = sum(obj[1] for obj in current_front_objs) / len(current_front_objs)
|
||
convergence_history.append((avg_cost, avg_tardiness))
|
||
|
||
# 判断是否改进（用于早停）
|
||
current_avg = avg_cost + avg_tardiness # 合并两个目标的平均值
|
||
if abs(current_avg - prev_best_avg) < 1e-4: # 变化小于阈值，视为无改进
|
||
no_improve_count += 1
|
||
else:
|
||
no_improve_count = 0 # 有改进，重置计数器
|
||
prev_best_avg = current_avg
|
||
else:
|
||
no_improve_count += 1 # 无前沿解，视为无改进
|
||
|
||
# 选择操作（锦标赛选择）
|
||
selected = nsga2.selection(population, objectives)
|
||
# 校验选择后的种群大小
|
||
assert len(selected) == config.pop_size, \
|
||
f"选择后的种群大小({len(selected)})与目标大小({config.pop_size})不符"
|
||
|
||
# 交叉操作（两点交叉）- 修复索引越界问题
|
||
offspring = [] # 子代种群
|
||
selected_len = len(selected) # selected的长度（等于pop_size）
|
||
i = 0
|
||
max_iter = 2 * config.pop_size # 最大迭代次数，避免无限循环
|
||
iter_count = 0
|
||
while len(offspring) < config.pop_size and iter_count < max_iter:
|
||
iter_count += 1
|
||
# 检查i是否超出selected范围，若超出则重置（循环使用个体）
|
||
if i >= selected_len:
|
||
i = 0
|
||
# 尝试两两交叉（需i+1在范围内，且满足交叉概率）
|
||
if i + 1 < selected_len and random.random() < config.crossover_prob:
|
||
parent1 = selected[i]
|
||
parent2 = selected[i + 1]
|
||
child1, child2 = genetic_op.two_point_crossover(parent1, parent2)
|
||
# 添加child1（若未达到种群大小）
|
||
if len(offspring) < config.pop_size:
|
||
offspring.append(child1)
|
||
# 添加child2（若未达到种群大小）
|
||
if len(offspring) < config.pop_size:
|
||
offspring.append(child2)
|
||
i += 2 # 处理下一对个体
|
||
else:
|
||
# 直接添加当前父代（避免i+1越界）
|
||
offspring.append(selected[i])
|
||
i += 1 # 处理下一个个体（步长改为1，避免快速越界）
|
||
|
||
# 若迭代次数用尽仍未生成足够子代，补充随机个体（健壮性处理）
|
||
while len(offspring) < config.pop_size:
|
||
offspring.append(encoder._generate_random_chromosome()) # 需确保该方法可访问
|
||
|
||
# 变异操作（均匀变异）
|
||
offspring = [
|
||
genetic_op.uniform_mutation(chrom) if random.random() < config.mutation_prob else chrom
|
||
for chrom in offspring
|
||
]
|
||
offspring = np.array(offspring[:config.pop_size]) # 确保子代大小严格等于pop_size
|
||
|
||
# 合并父代和子代，准备环境选择
|
||
combined = np.vstack([population, offspring]) # 合并种群
|
||
# 计算合并种群的目标函数值
|
||
combined_objs = objectives + [calculator.calculate_objectives(chrom) for chrom in offspring]
|
||
|
||
# 环境选择（保留最优的pop_size个个体）
|
||
population, objectives = nsga2.environmental_selection(combined, combined_objs)
|
||
# 校验环境选择后的种群大小
|
||
assert len(population) == config.pop_size, \
|
||
f"环境选择后的种群大小({len(population)})与目标大小({config.pop_size})不符"
|
||
|
||
# 早停检查（连续多代无改进则停止）
|
||
if no_improve_count >= config.early_stop_patience:
|
||
print(f"早停触发：连续{no_improve_count}代无改进，终止于第{generation}代")
|
||
break
|
||
|
||
# 每50代打印一次进度
|
||
if generation % 50 == 0:
|
||
front_size = len(current_front) if current_front else 0
|
||
print(f"第{generation}代完成，当前最优前沿大小: {front_size}，无改进计数: {no_improve_count}")
|
||
|
||
except Exception as e:
|
||
print(f"第{generation}代进化出错：{str(e)}，跳过当前代")
|
||
continue
|
||
|
||
# 7. 结果可视化与输出
|
||
print("进化完成，处理结果...")
|
||
if len(best_front_objs) > 0:
|
||
visualizer.plot_pareto_front(all_objectives, best_front_objs) # 绘制帕累托前沿
|
||
visualizer.plot_convergence(convergence_history) # 绘制收敛趋势
|
||
visualizer.print_pareto_solutions(best_front, best_front_objs) # 打印最优解详情
|
||
else:
|
||
print("未找到有效帕累托前沿解")
|
||
|
||
except Exception as e:
|
||
print(f"程序运行出错：{str(e)}")
|
||
import traceback
|
||
traceback.print_exc() # 打印详细错误栈
|
||
|
||
|
||
if __name__ == "__main__":
|
||
print("程序启动...")
|
||
main()
|
||
print("程序结束") |