def exec(self, p):
# 初始化基因，基因包含染色体和染色体的适应度
self.init_population()
self.gbest = self.find_best()
lb = self.particles.pop()
best_score = calculate_fitness(lb.solution, self.cloudlets, self.vms)
global_best_score = calculate_fitness(self.gbest.solution,
self.cloudlets, self.vms)
results = []
# 迭代过程 仅包含速度和位置的更新
for t in range(self.times):
for p in self.particles:
if p is self.gbest:
continue
if p is lb:
continue
self.update_velocity(p, lb, self.gbest)
self.update_position(p)
score = calculate_fitness(p.solution, self.cloudlets, self.vms)
if score > best_score:
best_score = score
lb = p
if score > global_best_score:
global_best_score = score
self.gbest = p
tempGbest = self.getSeries(self.gbest.solution)
if tempGbest.fitness > global_best_score:
global_best_score = tempGbest.fitness
self.gbest = tempGbest
results.append(global_best_score)
if t % 20 == 0:
print("CPSO iter: ", t, " / ", self.times, ", 适应度: ",
global_best_score)
# return gb.solution
return results
def scheduler_main(self):
results = [0 for _ in range(self.times)]
fitness = 0
for it in range(self.times):
best_time = 0
for ant_id in range(self.population_number):
topo_jobs = self.gen_topo_jobs()
fitness = calculate_fitness(topo_jobs, self.cloudlets,
self.vms)
if fitness > best_time:
self.best_topo = topo_jobs
best_time = fitness
assert self.best_topo is not None
self.update_topo()
results[it] = best_time
if it % 20 == 0:
print("ACO iter: ", it, " / ", self.times, ", 适应度: ", fitness)
plt.plot(range(self.times), results)
plt.xlabel("迭代次数")
plt.ylabel("适应度")
plt.rcParams['font.sans-serif'] = ['KaiTi'] # 指定默认字体
plt.rcParams['axes.unicode_minus'] = False # 解决保存图像是负号'-'显示为方块的问题
plt.title("蚁群算法求解云任务调度负载均衡问题")
# plt.savefig('img/ACScheduler-1.1_0.9-popu100-iter200.png', dpi=300,
# format='png') # bbox_inches="tight"解决X轴时间两个字不被保存的问题
plt.show()
return results
def mutation(self):
for i in range(self.population_number):
point = np.random.randint(0, self.cloudlet_num)
if np.random.rand() < self.mp:
self.genes[i].solution[point] = np.random.randint(
0, self.machine_number)
self.genes[i].fitness = calculate_fitness(
self.genes[i].solution, self.cloudlets, self.vms)
def init_population(self):
for i in range(self.population_number):
gene = Gene()
gene.solution = np.random.randint(0, self.machine_number,
self.cloudlet_num)
gene.fitness = calculate_fitness(gene.solution, self.cloudlets,
self.vms)
self.genes.append(gene)
def find_best(self):
best_score = 0
best_particle = None
for p in self.particles:
score = calculate_fitness(p.solution, self.cloudlets, self.vms)
if score > best_score:
best_score = score
best_particle = p
return best_particle
def init_population(self):
for _ in range(self.population_number):
size = self.cloudlet_num
p = DParticle(size)
p.solution = np.random.randint(0,
self.machine_number,
size=self.cloudlet_num)
p.velocity = -5 + 10 * np.random.rand(self.cloudlet_num)
p.fitness = calculate_fitness(p.solution, self.cloudlets, self.vms)
self.particles.append(p)
def crossover(self, g1, g2):
for i in range(self.population_number):
if np.random.rand() < self.cp:
point = np.random.randint(0, self.cloudlet_num)
new_gene = Gene()
new_gene.solution = np.hstack(
(g1.solution[0:point], g2.solution[point:]))
new_gene.fitness = calculate_fitness(new_gene.solution,
self.cloudlets, self.vms)
self.genes[i] = new_gene
def exec(self):
# 初始化基因，基因包含染色体和染色体的适应度
self.init_population()
self.gbest = self.find_best()
lb = self.particles.pop()
best_score = calculate_fitness(lb.solution, self.cloudlets, self.vms)
global_best_score = calculate_fitness(self.gbest.solution,
self.cloudlets, self.vms)
results = []
# 迭代过程 仅包含速度和位置的更新
for t in range(self.times):
# 升序排序
self.particles.sort(key=lambda x: x.fitness)
ind = 0
for p in self.particles:
if p is self.gbest:
continue
if p is lb:
continue
self.update_velocity(p, lb, self.gbest)
self.update_position(p)
if ind > self.population_number / 3 or ind < 2 / 3 * self.population_number:
self.update_position_copy1(p, lb, 2)
else:
self.update_position_copy1(p, self.gbest, 1)
ind += 1
score = calculate_fitness(p.solution, self.cloudlets, self.vms)
if score > best_score:
best_score = score
lb = p
if score > global_best_score:
global_best_score = score
self.gbest = p
tempGbest = self.getSeries(self.gbest.solution)
if tempGbest.fitness > global_best_score:
global_best_score = tempGbest.fitness
self.gbest = tempGbest
results.append(global_best_score)
if t % 20 == 0:
print("CRPSO iter: ", t, " / ", self.times, ", 适应度: ",
global_best_score)
# print(gb.solution)
return results
def sa_main(self):
self.best_way = np.random.randint(0, self.machine_number,
self.cloudlet_num)
# temp_solution = [0] * self.times
# temp_value = [0] * self.times
results = list()
for t in range(self.times):
if t % 20 == 0:
print("SA Iter, ", t, "/", self.times, ", 适应度：", self.best_bl)
# for j in range(self.Lk):
for j in range(self.population_number):
self.best_bl = calculate_fitness(self.best_way, self.cloudlets,
self.vms)
way1 = self.gen_new_way(self.best_way)
way1 = self.gen_new_way_change(way1)
# print("新解的适应度:", bl1)
bl1 = calculate_fitness(way1, self.cloudlets, self.vms)
if bl1 > self.best_bl:
self.best_way = way1
# if bl1 > temp_value[j]:
# temp_solution[j] = way1
# temp_value[j] = bl1
else:
p = np.exp((bl1 - self.best_bl) / self.T)
if np.random.rand() < p:
self.best_way = way1
results.append(self.best_bl)
# results.append(temp_value[t])
self.T = self.alpha * self.T
# print("最佳适应度", self.best_bl)
# print("最佳路径", self.best_way)
# print("最佳适应度：", self.best_bl)
# print("策略：", self.best_way)
return results
def getSeries(self, v0):
tempSet = [[0] * self.cloudlet_num] * 30
tempSet[0] = [x / self.machine_number for x in v0]
for i in range(1, 30):
tempSet[i] = [x for x in logistic_function(tempSet[i - 1], 3.5)]
for i in range(30):
tempSet[i] = [int(self.machine_number * x) for x in tempSet[i]]
# print(tempSet)
tempFit = []
for i in range(30):
tempFit.append(
calculate_fitness(tempSet[i], self.cloudlets, self.vms))
best_score = 0
best_solution = None
for i in range(self.chaosNum):
if tempFit[i] > best_score:
best_score = tempFit[i]
best_solution = tempSet[i]
res = DParticle(self.cloudlet_num)
res.velocity = self.gbest.velocity
res.solution = best_solution
res.fitness = best_score
return res
