当前位置：首页 > news >正文

无人机编队控制图案组合的实现方法与仿真技术详解

news 2025/10/21 17:05:42

1. 引言

随着无人机技术的快速发展，无人机编队飞行已经成为航空领域的研究热点。从军事侦察到商业表演，从物流配送到农业植保，无人机编队技术展现出了巨大的应用潜力。本文将详细介绍无人机编队进行控制图案组合的实现方法，包括理论基础、算法设计、仿真验证和代码实现。

2. 技术背景与理论基础

2.1 编队控制基本概念

无人机编队控制是指多架无人机通过协同控制算法，保持特定的几何构型或动态变换队形的技术。主要涉及以下关键技术：

队形保持：维持预定的几何构型
队形变换：从一种队形平滑过渡到另一种队形
避障避撞：确保无人机之间的安全距离
通信协调：实现无人机间的信息交换

2.2 数学模型

单个四旋翼无人机的动力学模型可以表示为：

位置动力学：
ẋ = v_x
ẏ = v_y  
ż = v_z速度动力学：
v̇_x = (1/m) * (cos(φ)sin(θ)cos(ψ) + sin(φ)sin(ψ)) * F
v̇_y = (1/m) * (cos(φ)sin(θ)sin(ψ) - sin(φ)cos(ψ)) * F
v̇_z = (1/m) * (cos(φ)cos(θ)) * F - g

其中，(x, y, z)为位置坐标，(φ, θ, ψ)为姿态角，F为总推力，m为无人机质量，g为重力加速度。

3. 编队控制架构设计

3.1 分层控制架构

我们采用三层控制架构：

任务规划层：负责图案设计和队形切换策略
路径规划层：生成每架无人机的运动轨迹
底层控制层：实现单机的位置和姿态控制

3.2 编队控制策略

主要采用以下三种控制策略：

3.2.1 领航-跟随法（Leader-Follower）

# 领航-跟随法核心思想
class LeaderFollowerFormation:def __init__(self, num_drones, formation_type):self.num_drones = num_dronesself.formation_type = formation_typeself.leader_pos = np.zeros(3)self.follower_offsets = self.calculate_offsets()def calculate_offsets(self):"""计算跟随者相对于领航者的位置偏移"""offsets = []if self.formation_type == "line":for i in range(1, self.num_drones):offsets.append([-2*i, 0, 0])  # 直线队形elif self.formation_type == "triangle":# 三角队形的偏移计算row = 1col = 0for i in range(1, self.num_drones):offsets.append([-2*row, 2*col-row, 0])col += 1if col > row:row += 1col = 0return np.array(offsets)

3.2.2 虚拟结构法（Virtual Structure）

class VirtualStructureFormation:def __init__(self, num_drones, formation_shape):self.num_drones = num_dronesself.formation_shape = formation_shapeself.center_pos = np.zeros(3)self.orientation = np.eye(3)def generate_formation_points(self, pattern_type, scale=5.0):"""生成不同图案的编队点"""points = []if pattern_type == "circle":# 圆形编队for i in range(self.num_drones):angle = 2 * np.pi * i / self.num_dronesx = scale * np.cos(angle)y = scale * np.sin(angle)points.append([x, y, 0])elif pattern_type == "star":# 五角星编队outer_radius = scaleinner_radius = scale * 0.4for i in range(self.num_drones):angle = 2 * np.pi * i / self.num_dronesif i % 2 == 0:r = outer_radiuselse:r = inner_radiusx = r * np.cos(angle - np.pi/2)y = r * np.sin(angle - np.pi/2)points.append([x, y, 0])elif pattern_type == "grid":# 网格编队grid_size = int(np.sqrt(self.num_drones))spacing = scale / grid_sizeidx = 0for i in range(grid_size):for j in range(grid_size):if idx < self.num_drones:x = (i - grid_size/2) * spacingy = (j - grid_size/2) * spacingpoints.append([x, y, 0])idx += 1return np.array(points)

3.2.3 基于行为的方法（Behavior-Based）

class BehaviorBasedFormation:def __init__(self, num_drones):self.num_drones = num_dronesself.behaviors = ['cohesion', 'separation', 'alignment', 'formation']def cohesion_behavior(self, drone_pos, neighbors_pos):"""聚集行为：向邻居中心移动"""if len(neighbors_pos) == 0:return np.zeros(3)center = np.mean(neighbors_pos, axis=0)return (center - drone_pos) * 0.01def separation_behavior(self, drone_pos, neighbors_pos, min_dist=2.0):"""分离行为：避免碰撞"""repulsion = np.zeros(3)for neighbor_pos in neighbors_pos:diff = drone_pos - neighbor_posdist = np.linalg.norm(diff)if dist < min_dist and dist > 0:repulsion += diff / (dist ** 2)return repulsion * 0.1def alignment_behavior(self, drone_vel, neighbors_vel):"""对齐行为：速度一致"""if len(neighbors_vel) == 0:return np.zeros(3)avg_vel = np.mean(neighbors_vel, axis=0)return (avg_vel - drone_vel) * 0.05

4. 图案组合实现方法

4.1 图案定义与参数化

class PatternLibrary:def __init__(self):self.patterns = {}self.load_basic_patterns()def load_basic_patterns(self):"""加载基础图案库"""# 2D图案self.patterns['heart'] = self.generate_heart_patternself.patterns['letter'] = self.generate_letter_patternself.patterns['logo'] = self.generate_logo_pattern# 3D图案self.patterns['sphere'] = self.generate_sphere_patternself.patterns['helix'] = self.generate_helix_patternself.patterns['cube'] = self.generate_cube_patterndef generate_heart_pattern(self, num_drones, scale=10):"""生成心形图案"""points = []t = np.linspace(0, 2*np.pi, num_drones)for i in range(num_drones):x = scale * 16 * np.sin(t[i])**3 / 16y = scale * (13*np.cos(t[i]) - 5*np.cos(2*t[i]) - 2*np.cos(3*t[i]) - np.cos(4*t[i])) / 16z = 0points.append([x, y, z])return np.array(points)def generate_sphere_pattern(self, num_drones, radius=10):"""生成球形图案"""points = []# 使用斐波那契螺旋均匀分布点golden_angle = np.pi * (3 - np.sqrt(5))for i in range(num_drones):theta = golden_angle * iy = 1 - (i / float(num_drones - 1)) * 2radius_at_y = np.sqrt(1 - y * y)x = np.cos(theta) * radius_at_yz = np.sin(theta) * radius_at_ypoints.append([x*radius, y*radius, z*radius])return np.array(points)

4.2 队形切换算法

class FormationTransition:def __init__(self):self.transition_time = 10.0  # 切换时间self.safety_distance = 1.5   # 安全距离def hungarian_assignment(self, current_positions, target_positions):"""使用匈牙利算法进行最优分配"""from scipy.optimize import linear_sum_assignment# 计算代价矩阵（距离矩阵）cost_matrix = np.zeros((len(current_positions), len(target_positions)))for i in range(len(current_positions)):for j in range(len(target_positions)):cost_matrix[i, j] = np.linalg.norm(current_positions[i] - target_positions[j])# 求解最优分配row_ind, col_ind = linear_sum_assignment(cost_matrix)# 返回分配结果assignments = {}for i, j in zip(row_ind, col_ind):assignments[i] = jreturn assignmentsdef generate_transition_trajectory(self, start_pos, end_pos, t_total):"""生成平滑过渡轨迹（五次多项式）"""def quintic_polynomial(t, t_f, x_0, x_f):"""五次多项式轨迹生成"""tau = t / t_freturn x_0 + (x_f - x_0) * (10*tau**3 - 15*tau**4 + 6*tau**5)# 生成时间序列dt = 0.1time_steps = np.arange(0, t_total + dt, dt)trajectory = []for t in time_steps:pos = quintic_polynomial(t, t_total, start_pos, end_pos)trajectory.append(pos)return np.array(trajectory)def check_collision_free(self, trajectories, safety_distance):"""检查轨迹是否无碰撞"""num_drones = len(trajectories)num_steps = len(trajectories[0])for step in range(num_steps):for i in range(num_drones):for j in range(i+1, num_drones):dist = np.linalg.norm(trajectories[i][step] - trajectories[j][step])if dist < safety_distance:return False, (i, j, step)return True, None

5. 控制算法实现

5.1 PID控制器

class PIDController:def __init__(self, kp, ki, kd, dt=0.01):self.kp = kpself.ki = kiself.kd = kdself.dt = dtself.error_sum = np.zeros(3)self.last_error = np.zeros(3)def update(self, setpoint, current):"""PID控制更新"""error = setpoint - current# P项p_term = self.kp * error# I项self.error_sum += error * self.dti_term = self.ki * self.error_sum# D项error_diff = (error - self.last_error) / self.dtd_term = self.kd * error_diffself.last_error = errorreturn p_term + i_term + d_term

5.2 模型预测控制（MPC）

import cvxpy as cpclass MPCController:def __init__(self, prediction_horizon=20, control_horizon=10):self.N = prediction_horizonself.M = control_horizonself.dt = 0.1# 系统矩阵（简化的线性模型）self.A = np.eye(6) + self.dt * np.array([[0, 0, 0, 1, 0, 0],[0, 0, 0, 0, 1, 0],[0, 0, 0, 0, 0, 1],[0, 0, 0, 0, 0, 0],[0, 0, 0, 0, 0, 0],[0, 0, 0, 0, 0, 0]])self.B = self.dt * np.array([[0, 0, 0],[0, 0, 0],[0, 0, 0],[1, 0, 0],[0, 1, 0],[0, 0, 1]])def solve(self, x0, x_ref, u_max=5.0):"""求解MPC优化问题"""# 定义优化变量x = cp.Variable((6, self.N + 1))u = cp.Variable((3, self.N))# 定义代价函数Q = np.diag([10, 10, 10, 1, 1, 1])  # 状态权重R = np.eye(3) * 0.1                  # 控制权重cost = 0constraints = [x[:, 0] == x0]for k in range(self.N):cost += cp.quad_form(x[:, k] - x_ref[:, k], Q)cost += cp.quad_form(u[:, k], R)# 系统动力学约束constraints += [x[:, k + 1] == self.A @ x[:, k] + self.B @ u[:, k]]# 控制输入约束constraints += [cp.norm(u[:, k], 'inf') <= u_max]# 终端代价cost += cp.quad_form(x[:, self.N] - x_ref[:, self.N], Q * 10)# 求解优化问题prob = cp.Problem(cp.Minimize(cost), constraints)prob.solve(solver=cp.OSQP, warm_start=True)if prob.status == cp.OPTIMAL:return u[:, 0].valueelse:return np.zeros(3)

6. 仿真环境搭建

6.1 基于Python的仿真框架

import numpy as np
import matplotlib.pyplot as plt
from matplotlib.animation import FuncAnimation
from mpl_toolkits.mplot3d import Axes3Dclass DroneSwarmSimulator:def __init__(self, num_drones, dt=0.01):self.num_drones = num_dronesself.dt = dt# 无人机状态self.positions = np.zeros((num_drones, 3))self.velocities = np.zeros((num_drones, 3))self.accelerations = np.zeros((num_drones, 3))# 目标位置self.target_positions = np.zeros((num_drones, 3))# 控制器self.controllers = []for i in range(num_drones):self.controllers.append(PIDController(kp=np.array([2.0, 2.0, 2.0]),ki=np.array([0.1, 0.1, 0.1]),kd=np.array([1.0, 1.0, 1.0]),dt=dt))# 历史记录self.position_history = []def set_formation(self, formation_points):"""设置编队目标位置"""self.target_positions = formation_pointsdef update(self):"""更新仿真状态"""for i in range(self.num_drones):# 计算控制输入control_input = self.controllers[i].update(self.target_positions[i], self.positions[i])# 限制最大加速度max_acc = 5.0control_input = np.clip(control_input, -max_acc, max_acc)# 更新加速度self.accelerations[i] = control_input# 更新速度（考虑最大速度限制）self.velocities[i] += self.accelerations[i] * self.dtmax_vel = 10.0vel_norm = np.linalg.norm(self.velocities[i])if vel_norm > max_vel:self.velocities[i] = self.velocities[i] / vel_norm * max_vel# 更新位置self.positions[i] += self.velocities[i] * self.dt# 记录历史self.position_history.append(self.positions.copy())def simulate(self, duration):"""运行仿真"""steps = int(duration / self.dt)for _ in range(steps):self.update()return self.position_historydef animate_3d(self, save_path=None):"""3D动画显示"""fig = plt.figure(figsize=(12, 9))ax = fig.add_subplot(111, projection='3d')# 设置坐标轴ax.set_xlabel('X (m)')ax.set_ylabel('Y (m)')ax.set_zlabel('Z (m)')ax.set_title('UAV Swarm Formation Simulation')# 设置显示范围limit = 20ax.set_xlim([-limit, limit])ax.set_ylim([-limit, limit])ax.set_zlim([0, limit*2])# 初始化散点图scatter = ax.scatter([], [], [], c='b', marker='o', s=50)target_scatter = ax.scatter([], [], [], c='r', marker='x', s=100)# 轨迹线lines = [ax.plot([], [], [], 'b-', alpha=0.3)[0] for _ in range(self.num_drones)]def init():scatter._offsets3d = ([], [], [])target_scatter._offsets3d = ([], [], [])for line in lines:line.set_data([], [])line.set_3d_properties([])return [scatter, target_scatter] + linesdef update(frame):if frame < len(self.position_history):positions = self.position_history[frame]# 更新无人机位置scatter._offsets3d = (positions[:, 0], positions[:, 1], positions[:, 2])# 更新目标位置target_scatter._offsets3d = (self.target_positions[:, 0],self.target_positions[:, 1],self.target_positions[:, 2])# 更新轨迹trail_length = min(50, frame)if frame > trail_length:for i, line in enumerate(lines):trail = np.array(self.position_history[frame-trail_length:frame+1])line.set_data(trail[:, i, 0], trail[:, i, 1])line.set_3d_properties(trail[:, i, 2])return [scatter, target_scatter] + linesanim = FuncAnimation(fig, update, init_func=init,frames=len(self.position_history),interval=50, blit=True)if save_path:anim.save(save_path, writer='pillow', fps=20)plt.show()return anim

6.2 ROS2仿真接口

import rclpy
from rclpy.node import Node
from geometry_msgs.msg import PoseStamped, Twist
from nav_msgs.msg import Path
from std_msgs.msg import String
import jsonclass SwarmControlNode(Node):def __init__(self):super().__init__('swarm_control_node')self.num_drones = 10# 发布器self.cmd_publishers = []self.path_publishers = []for i in range(self.num_drones):# 速度指令发布cmd_pub = self.create_publisher(Twist, f'/drone_{i}/cmd_vel', 10)self.cmd_publishers.append(cmd_pub)# 路径发布path_pub = self.create_publisher(Path, f'/drone_{i}/planned_path', 10)self.path_publishers.append(path_pub)# 订阅器self.pose_subscribers = []for i in range(self.num_drones):pose_sub = self.create_subscription(PoseStamped,f'/drone_{i}/pose',lambda msg, idx=i: self.pose_callback(msg, idx),10)self.pose_subscribers.append(pose_sub)# 编队指令订阅self.formation_sub = self.create_subscription(String,'/formation_command',self.formation_callback,10)# 当前位置和目标位置self.current_poses = [None] * self.num_dronesself.target_formation = None# 控制定时器self.control_timer = self.create_timer(0.1, self.control_loop)self.get_logger().info('Swarm Control Node initialized')def pose_callback(self, msg, drone_idx):"""处理无人机位置反馈"""self.current_poses[drone_idx] = msgdef formation_callback(self, msg):"""处理编队指令"""try:command = json.loads(msg.data)pattern_type = command.get('pattern_type', 'circle')scale = command.get('scale', 10.0)# 生成编队图案pattern_gen = PatternLibrary()if pattern_type in pattern_gen.patterns:self.target_formation = pattern_gen.patterns[pattern_type](self.num_drones, scale)self.get_logger().info(f'Formation changed to: {pattern_type}')except Exception as e:self.get_logger().error(f'Failed to parse formation command: {e}')def control_loop(self):"""主控制循环"""if self.target_formation is None:return# 检查是否所有无人机位置都已更新if None in self.current_poses:return# 计算并发布控制指令for i in range(self.num_drones):if self.current_poses[i] is not None:# 获取当前位置current_pos = np.array([self.current_poses[i].pose.position.x,self.current_poses[i].pose.position.y,self.current_poses[i].pose.position.z])# 计算速度指令（简单的P控制）target_pos = self.target_formation[i]error = target_pos - current_pos# 创建Twist消息cmd_msg = Twist()cmd_msg.linear.x = np.clip(error[0] * 0.5, -2.0, 2.0)cmd_msg.linear.y = np.clip(error[1] * 0.5, -2.0, 2.0)cmd_msg.linear.z = np.clip(error[2] * 0.5, -2.0, 2.0)# 发布控制指令self.cmd_publishers[i].publish(cmd_msg)def main(args=None):rclpy.init(args=args)node = SwarmControlNode()try:rclpy.spin(node)except KeyboardInterrupt:passfinally:node.destroy_node()rclpy.shutdown()if __name__ == '__main__':main()

7. 完整示例：动态图案切换

class DynamicFormationDemo:def __init__(self, num_drones=20):self.num_drones = num_dronesself.simulator = DroneSwarmSimulator(num_drones)self.pattern_lib = PatternLibrary()self.transition_planner = FormationTransition()def run_demo(self):"""运行完整的动态编队演示"""# 定义演示序列demo_sequence = [{'pattern': 'grid', 'duration': 5.0, 'scale': 15.0},{'pattern': 'circle', 'duration': 5.0, 'scale': 10.0},{'pattern': 'star', 'duration': 5.0, 'scale': 12.0},{'pattern': 'heart', 'duration': 5.0, 'scale': 8.0},{'pattern': 'sphere', 'duration': 5.0, 'scale': 10.0},]# 初始化起始位置（随机分布）initial_positions = np.random.randn(self.num_drones, 3) * 5initial_positions[:, 2] = np.abs(initial_positions[:, 2]) + 5  # 确保高度为正self.simulator.positions = initial_positionsall_trajectories = []for i, config in enumerate(demo_sequence):print(f"Switching to pattern: {config['pattern']}")# 生成目标图案if config['pattern'] == 'grid':pattern_gen = VirtualStructureFormation(self.num_drones, 'grid')target_positions = pattern_gen.generate_formation_points('grid', config['scale'])else:target_positions = self.pattern_lib.patterns[config['pattern']](self.num_drones, config['scale'])# 添加高度偏移target_positions[:, 2] += 15# 计算最优分配current_positions = self.simulator.positions.copy()assignments = self.transition_planner.hungarian_assignment(current_positions, target_positions)# 重新排列目标位置reordered_targets = np.zeros_like(target_positions)for drone_id, target_id in assignments.items():reordered_targets[drone_id] = target_positions[target_id]# 生成过渡轨迹transition_trajectories = []for j in range(self.num_drones):trajectory = self.transition_planner.generate_transition_trajectory(current_positions[j], reordered_targets[j], config['duration'])transition_trajectories.append(trajectory)# 检查碰撞is_safe, collision_info = self.transition_planner.check_collision_free(transition_trajectories, self.transition_planner.safety_distance)if not is_safe:print(f"Warning: Potential collision detected at step {collision_info[2]} "f"between drones {collision_info[0]} and {collision_info[1]}")# 这里可以添加避障策略# 执行轨迹for step in range(len(transition_trajectories[0])):for j in range(self.num_drones):self.simulator.positions[j] = transition_trajectories[j][step]self.simulator.position_history.append(self.simulator.positions.copy())# 在目标位置保持一段时间hold_steps = 50for _ in range(hold_steps):self.simulator.position_history.append(self.simulator.positions.copy())# 可视化结果self.visualize_results()def visualize_results(self):"""可视化仿真结果"""print("Generating animation...")self.simulator.animate_3d(save_path='swarm_formation_demo.gif')# 绘制性能指标self.plot_performance_metrics()def plot_performance_metrics(self):"""绘制性能指标"""fig, axes = plt.subplots(2, 2, figsize=(12, 8))# 计算指标positions_array = np.array(self.simulator.position_history)num_steps = len(positions_array)# 1. 平均队形误差formation_errors = []for i in range(num_steps):if hasattr(self.simulator, 'target_positions'):errors = np.linalg.norm(positions_array[i] - self.simulator.target_positions, axis=1)formation_errors.append(np.mean(errors))if formation_errors:axes[0, 0].plot(formation_errors)axes[0, 0].set_title('Average Formation Error')axes[0, 0].set_xlabel('Time Step')axes[0, 0].set_ylabel('Error (m)')axes[0, 0].grid(True)# 2. 最小间距min_distances = []for i in range(num_steps):min_dist = float('inf')for j in range(self.num_drones):for k in range(j+1, self.num_drones):dist = np.linalg.norm(positions_array[i, j] - positions_array[i, k])min_dist = min(min_dist, dist)min_distances.append(min_dist)axes[0, 1].plot(min_distances)axes[0, 1].axhline(y=self.transition_planner.safety_distance, color='r', linestyle='--', label='Safety Distance')axes[0, 1].set_title('Minimum Inter-Drone Distance')axes[0, 1].set_xlabel('Time Step')axes[0, 1].set_ylabel('Distance (m)')axes[0, 1].legend()axes[0, 1].grid(True)# 3. 速度分布if num_steps > 1:velocities = np.diff(positions_array, axis=0) / self.simulator.dtavg_velocities = np.mean(np.linalg.norm(velocities, axis=2), axis=1)axes[1, 0].plot(avg_velocities)axes[1, 0].set_title('Average Velocity')axes[1, 0].set_xlabel('Time Step')axes[1, 0].set_ylabel('Velocity (m/s)')axes[1, 0].grid(True)# 4. 能量消耗（简化模型）if num_steps > 2:accelerations = np.diff(velocities, axis=0) / self.simulator.dtenergy = np.sum(np.linalg.norm(accelerations, axis=2) ** 2, axis=1)axes[1, 1].plot(np.cumsum(energy))axes[1, 1].set_title('Cumulative Energy Consumption')axes[1, 1].set_xlabel('Time Step')axes[1, 1].set_ylabel('Energy (arbitrary units)')axes[1, 1].grid(True)plt.tight_layout()plt.savefig('performance_metrics.png')plt.show()

8. 高级特性实现

8.1 避障功能

class ObstacleAvoidance:def __init__(self, obstacle_list):self.obstacles = obstacle_list  # [(center, radius), ...]def compute_avoidance_force(self, drone_pos, max_force=5.0):"""计算避障力"""total_force = np.zeros(3)for obs_center, obs_radius in self.obstacles:# 计算到障碍物的距离diff = drone_pos - obs_centerdistance = np.linalg.norm(diff)# 如果在影响范围内influence_radius = obs_radius + 3.0if distance < influence_radius:# 计算排斥力if distance > 0:force_magnitude = max_force * (1 - distance/influence_radius)**2force_direction = diff / distancetotal_force += force_magnitude * force_directionreturn total_force

8.2 通信约束处理

class CommunicationNetwork:def __init__(self, num_drones, comm_range=20.0):self.num_drones = num_dronesself.comm_range = comm_rangeself.adjacency_matrix = np.zeros((num_drones, num_drones))def update_network(self, positions):"""更新通信网络拓扑"""for i in range(self.num_drones):for j in range(i+1, self.num_drones):distance = np.linalg.norm(positions[i] - positions[j])if distance <= self.comm_range:self.adjacency_matrix[i, j] = 1self.adjacency_matrix[j, i] = 1else:self.adjacency_matrix[i, j] = 0self.adjacency_matrix[j, i] = 0def is_connected(self):"""检查网络连通性"""# 使用深度优先搜索visited = [False] * self.num_dronesself._dfs(0, visited)return all(visited)def _dfs(self, node, visited):visited[node] = Truefor neighbor in range(self.num_drones):if self.adjacency_matrix[node, neighbor] == 1 and not visited[neighbor]:self._dfs(neighbor, visited)

9. 性能优化技巧

9.1 并行计算优化

from multiprocessing import Pool
import numba@numba.jit(nopython=True)
def fast_distance_calculation(pos1, pos2):"""使用Numba加速距离计算"""return np.sqrt(np.sum((pos1 - pos2)**2))class ParallelFormationControl:def __init__(self, num_drones):self.num_drones = num_dronesself.pool = Pool()def parallel_trajectory_planning(self, start_positions, end_positions):"""并行轨迹规划"""tasks = [(start_positions[i], end_positions[i]) for i in range(self.num_drones)]trajectories = self.pool.starmap(self.plan_single_trajectory, tasks)return trajectories@staticmethoddef plan_single_trajectory(start, end):"""单个无人机轨迹规划"""# 实现轨迹规划算法pass

9.2 实时性能监控

class PerformanceMonitor:def __init__(self):self.metrics = {'computation_time': [],'formation_error': [],'communication_load': [],'energy_consumption': []}def record_metric(self, metric_name, value):"""记录性能指标"""if metric_name in self.metrics:self.metrics[metric_name].append(value)def generate_report(self):"""生成性能报告"""report = {}for metric, values in self.metrics.items():if values:report[metric] = {'mean': np.mean(values),'std': np.std(values),'max': np.max(values),'min': np.min(values)}return report