无人机三维路径规划算法对比：蚁群、A*与RRT*的 Matlab 实战

数据解读：改进蚁群算法能耗较低（64.08 J），因路径平滑减少姿态调整；A*在风险路径检测上表现最优，得益于精确栅格碰撞检测。

四、应用场景推荐

• 电力巡检（静态精细）：推荐 A*，路径最短且能耗最低。
• 城市物流（动态密集）：推荐 RRT*，实时避障能力强。
• 农业植保（大范围非结构）：推荐改进蚁群，适应复杂地形且能耗低。
• 实时搜救（未知环境）：推荐 RRT* + 蚁群混合，快速生成初始路径并优化全局安全。

五、Matlab 代码实现

理论分析之后，我们来看具体的 Matlab 实现逻辑。主函数负责加载地图、设置起点终点，并依次调用三个算法模块进行求解与可视化。

% Wishing you to encourage yourself！ 
clc; clear; close all 
% 载入函数路径
addpath(genpath('./ACO3D'))
addpath(genpath('./Astar3D'))
addpath(genpath('./RRT3D'))
addpath(genpath('./Evaluation'))

Algorithm_name = {'ACO', 'Astar', 'RRT'};
map = Makemap3D; % 地图规模在 Makemap3D 函数中设置
source = [10, 10, 1]; % 起点
goal = [450, 400, 50]; % 终点
max_item = 1000; % 最大迭代次数
comparative_data = {};
Global_data = {};
Straight_distance = sqrt(sum((source - goal).^2, 2)); % 直线距离
fprintf('起点到终点的直线距离：%d\n\n', Straight_distance);

Global_data(1, end+1) = {num2str(source)};
Global_data(1, end+1) = {num2str(goal)};
Global_data(1, end+1) = {Straight_distance};
t1 = clock;

%% ********蚁群算法********************************
figure(1)
plot3DMap(map);
text(source(1), source(2), source(3), '起点', 'color', 'r');
text(goal(1), goal(2), goal(3), '终点', 'color', 'r');
scatter3(source(1), source(2), source(3), "filled", "g");
scatter3(goal(1), goal(2), goal(3), "filled", "b");
title('ACO');
popNum = 10; % 蚁群数量
tic
[aco_path, aco_cost, aco_Number_of_searches, ...]
    = aco(source, goal, map, popNum); % 蚁群主函数
aco_time = toc;
fprintf('ACO 历时：%0.3f 秒\n', aco_time);
hold on
plot3(aco_path(:,1), aco_path(:,2), aco_path(:,3), 'LineWidth', 2, 'color', 'r');
view(-30, 30);
fprintf('ACO 路径长度：%d \n\n', aco_cost);
[aco_max_turning_angle, aco_turning_num, aco_index] = Max_turning_angle(aco_path, 1);
comparative_data(1, end+1) = {aco_time};
comparative_data(2, end) = {aco_cost};
comparative_data(3, end) = {size(aco_path, 1)};
comparative_data(4, end) = {aco_Number_of_searches};
comparative_data(5, end) = {aco_Number_of_successful_searches};
comparative_data(6, end) = {aco_Number_of_successful_searches / aco_Number_of_searches};
comparative_data(7, end) = {aco_max_turning_angle};
comparative_data(8, end) = {aco_turning_num};

%% ************Astar********************************
figure(2)
plot3DMap(map);
text(source(1), source(2), source(3), '起点', 'color', 'r');
text(goal(1), goal(2), goal(3), '终点', 'color', 'r');
scatter3(source(1), source(2), source(3), "filled", "g");
scatter3(goal(1), goal(2), goal(3), "filled", "b");
title('Astar');
tic % 计算运行时间
[astar_path, astar_cost, astar_Number_of_searches, ...]
    = Astar_main(source, goal, map, max_item); % A*主函数
astar_time = toc;
fprintf('Astar 历时：%0.3f 秒\n', astar_time);
plot3(astar_path(:,1), astar_path(:,2), astar_path(:,3), 'LineWidth', 2, 'color', 'r');
view(-30, 30);
fprintf('Astar 路径长度：%d\n\n', astar_cost);
[astar_max_turning_angle, astar_turning_num, astar_index] = Max_turning_angle(astar_path, 1);
comparative_data(1, end+1) = {astar_time};
comparative_data(2, end) = {astar_cost};
comparative_data(3, end) = {size(astar_path, 1)};
comparative_data(4, end) = {astar_Number_of_searches};
comparative_data(5, end) = {astar_Number_of_successful_searches};
comparative_data(6, end) = {astar_Number_of_successful_searches / astar_Number_of_searches};
comparative_data(7, end) = {astar_max_turning_angle};
comparative_data(8, end) = {astar_turning_num};

%% ****************RRT********************************
figure(3)
plot3DMap(map);
text(source(1), source(2), source(3), '起点', 'color', 'r');
text(goal(1), goal(2), goal(3), '终点', 'color', 'r');
scatter3(source(1), source(2), source(3), "filled", "g");
scatter3(goal(1), goal(2), goal(3), "filled", "b");
title('RRT');
step = 10; % 设置步长
tic
[rrt_path, rrt_cost, rrt_Number_of_searches, ...]
    = RRT_main(source, goal, map, step); % RRT 主函数
rrt_time = toc;
fprintf('RRT 历时：%0.3f 秒\n', rrt_time);
plot3(rrt_path(:,1), rrt_path(:,2), rrt_path(:,3), 'LineWidth', 2, 'color', 'r');
view(-30, 30);
fprintf('RRT 路径长度：%d \n\n', rrt_cost);
[rrt_max_turning_angle, rrt_turning_num, rrt_index] = Max_turning_angle(rrt_path, 1);
comparative_data(1, end+1) = {toc};
comparative_data(2, end) = {rrt_cost};
comparative_data(3, end) = {size(rrt_path, 1)};
comparative_data(4, end) = {rrt_Number_of_searches};
comparative_data(5, end) = {rrt_Number_of_successful_searches};
comparative_data(6, end) = {rrt_Number_of_successful_searches / rrt_Number_of_searches};
comparative_data(7, end) = {rrt_max_turning_angle};
comparative_data(8, end) = {rrt_turning_num};

%% 打印运行时间
t2 = clock;
fprintf('程序总运行时间：%0.3f 秒\n\n', etime(t2, t1));

% 计算最优项，展示表格顺序为 aco、astar、rrt
comparative_data(1, end+1) = {Algorithm_name{Min_value(comparative_data{1,1}, comparative_data{1,2}, comparative_data{1,3})}};
for i = 2:size(comparative_data, 1)
    comparative_data(i, end) = {Algorithm_name{Min_value(comparative_data{i,1}, comparative_data{i,2}, comparative_data{i,3})}};
end
comparative_data(6, end) = {Algorithm_name{Max_value(comparative_data{6,1}, comparative_data{6,2}, comparative_data{6,3})}};
Show_Comparative_result(Global_data, comparative_data);

% 三种算法展示到一张图中
figure(4)
plot3DMap(map);
text(source(1), source(2), source(3), '起点', 'color', 'r');
text(goal(1), goal(2), goal(3), '终点', 'color', 'r');
h1 = plot3(aco_path(:,1), aco_path(:,2), aco_path(:,3), 'LineWidth', 2, 'color', 'r');
h2 = plot3(astar_path(:,1), astar_path(:,2), astar_path(:,3), 'LineWidth', 2, 'color', 'k');
h3 = plot3(rrt_path(:,1), rrt_path(:,2), rrt_path(:,3), 'LineWidth', 2, 'color', 'm');
legend([h1, h2, h3], '蚁群', 'A*', 'RRT')

六、总结与展望

算法融合是未来的主要趋势。例如 A结合贝塞尔曲线可解决平滑问题，RRT结合梯度优化能提升初始路径质量。在实际工程中，建议根据场景约束（如实时性要求、硬件算力）综合权衡选型。未来还可探索轻量化神经网络加速在线规划，或利用强化学习优化多目标权重。

参考文献

[1] 陆璐，孙昱浩，孙启光。基于改进蚁群算法的无人机三维路径规划研究 [J].广东交通职业技术学院学报，2025,24(01):46-51+114. [2] 温佳霖，梁丰，许雯。基于改进蚁群算法的无人机三维路径规划研究 [J].现代信息科技，2023,7(20):84-87+91.DOI:10.19850/j.cnki.2096-4706.2023.20.018.