✅作者简介：热爱科研的Matlab仿真开发者，修心和技术同步精进，matlab项目合作可私信。

🍎个人主页：Matlab科研工作室

🍊个人信条：格物致知。

⛄ 内容介绍

本文研究了无线网络中无人驾驶飞行器(UAV)的最佳放置问题。无人机作为一个飞行的无线中继器，为基站(BS)提供覆盖范围的扩展，并为被障碍物遮挡的用户提供容量提升。虽然现有的方法依赖于直接传播链路的潜在阻塞的统计模型，但我们提出了一种能够利用当地地形信息来提供性能保证的方法。所提出的方法允许在最小化到地面终端的传播距离和发现良好的传播条件之间进行最佳权衡。该算法只需要几个传播参数，但它能够避免深度传播阴影，并被证明能够找到全球最佳的无人机位置。只需要对目标区域进行局部探索，而且搜索轨迹的最大长度与地理尺度成线性关系。因此，它适合于在线搜索。与其他基于统计传播模型的定位方法相比，发现有明显的吞吐量提高。

⛄ 部分代码

% Massive simulation

clear

addpath(genpath('lib')),

Nue = 10000;    % <- reduce this number to shorter simulation time (coarser results)

DATA = load('citymap/urbanMapSingleUserK2.mat');

U = DATA.U; PosBS = DATA.PosBS; 

DATA = load('citymap/losStatistics.mat');

losStat.Plos = DATA.Plos;

losStat.ElvAngles = DATA.ElvAngles;

clear DATA

load('citymap/topologyK2.mat');

U.K = 2;

if U.K == 2

    U.Alpha = [-21.4, -30.3];

    U.Beta =[-36.92, -38.42];

elseif U.K == 3

    U.Alpha = [-22, -28, -36];

    U.Beta =[-28, -24, -22];

else

    error('K should be 2 or 3.');

end

U.A0 = -20.8; U.B0 = -38.5;

U.A1 = U.Alpha(1); U.B1 = U.Beta(1); 

U.A2 = U.Alpha(2); U.B2 = U.Beta(2);

Noise_dBm = -80;

Power_BS_dBm = 33;

Power_UAV_dBm = 33;

U.Noise = 10^(Noise_dBm/10) / 1000; % Watt in linear scale

U.Pb = 10^(Power_BS_dBm/10) / 1000; 

U.Pd = 10^(Power_UAV_dBm/10) / 1000; 

U.Hbs = 45;     % meter, BS height

U.Hmin = 45;    % meter, minimum UAV operation height

U.Hdrone = 50;  % meter, UAV search height

stepSizeMeter = 5;  % UAV search step size

fun = @(x,y) max(-log2(1 + U.Pd * real(x)), -log2(1 + U.Pb * real(y)));

fun0 = @(x) -log2(1 + U.Pb * x);

% Ergodic capacity

SNRs_dB = -10:2:20; Ks_dB = [9, -Inf];

Rerg = capacity_ergodic(Ks_dB, SNRs_dB);

fun1 = @(x,y) max(- max(0, ppval(spline(SNRs_dB, Rerg(1, :)), 10 * log10(U.Pd * real(x)))), ... 

                  - log2(1 + U.Pb * real(y))); % UAV-UE_LOS(K-factor = 9dB, 

fun2 = @(x,y) max(- max(0, ppval(spline(SNRs_dB, Rerg(2, :)), 10 * log10(U.Pd * real(x)))), ... 

                  - log2(1 + U.Pb * real(y))); % UAV-UE_NLOS, Rayleigh fading 

%%

N_scheme = 6;

tic

Nue = min(size(Topology, 1), Nue);

Rates0 = zeros(Nue, N_scheme);

strongUserIds = zeros(Nue, 1);

failIds = zeros(Nue, 1);

parfor i = 1:Nue

    PosUE = Topology{i}.PosUE; 

    Blds = Topology{i}.Blds;

    BldTypes = Topology{i}.BldTypes;

    BldLines = Topology{i}.BldLines; 

    BldHeight = Topology{i}.BldHeight; 

    Nbld = size(Blds, 1);

    los = IsLosK(PosUE, [PosBS, U.Hbs], BldLines, BldHeight, U.Hdrone, BldTypes);

    if los == 1

        strongUserIds(i) = 1;

        % continue    % We are only interested in the case where the direct BS-user link is blocked

    end

    urbanMap = struct();

    urbanMap.BldLines = BldLines;

    urbanMap.BldHeight = BldHeight;

    urbanMap.BldTypes = BldTypes;

    % Direct BS-user link

    k = round((1 - los) * (U.K - 1) + 1);   % propagation segment index

    d = norm([PosBS, U.Hbs] - [PosUE, 0], 2);

    snr = 10 ^ ((U.Alpha(k) * log10(d) + U.Beta(k)) / 10) / U.Noise;

    F0 = fun0(snr);

    try

        % [Fmin3, Xhat3] = finduavpos3d(PosUE, PosBS, U, fun, stepSizeMeter, urbanMap);

        % Fmin3 = min(Fmin3, F0);

        Fmin3 = 0;

        [~, Xhat2] = finduavpos(PosUE, PosBS, U, fun, stepSizeMeter, urbanMap);

        los = IsLosK(PosUE, [Xhat2, U.Hdrone], BldLines, BldHeight, U.Hdrone, BldTypes);

        Fmin2 = getcostf2DK_ergodic([Xhat2, U.Hdrone], [PosUE, 0], [PosBS, U.Hbs], los, U, fun1, fun2);

        % Fmin2 = min(Fmin2, F0);

        [~, Xhat1] = finduavpos1d(PosUE, PosBS, U, fun, stepSizeMeter, urbanMap);

        los = IsLosK(PosUE, [Xhat1, U.Hdrone], BldLines, BldHeight, U.Hdrone, BldTypes);

        Fmin1 = getcostf2DK_ergodic([Xhat1, U.Hdrone], [PosUE, 0], [PosBS, U.Hbs], los, U, fun1, fun2);

        % Fmin1 = min(Fmin1, F0);

        % [Fmin_exhst, Xhat_exhst] = finduavpos2d_exhst(PosUE, PosBS, U, fun, stepSizeMeter, urbanMap);

        Fmin_exhst = Fmin3;

        [~, XhatStat] = finduavposStat(PosUE, PosBS, U, fun, stepSizeMeter, urbanMap, losStat);

        los = IsLosK(PosUE, [XhatStat, U.Hdrone], BldLines, BldHeight, U.Hdrone, BldTypes);

        FminStat = getcostf2DK_ergodic([XhatStat, U.Hdrone], [PosUE, 0], [PosBS, U.Hbs], los, U, fun1, fun2);

        % FminStat = min(FminStat, F0);

    catch

        Fmin1 = 0;

        Fmin2 = 0;

        Fmin3 = 0;

        Fmin_exhst = 0;

        FminStat = 0;

        failIds(i) = 1;

    end

    Rates0(i, :) = - [F0, FminStat, Fmin1, Fmin2, Fmin3, Fmin_exhst];

end

toc

%% Plot results

my_line_styles = {'-', '--', '-.', ':'}.';

Alg_scheme_name = {

    'Direct BS-User linkxx'

    'Probabilistic Alg'

    'Simple Search'

    'Proposed'

    'Proposed (3D)'

    'Exhaustive'

};

schemes_to_show = [1 2 3 4 6];

N_scheme_to_show = length(schemes_to_show);

validUserId = failIds < 1;

Rates = Rates0(validUserId, :);

Nue = size(Rates, 1);

maxdata = max(Rates(:));

Npt = 40;

XI = sort([0.1 0.17 0.3 0.5 (0:1/(Npt - 1 - 4):1) * maxdata], 'ascend');

X_data = zeros(Npt, N_scheme_to_show);

F_data = zeros(Npt, N_scheme_to_show);

for i = 1:N_scheme_to_show

    n = schemes_to_show(i);

    r_vec = Rates(:, n);

    [F1,X1] = ksdensity(r_vec, XI, 'function', 'cdf');

    X_data(:, i) = X1(:);

    F_data(:, i) = F1;

end

figure(1),

h = plot(X_data, F_data,'linewidth', 2);

set(gca, 'FontSize', 14);

legend(Alg_scheme_name{schemes_to_show}, 'location', 'southeast');

xlim([0 ceil(max(Rates(:)))]);

set(gca, 'YTick', 0:0.2:1);

xlabel('bps/Hz');

ylabel('CDF');

tune_figure,

set(h(1), 'linewidth', 2);

set(h(1), 'Marker', '*', 'Markersize', 6);

set(h(1), 'LineStyle', ':');

set(h(2), 'LineStyle', '-.');

set(h(3), 'LineStyle', ':');

set(h(4), 'LineStyle', '-');

set(h(4), 'LineWidth', 3);

set(h(5), 'linestyle', '--');

set(h(5), 'LineWidth', 3);

% ----

schemes_to_show = [1 2 3 4];

figure(2),

rateNoUav = Rates(:, 1);

[~, sortedIndex] = sort(rateNoUav, 'ascend');

low20percentileIndex = sortedIndex(1:round(Nue * 0.2));

high20percentileIndex = sortedIndex(round(Nue * 0.8): end);

RateLow = mean(Rates(low20percentileIndex, schemes_to_show), 1);

RateMean = mean(Rates(:, schemes_to_show), 1);

RateHigh = mean(Rates(high20percentileIndex, schemes_to_show), 1);

h = bar([RateLow

         RateMean

         RateHigh]);

set(gca, 'FontSize', 14);

set(h, 'linewidth', 2);

ylim([0, 10]);

legend(Alg_scheme_name{schemes_to_show}, 'location', 'northwest');

set(gca, 'XTickLabel', {'20th percentile', 'Mean', 'Top 20th percentile'});

set(gca, 'YTick', 0:2:10);

ylabel('Average end-to-end throughput [bps/Hz]');

% label the bars

Xdata = [RateLow

         RateMean

         RateHigh];

bartext = [];

for i = 1:size(Xdata, 1)

    for j = 1:size(Xdata, 2)

        bartext(i, j) = text(i + (j - 2.5) * 0.18, Xdata(i, j) + 0.05, ...

            sprintf('%1.2f', Xdata(i, j)), 'fontsize', 12);

    end

end

% Use the handles TH to modify some properties

set(bartext,'Horizontalalignment','center',...

'verticalalignment','bottom') ;

tune_figure,

[im_hatch,colorlist] = applyhatch_pluscolor(gcf,'\-x./+',0,0,[],150,2,2);

⛄ 运行结果

【中继优化】基于高效局部地图搜索算法实现无人机通信中继位置优化附matlab代码和复现论文_无人机

【中继优化】基于高效局部地图搜索算法实现无人机通信中继位置优化附matlab代码和复现论文_Power_02

⛄ 参考文献

[1] Chen J ,  Gesbert D . Efficient Local Map Search Algorithms for the Placement of Flying Relays:, 10.48550/arXiv.1801.03595[P]. 2018.

⛳️ 代码获取关注我

❤️部分理论引用网络文献，若有侵权联系博主删除
❤️ 关注我领取海量matlab电子书和数学建模资料