【无人机路径规划】基于视觉 SLAM 算法的粒子群优化无人机路径规划附matlab复现

最新推荐文章于 2024-10-08 17:52:16 发布

Matlab科研辅导帮

最新推荐文章于 2024-10-08 17:52:16 发布

阅读量1k

点赞数 17

分类专栏：无人机文章标签：无人机算法 matlab

本文链接：https://blog.csdn.net/m0_60703264/article/details/136954687

版权

无人机专栏收录该内容

226 篇文章 33 订阅

订阅专栏

本文介绍了一种结合视觉同步定位与建图（VSLAM）的无人机路径规划方法，利用粒子群优化（PSO）和区域敏感性策略，设计了一个能避开危险区域并优化飞行风险、能量消耗的路径规划器。动态适应性函数确保了路径的有效性和安全性，仿真结果验证了该系统的有效性。

摘要由CSDN通过智能技术生成

✅作者简介：热爱科研的Matlab仿真开发者，修心和技术同步精进，代码获取、论文复现及科研仿真合作可私信。

🍎个人主页：Matlab科研工作室

🍊个人信条：格物致知。

更多Matlab完整代码及仿真定制内容点击👇

智能优化算法神经网络预测雷达通信无线传感器电力系统

信号处理图像处理路径规划元胞自动机无人机

物理应用机器学习

🔥 内容介绍

我们使用最先进的视觉同步定位和建图（VSLAM）方法来追踪无人机位姿，同时逐步构建周围环境的增量地图。在这方面，首先使用单目视觉方法来绘制感兴趣区域的地图。构建的地图被处理为优化算法的输入手段，以规划多架无人机的最佳路径。我们设计了一个基于粒子群优化（PSO）的路径规划器，并提出了一个基于区域敏感性（RS）的路径更新机制，以避免在执行最终路径时检测到危险事件时的敏感区域。此外，我们提出了一个动态适应度函数（DFF），以评估路径规划器的规划策略，同时考虑各种优化参数，如飞行风险估计、能量消耗和操作完成时间。所提出的规划器获得了较高的适应度值，并安全地到达目的地，同时遵循最短路径，避免所有意外的危险事件和限制区域，这验证了我们提出的 PSO-VSLAM 系统的有效性，如仿真结果所示。**避开了所有意外的危险事件和限制区域，这验证了我们提出的 PSO-VSLAM 系统的有效性，如仿真结果所示。

📣 部分代码

function [isLoopClosed, mapPoints, vSetKeyFrames] = helperAddLoopConnections(...    mapPoints, vSetKeyFrames, loopCandidates, currKeyFrameId, currFeatures, currPoints, intrinsics)%helperAddLoopConnections add connections between the current key frame and %   the valid loop candidate key frames. A loop candidate is valid if it has %   enough covisible map points with the current key frame.%   This is an example helper function that is subject to change or removal %   in future releases.%   Copyright 2019 The MathWorks, Inc.scaleFactor     = 1.2;imageSize       = [480, 640];% Minimum number of matched features for loop edgeminNumMatches   = 60;numCandidates   = size(loopCandidates,1);loopConnections = [];[index3d1, index2d1] = getProjectionIndexPair(mapPoints, currKeyFrameId);validFeatures1  = currFeatures.Features(index2d1, :);validPoints1    = currPoints(index2d1).Location;for k = numCandidates : -1 : 1    [index3d2, index2d2] = getProjectionIndexPair(mapPoints, loopCandidates(k));    allFeatures2 = vSetKeyFrames.Views.Features{loopCandidates(k)};    validFeatures2 = allFeatures2(index2d2, :);    allPoints2 = vSetKeyFrames.Views.Points{loopCandidates(k)};    validPoints2 = allPoints2(index2d2);        indexPairs = matchFeatures(binaryFeatures(validFeatures1), binaryFeatures(validFeatures2), ...        'Unique', true, 'MaxRatio', 0.9, 'MatchThreshold', 90);        % Check if all the candidate key frames have strong connection with the    % current keyframe    if size(indexPairs, 1) < minNumMatches        isLoopClosed = false;        return    end        % Estimate the relative pose of the current key frame with respect to the    % loop candidate keyframe with the highest similarity score    if k == 1         worldPoints = mapPoints.Locations(index3d2(indexPairs(:,2)),:);                matchedImagePoints = cast(validPoints1(indexPairs(:,1),:), 'like', worldPoints);        [worldOrientation, worldLocation] = estimateWorldCameraPose(matchedImagePoints, worldPoints, intrinsics, ...            'Confidence', 90, 'MaxReprojectionError', 6, 'MaxNumTrials', 1e4);        cameraPose = rigid3d(worldOrientation, worldLocation);        [R, t]    = cameraPoseToExtrinsics(cameraPose.Rotation, cameraPose.Translation);        xyzPoints = mapPoints.Locations(index3d2,:);        projectedPoints = worldToImage(intrinsics, R, t, xyzPoints);                isInImage = find(projectedPoints(:,1)<imageSize(2) & projectedPoints(:,1)>0 & ...            projectedPoints(:,2)< imageSize(1) & projectedPoints(:,2)>0);                minScales    = validPoints2.Scale(isInImage)/scaleFactor;        maxScales    = validPoints2.Scale(isInImage)*scaleFactor;        r            = 3;        searchRadius = r*validPoints2.Scale(isInImage);                matchedIndexPairs = helperMatchFeaturesInRadius(validFeatures2(isInImage,:), currFeatures.Features, ...            validPoints2(isInImage), currPoints, projectedPoints(isInImage,:), searchRadius, minScales, maxScales);        matchedIndexPairs(:,1) = isInImage(matchedIndexPairs(:,1));                visiblePointsIndex = index3d2(matchedIndexPairs(:,1));        validWorldPoints   = mapPoints.Locations(visiblePointsIndex, :);        matchedImagePoints = currPoints.Location(matchedIndexPairs(:,2),:);                % Refine the pose        cameraPose = bundleAdjustmentMotion(validWorldPoints, matchedImagePoints, ...            cameraPose, intrinsics, 'PointsUndistorted', true, 'AbsoluteTolerance', 1e-7,...            'RelativeTolerance', 1e-15, 'MaxIteration', 50);                % Fuse covisible map points        [matchedIndex2d1, ia1, ib1] = intersect(index2d1, matchedIndexPairs(:,2), 'stable');        matchedIndex3d1 = index3d1(ia1);        matchedIndex3d2 = index3d2(matchedIndexPairs(ib1,1));        matchedIndex2d2 = index2d2(matchedIndexPairs(ib1,1));                mapPoints = updateLocation(mapPoints, mapPoints.Locations(matchedIndex3d2, :), matchedIndex3d1);                % Add connection between the current key frame and the loop key frame        pose1   = vSetKeyFrames.Views.AbsolutePose(loopCandidates(k));        pose2   = cameraPose;        relPose = rigid3d(pose2.Rotation*pose1.Rotation', (pose2.Translation-pose1.Translation)*pose1.Rotation');        matches = [matchedIndex2d2, matchedIndex2d1];        vSetKeyFrames = addConnection(vSetKeyFrames, loopCandidates(k), currKeyFrameId, relPose, 'Matches', matches);        disp(['Loop edge added between keyframe: ', num2str(loopCandidates(k)), ' and ', num2str(currKeyFrameId)]);                % Add connections between the current key frame and the connected         % key frames of the loop key frame        neighborViews = connectedViews(vSetKeyFrames, loopCandidates(k));        for m = 1:numel(neighborViews.ViewId)            neighborViewId = neighborViews.ViewId(m);            [index3d3, index2d3] = getProjectionIndexPair(mapPoints, neighborViewId);            [covPointsIndices, ia2, ib2] = intersect(index3d3, matchedIndex3d2, 'stable');             if numel(covPointsIndices) > minNumMatches                pose1   = neighborViews.AbsolutePose(m);                pose2   = cameraPose;                relPose = rigid3d(pose2.Rotation*pose1.Rotation', (pose2.Translation-pose1.Translation)*pose1.Rotation');                 matches = [index2d3(ia2), matchedIndex2d1(ib2)];                if ~hasConnection(vSetKeyFrames, neighborViewId, currKeyFrameId)                    vSetKeyFrames = addConnection(vSetKeyFrames, neighborViewId, currKeyFrameId, relPose, 'Matches', matches);                end                disp(['Loop edge added between keyframe: ', num2str(neighborViewId), ' and ', num2str(currKeyFrameId)]);            end        end                isLoopClosed = true;    endendend

⛳️ 运行结果

🔗 参考文献

[1] Mughal U A , Ahmad I , Pawase C J ,et al.UAVs Path Planning by Particle Swarm Optimization Based on Visual-SLAM Algorithm[J]. 2022.DOI:10.1007/978-981-19-1292-4_8.