针对弱纹理和变光照环境下基于点特征的视觉SLAM(simultaneous localization and mapping)算法轨迹漂移的问题,提出了一种基于改进自适应阈值ELSED算法(Adaptive-ELSED)的快速点线融合双目视觉SLAM算法。通过在ELSED算法中添加自适应阈...针对弱纹理和变光照环境下基于点特征的视觉SLAM(simultaneous localization and mapping)算法轨迹漂移的问题,提出了一种基于改进自适应阈值ELSED算法(Adaptive-ELSED)的快速点线融合双目视觉SLAM算法。通过在ELSED算法中添加自适应阈值矩阵,动态调整不同光照条件下梯度阈值,并使用长度抑制和短线合并策略,提高线特征的质量。利用基于双目几何约束和图像结构相似性(SSIM)进行快速线段特征三角化。基于历史位姿及误差分析获取初始位姿,通过自适应因子实现光束法平差过程中点线特征的更有效融合。实验结果表明,所提算法在提高线特征质量的同时,耗时仅为LSD算法的50%,线特征匹配速度较传统LBD算法提升67%,挑战性场景下轨迹误差较ORB-SLAM3降低62.2%,系统的平均跟踪帧率为27帧/s,在保证系统实时性的同时,显著提升了系统在弱纹理、变光照环境下的精度和鲁棒性。展开更多
Simultaneous localization and mapping （SLAM） is a key technology for mobile robots operating under unknown environment. While FastSLAM algorithm is a popular solution to the SLAM problem, it suffers from two major d...Simultaneous localization and mapping （SLAM） is a key technology for mobile robots operating under unknown environment. While FastSLAM algorithm is a popular solution to the SLAM problem, it suffers from two major drawbacks： one is particle set degeneracy due to lack of observation information in proposal distribution design of the particle filter; the other is errors accumulation caused by linearization of the nonlinear robot motion model and the nonlinear environment observation model. For the purpose of overcoming the above problems, a new iterated sigma point FastSLAM （ISP-FastSLAM） algorithm is proposed. The main contribution of the algorithm lies in the utilization of iterated sigma point Kalman filter （ISPKF）, which minimizes statistical linearization error through Gaussian-Newton iteration, to design an optimal proposal distribution of the particle filter and to estimate the environment landmarks. On the basis of Rao-Blackwellized particle filter, the proposed ISP-FastSLAM algorithm is comprised by two main parts： in the first part, an iterated sigma point particle filter （ISPPF） to localize the robot is proposed, in which the proposal distribution is accurately estimated by the ISPKF; in the second part, a set of ISPKFs is used to estimate the environment landmarks. The simulation test of the proposed ISP-FastSLAM algorithm compared with FastSLAM2.0 algorithm and Unscented FastSLAM algorithm is carried out, and the performances of the three algorithms are compared. The simulation and comparing results show that the proposed ISP-FastSLAM outperforms other two algorithms both in accuracy and in robustness. The proposed algorithm provides reference for the optimization research of FastSLAM algorithm.展开更多
针对复杂室内环境中视觉同步定位与建图(simultaneous localization and mapping,SLAM)算法在高质量三维重建中的效率问题,提出了一种高效的神经辐射场SLAM(NeRF-SLAM)算法——EN-SLAM。该算法利用多分辨率哈希网格表示场景,结合其快速...针对复杂室内环境中视觉同步定位与建图(simultaneous localization and mapping,SLAM)算法在高质量三维重建中的效率问题,提出了一种高效的神经辐射场SLAM(NeRF-SLAM)算法——EN-SLAM。该算法利用多分辨率哈希网格表示场景,结合其快速收敛特性及高频局部特征表示能力,显著提升了三维重建效率。为进一步增强未观测区域的表面连贯性及细节补全,算法引入球谐函数进行方向编码,从而保证了重建结果的一致性与细节完整性,同时提高实时性。此外,设计了一种信息引导采样策略,优先采样对重建贡献较大的光线,同时实现全局优化(BA)在所有关键帧上的高效执行。在Replica、ScanNet、TUM RGBD和Neural RGB-D数据集上的实验表明,该算法在提高建图精度、跟踪精度及渲染质量的同时,在Replica数据集上的运行时间较iMAP、NICE-SLAM、Vox-Fusion、ESLAM和Co-SLAM分别提升了98.99%、92.80%、91.97%、63.77%和19.15%,且场景重建完成率达到94.14%。展开更多
Geo-monitoring provides quantitative and reliable information to identify hazards and adopt appropriate measures timely.However,this task inherently exposes monitoring staff to hazardous environments,especially in und...Geo-monitoring provides quantitative and reliable information to identify hazards and adopt appropriate measures timely.However,this task inherently exposes monitoring staff to hazardous environments,especially in underground settings.Since 2000s,robots have been widely applied in various fields and many studies have focused on establishing autonomous mobile robotic systems as well as solving the issue of underground navigation and mapping.However,only a few studies have conducted quantitative evaluations of these methods,and almost none have provided a systematic and comprehensive assessment of the suitability of mapping robots for underground geo-monitoring.In this study,a methodology for objective and quantitative assessment of the applicability of SLAM methods in underground geo-monitoring is proposed.This involves the development of an underground test field and some specific metrics,which allow detailed local accuracy analysis of point measurements,line segments,and areas using artificial targets.With this proposed methodology,a series of repeated experimental measurements has been performed with an autonomous driving robot and the selected LiDAR-and visual-based SLAM methods.The resulting point cloud was compared with the reference data measured by a total station and a terrestrial laser scanner.The accuracy and precision of the selected SLAM methods as well as the verifiability and reliability of the results are evaluated and discussed by analysing quantities such as the deviations of the control points coordinates,cloudto-cloud distances between the test and reference point cloud,normal vector,centre point coordinates and area of the planar objects.The results demonstrate that the HDL Graph SLAM achieves satisfactory precision,accuracy,and repeatability with a mean cloud-to-cloud distance of 0.12 m(with a standard deviation of 0.13 m)in an 80 m closed-loop measurement area.Although RTAB-Map exhibits better plane-capturing capabilities,the measurement results reveal instability and inaccuracies.展开更多
The development of wind power clusters has scaled in terms of both scale and coverage,and the impact of weather fluctuations on cluster output changes has become increasingly complex.Accurately identifying the forward...The development of wind power clusters has scaled in terms of both scale and coverage,and the impact of weather fluctuations on cluster output changes has become increasingly complex.Accurately identifying the forward-looking information of key wind farms in a cluster under different weather conditions is an effective method to improve the accuracy of ultrashort-term cluster power forecasting.To this end,this paper proposes a refined modeling method for ultrashort-term wind power cluster forecasting based on a convergent cross-mapping algorithm.From the perspective of causality,key meteorological forecasting factors under different cluster power fluctuation processes were screened,and refined training modeling was performed for different fluctuation processes.First,a wind process description index system and classification model at the wind power cluster level are established to realize the classification of typical fluctuation processes.A meteorological-cluster power causal relationship evaluation model based on the convergent cross-mapping algorithm is pro-posed to screen meteorological forecasting factors under multiple types of typical fluctuation processes.Finally,a refined modeling meth-od for a variety of different typical fluctuation processes is proposed,and the strong causal meteorological forecasting factors of each scenario are used as inputs to realize high-precision modeling and forecasting of ultra-short-term wind cluster power.An example anal-ysis shows that the short-term wind power cluster power forecasting accuracy of the proposed method can reach 88.55%,which is 1.57-7.32%higher than that of traditional methods.展开更多
针对现代化鹅养殖场景中饲料投喂移动小车受动态鹅群干扰,致使同时定位与地图构建(Simultaneous Localization And Mapping,SLAM)算法的定位精度、建图质量下降的问题,提出基于多传感融合目标检测的动态SLAM算法。该算法以LIO-SAM框架...针对现代化鹅养殖场景中饲料投喂移动小车受动态鹅群干扰,致使同时定位与地图构建(Simultaneous Localization And Mapping,SLAM)算法的定位精度、建图质量下降的问题,提出基于多传感融合目标检测的动态SLAM算法。该算法以LIO-SAM框架为基础,融合激光雷达与惯性测量单元搭建SLAM系统,采用前后端架构优化定位与建图性能;运用匈牙利算法实时追踪鹅群运动状态,结合多传感融合目标检测算法,精准识别并剔除动态鹅群产生的特征点,有效降低定位与建图误差。经KITTI、UrbanNav等公共数据集与实际养殖场景数据测试,在KITTI07序列中,较LeGO-LOAM、LIO-SAM和LVI-SAM等经典算法,均方根误差(RMSE)降低33.18%;在实际鹅养殖环境中,可以快速滤除动态鹅群干扰,提升建图质量与导航可靠性。本研究为智能化鹅养殖饲料投喂提供了新的技术方案,推动了畜牧业自动化发展。展开更多
By integrating self-localization,environment mapping,and dynamic object tracking into a unified framework,visual simultaneous localization and mapping with multiple object tracking(SLAMMOT)enhances decision-making and...By integrating self-localization,environment mapping,and dynamic object tracking into a unified framework,visual simultaneous localization and mapping with multiple object tracking(SLAMMOT)enhances decision-making and interaction capabilities in applications such as autonomous driving,robotic navigation,and augmented reality.While numerous outstanding visual SLAMMOT methods have been proposed,the majority rely only on point features,overlooking the abundant and stable planar features in artificial objects that can provide valuable constraints.To address this limitation,we propose OP(object planar)-SLAM,an RGB-D SLAMMOT system that leverages planar features to improve object pose estimation and reconstruction accuracy.Specifically,we introduce an accurate object planar feature extraction and association method using normal images,alongside a novel object bundle adjustment framework that incorporates planar constraints for enhanced optimization.The proposed system is evaluated on both synthetic and public real-world datasets,including Oxford multimotion dataset(OMD)and KITTI tracking dataset.Especially on the OMD,where planar features are prominent,our method improves object pose estimation accuracy by approximately 60%.Extensive experiments demonstrate its effectiveness in enhancing object pose estimation and reconstruction,achieving notable performance compared with existing methods.Furthermore,OP-SLAM runs in real time,making it suitable for practical robots and augmented reality applications.展开更多
With the increasing complexity of substation inspection tasks,achieving efficient and safe path planning for Unmanned Aerial Vehicles in densely populated and structurally complex three-dimensional(3D)environments rem...With the increasing complexity of substation inspection tasks,achieving efficient and safe path planning for Unmanned Aerial Vehicles in densely populated and structurally complex three-dimensional(3D)environments remains a critical challenge.To address this problem,this paper proposes an improved path planning algorithm—Random Geometric Graph(RGG)-guided Rapidly-exploring Random Tree(R-RRT)—based on the classical Rapidly-exploring Random Tree(RRT)framework.First,a refined 3D occupancy grid map is constructed from Light Detection and Ranging point cloud data through ground filtering,noise removal,coordinate transformation,and obstacle inflation using spherical structuring elements.During the planning stage,a dynamic goal-biasing strategy is introduced to adaptively adjust the sampling direction,the sampling distribution is optimized using a pre-generated RGG,and collision detection is accelerated via a K-Dimensional Tree structure.After initial trajectory generation,redundant nodes are eliminated via greedy pruning,and a curvature-minimizing gradient-based optimizationmethod is applied to smooth the trajectory.Experimental results conducted in a simulated substation environment demonstrate that,compared with mainstream path planning algorithms,the proposed R-RRT achieves superior performance in terms of path length,planning time,and trajectory smoothness.Comprehensive analysis shows that the proposed method significantly enhances trajectory quality,planning efficiency,and operational safety,validating its applicability and advantages for high-precision 3D path planning in complex substation inspection scenarios.展开更多
针对动态场景导致视觉定位与建图(simultaneous localization and mapping,SLAM)算法位姿估计精度低和地图质量差等问题,提出一种结合深度学习的动态视觉SLAM算法。该算法在ORB-SLAM3前端引入轻量化且目标识别率高的YOLO11n目标检测网络...针对动态场景导致视觉定位与建图(simultaneous localization and mapping,SLAM)算法位姿估计精度低和地图质量差等问题,提出一种结合深度学习的动态视觉SLAM算法。该算法在ORB-SLAM3前端引入轻量化且目标识别率高的YOLO11n目标检测网络,检测潜在动态区域,并结合Lucas-Kanade(LK)光流法识别其中的动态特征点,从而在剔除动态特征点的同时保留静态特征点,提高特征点利用率和位姿估计精度。此外,新增语义地图构建线程,通过去除YOLO11n识别到的动态物体点云,并融合前端提取的语义信息,实现静态语义地图的构建。在TUM数据集上的实验结果表明,相较于ORB-SLAM3,该算法在高动态序列数据集中的定位精度提升了95.02%,验证了该算法在动态环境下的有效性,能显著提升视觉SLAM系统的定位精度和地图构建质量。展开更多
文摘针对弱纹理和变光照环境下基于点特征的视觉SLAM(simultaneous localization and mapping)算法轨迹漂移的问题,提出了一种基于改进自适应阈值ELSED算法(Adaptive-ELSED)的快速点线融合双目视觉SLAM算法。通过在ELSED算法中添加自适应阈值矩阵,动态调整不同光照条件下梯度阈值,并使用长度抑制和短线合并策略,提高线特征的质量。利用基于双目几何约束和图像结构相似性(SSIM)进行快速线段特征三角化。基于历史位姿及误差分析获取初始位姿,通过自适应因子实现光束法平差过程中点线特征的更有效融合。实验结果表明,所提算法在提高线特征质量的同时,耗时仅为LSD算法的50%,线特征匹配速度较传统LBD算法提升67%,挑战性场景下轨迹误差较ORB-SLAM3降低62.2%,系统的平均跟踪帧率为27帧/s,在保证系统实时性的同时,显著提升了系统在弱纹理、变光照环境下的精度和鲁棒性。
基金supported by Open Foundation of State Key Laboratory of Robotics and System, China (Grant No. SKLRS-2009-ZD-04)National Natural Science Foundation of China (Grant No. 60909055, Grant No.61005070)Fundamental Research Funds for the Central Universities of China (Grant No. 2009JBZ001-2)
文摘Simultaneous localization and mapping （SLAM） is a key technology for mobile robots operating under unknown environment. While FastSLAM algorithm is a popular solution to the SLAM problem, it suffers from two major drawbacks： one is particle set degeneracy due to lack of observation information in proposal distribution design of the particle filter; the other is errors accumulation caused by linearization of the nonlinear robot motion model and the nonlinear environment observation model. For the purpose of overcoming the above problems, a new iterated sigma point FastSLAM （ISP-FastSLAM） algorithm is proposed. The main contribution of the algorithm lies in the utilization of iterated sigma point Kalman filter （ISPKF）, which minimizes statistical linearization error through Gaussian-Newton iteration, to design an optimal proposal distribution of the particle filter and to estimate the environment landmarks. On the basis of Rao-Blackwellized particle filter, the proposed ISP-FastSLAM algorithm is comprised by two main parts： in the first part, an iterated sigma point particle filter （ISPPF） to localize the robot is proposed, in which the proposal distribution is accurately estimated by the ISPKF; in the second part, a set of ISPKFs is used to estimate the environment landmarks. The simulation test of the proposed ISP-FastSLAM algorithm compared with FastSLAM2.0 algorithm and Unscented FastSLAM algorithm is carried out, and the performances of the three algorithms are compared. The simulation and comparing results show that the proposed ISP-FastSLAM outperforms other two algorithms both in accuracy and in robustness. The proposed algorithm provides reference for the optimization research of FastSLAM algorithm.
基金supported by the German Academic Scholarship Foundation,the Deutsche Forschungsgemeinschaft(DFG,German Research Foundation,Project number 422117092)the Saxon Ministry of Science and Arts.
文摘Geo-monitoring provides quantitative and reliable information to identify hazards and adopt appropriate measures timely.However,this task inherently exposes monitoring staff to hazardous environments,especially in underground settings.Since 2000s,robots have been widely applied in various fields and many studies have focused on establishing autonomous mobile robotic systems as well as solving the issue of underground navigation and mapping.However,only a few studies have conducted quantitative evaluations of these methods,and almost none have provided a systematic and comprehensive assessment of the suitability of mapping robots for underground geo-monitoring.In this study,a methodology for objective and quantitative assessment of the applicability of SLAM methods in underground geo-monitoring is proposed.This involves the development of an underground test field and some specific metrics,which allow detailed local accuracy analysis of point measurements,line segments,and areas using artificial targets.With this proposed methodology,a series of repeated experimental measurements has been performed with an autonomous driving robot and the selected LiDAR-and visual-based SLAM methods.The resulting point cloud was compared with the reference data measured by a total station and a terrestrial laser scanner.The accuracy and precision of the selected SLAM methods as well as the verifiability and reliability of the results are evaluated and discussed by analysing quantities such as the deviations of the control points coordinates,cloudto-cloud distances between the test and reference point cloud,normal vector,centre point coordinates and area of the planar objects.The results demonstrate that the HDL Graph SLAM achieves satisfactory precision,accuracy,and repeatability with a mean cloud-to-cloud distance of 0.12 m(with a standard deviation of 0.13 m)in an 80 m closed-loop measurement area.Although RTAB-Map exhibits better plane-capturing capabilities,the measurement results reveal instability and inaccuracies.
基金funded by the State Grid Science and Technology Project“Research on Key Technologies for Prediction and Early Warning of Large-Scale Offshore Wind Power Ramp Events Based on Meteorological Data Enhancement”(4000-202318098A-1-1-ZN).
文摘The development of wind power clusters has scaled in terms of both scale and coverage,and the impact of weather fluctuations on cluster output changes has become increasingly complex.Accurately identifying the forward-looking information of key wind farms in a cluster under different weather conditions is an effective method to improve the accuracy of ultrashort-term cluster power forecasting.To this end,this paper proposes a refined modeling method for ultrashort-term wind power cluster forecasting based on a convergent cross-mapping algorithm.From the perspective of causality,key meteorological forecasting factors under different cluster power fluctuation processes were screened,and refined training modeling was performed for different fluctuation processes.First,a wind process description index system and classification model at the wind power cluster level are established to realize the classification of typical fluctuation processes.A meteorological-cluster power causal relationship evaluation model based on the convergent cross-mapping algorithm is pro-posed to screen meteorological forecasting factors under multiple types of typical fluctuation processes.Finally,a refined modeling meth-od for a variety of different typical fluctuation processes is proposed,and the strong causal meteorological forecasting factors of each scenario are used as inputs to realize high-precision modeling and forecasting of ultra-short-term wind cluster power.An example anal-ysis shows that the short-term wind power cluster power forecasting accuracy of the proposed method can reach 88.55%,which is 1.57-7.32%higher than that of traditional methods.
文摘针对现代化鹅养殖场景中饲料投喂移动小车受动态鹅群干扰,致使同时定位与地图构建(Simultaneous Localization And Mapping,SLAM)算法的定位精度、建图质量下降的问题,提出基于多传感融合目标检测的动态SLAM算法。该算法以LIO-SAM框架为基础,融合激光雷达与惯性测量单元搭建SLAM系统,采用前后端架构优化定位与建图性能;运用匈牙利算法实时追踪鹅群运动状态,结合多传感融合目标检测算法,精准识别并剔除动态鹅群产生的特征点,有效降低定位与建图误差。经KITTI、UrbanNav等公共数据集与实际养殖场景数据测试,在KITTI07序列中,较LeGO-LOAM、LIO-SAM和LVI-SAM等经典算法,均方根误差(RMSE)降低33.18%;在实际鹅养殖环境中,可以快速滤除动态鹅群干扰,提升建图质量与导航可靠性。本研究为智能化鹅养殖饲料投喂提供了新的技术方案,推动了畜牧业自动化发展。
基金Supported by Major Science and Technology Project of Hubei Province(2022AAA009)。
文摘By integrating self-localization,environment mapping,and dynamic object tracking into a unified framework,visual simultaneous localization and mapping with multiple object tracking(SLAMMOT)enhances decision-making and interaction capabilities in applications such as autonomous driving,robotic navigation,and augmented reality.While numerous outstanding visual SLAMMOT methods have been proposed,the majority rely only on point features,overlooking the abundant and stable planar features in artificial objects that can provide valuable constraints.To address this limitation,we propose OP(object planar)-SLAM,an RGB-D SLAMMOT system that leverages planar features to improve object pose estimation and reconstruction accuracy.Specifically,we introduce an accurate object planar feature extraction and association method using normal images,alongside a novel object bundle adjustment framework that incorporates planar constraints for enhanced optimization.The proposed system is evaluated on both synthetic and public real-world datasets,including Oxford multimotion dataset(OMD)and KITTI tracking dataset.Especially on the OMD,where planar features are prominent,our method improves object pose estimation accuracy by approximately 60%.Extensive experiments demonstrate its effectiveness in enhancing object pose estimation and reconstruction,achieving notable performance compared with existing methods.Furthermore,OP-SLAM runs in real time,making it suitable for practical robots and augmented reality applications.
基金Funding for this research was provided by the Program for Scientific Research Innovation Team in Colleges and Universities of Anhui Province(No.2022AH010095)the Hefei Key Technology R&D“Champion-Based Selection”Project(No.2023SGJ011).
文摘With the increasing complexity of substation inspection tasks,achieving efficient and safe path planning for Unmanned Aerial Vehicles in densely populated and structurally complex three-dimensional(3D)environments remains a critical challenge.To address this problem,this paper proposes an improved path planning algorithm—Random Geometric Graph(RGG)-guided Rapidly-exploring Random Tree(R-RRT)—based on the classical Rapidly-exploring Random Tree(RRT)framework.First,a refined 3D occupancy grid map is constructed from Light Detection and Ranging point cloud data through ground filtering,noise removal,coordinate transformation,and obstacle inflation using spherical structuring elements.During the planning stage,a dynamic goal-biasing strategy is introduced to adaptively adjust the sampling direction,the sampling distribution is optimized using a pre-generated RGG,and collision detection is accelerated via a K-Dimensional Tree structure.After initial trajectory generation,redundant nodes are eliminated via greedy pruning,and a curvature-minimizing gradient-based optimizationmethod is applied to smooth the trajectory.Experimental results conducted in a simulated substation environment demonstrate that,compared with mainstream path planning algorithms,the proposed R-RRT achieves superior performance in terms of path length,planning time,and trajectory smoothness.Comprehensive analysis shows that the proposed method significantly enhances trajectory quality,planning efficiency,and operational safety,validating its applicability and advantages for high-precision 3D path planning in complex substation inspection scenarios.
文摘针对动态场景导致视觉定位与建图(simultaneous localization and mapping,SLAM)算法位姿估计精度低和地图质量差等问题,提出一种结合深度学习的动态视觉SLAM算法。该算法在ORB-SLAM3前端引入轻量化且目标识别率高的YOLO11n目标检测网络,检测潜在动态区域,并结合Lucas-Kanade(LK)光流法识别其中的动态特征点,从而在剔除动态特征点的同时保留静态特征点,提高特征点利用率和位姿估计精度。此外,新增语义地图构建线程,通过去除YOLO11n识别到的动态物体点云,并融合前端提取的语义信息,实现静态语义地图的构建。在TUM数据集上的实验结果表明,相较于ORB-SLAM3,该算法在高动态序列数据集中的定位精度提升了95.02%,验证了该算法在动态环境下的有效性,能显著提升视觉SLAM系统的定位精度和地图构建质量。
