[{"data":1,"prerenderedAt":312},["ShallowReactive",2],{"method-brossard2020icpcov":3},{"method":4,"reference":47,"equipment":66,"figures":74,"results":75},{"id":5,"label":6,"shortName":7,"title":8,"year":9,"era":10,"cluster":11,"scope":12,"keyIdeaZh":13,"keyIdeaEn":14,"fulltextStatus":15,"publicationStatus":16,"recommendation":17,"constructionRelevance":18,"validationEnvironment":19,"strengths":21,"limitations":25,"sensors":30,"platform":32,"estimator":34,"association":35,"timeModel":36,"deskew":18,"loopClosure":36,"globalOptimization":37,"mapRepresentation":36,"prior":38,"outputGeometry":39,"compute":40,"codeUrl":41,"codeLicense":42,"relatedVersions":43},"brossard2020icpcov","Brossard et al., 2020","3D ICP covariance (unscented)","A New Approach to 3D ICP Covariance Estimation",2020,"recent","C13","registration_component","作者主張 ICP 結果的不確定性取決於初始值（通常來自里程計）的不確定性，因此以無跡轉換（unscented transform）額外執行 12 次 ICP 配準來傳遞初始化不確定性，並輸出含初始值與 ICP 結果相關項的聯合共變異數；感測器白雜訊與所有點共有的校正偏差（各假設約 5 cm）則以封閉式公式另計。在 Challenging data sets 八個序列、1020 組配準上，所提方法的 NNE 為平移 4.2、旋轉 34，而 Censi 公式與 65 次 Monte Carlo 為 10^2 至 10^3 量級；在軌跡一致性上平均也優於未考慮相關項的組合，但仍略為樂觀。","Treats ICP covariance as conditional on initialization uncertainty (propagated with an unscented transform) and adds a term for sensor noise and calibration bias.","full_text_reviewed","peer_reviewed_published","background","not_reported",[20],"public_benchmark",[22,23,24],"Averaged over 8 sequences, NNE of 4.2 (translation) and 34 (rotation) versus about 10^3 for Censi's formula and 10^3 and 10^2 for 65-sample Monte Carlo; robust NNE 0.8 and 3.8 (Table 1)","Best average trajectory consistency (Mahalanobis distance) among CELLO-3D, ini.+ICP and the proposed covariance, with a clear gain on Apartment (2.3 and 9.8 vs 3.5 and 15) (Table 2)","Deterministic, needs only 12 extra ICP runs, and flags wrong ICP convergence through very large covariances (Sec. III-B, Sec. V-B)",[26,27,28,29],"The closed-form part is not valid for point-to-point ICP (Sec. VI)","Relies on a Gaussian error assumption and cannot describe non-Gaussian ICP error distributions (Sec. V-B, V-C)","Requires the initialization covariance as input; with local minima the output inherits an optimistic or pessimistic Q_ini (Sec. V-B, V-C, Table 3)","Trajectory covariances remain slightly optimistic (Table 2 caption)",[31],"3D laser scans from a Hokuyo sensor (Challenging data sets for point cloud registration, Pomerleau et al. 2012; model number not given in the paper)",[33],"public dataset only: Challenging data sets for point cloud registration (8 sequences of 31 to 45 scans, 268 scans, 1020 registrations; structured to unstructured, indoor to outdoor); carrying platform not described in the paper","unscented transform over initialization uncertainty plus closed-form sensor-noise term; point-to-plane ICP","ICP configured as in Pomerleau et al. (2013): 95% random subsampling, kd-tree data association, point-to-plane error metric, 70% closest associations kept for outlier rejection","not_applicable","none","initialization covariance Q_ini assumed known (main tests: 0.1 m and 10 deg standard deviation, 'easy'; robustness tests with 0.5 m and 20 deg, and 1 m and 50 deg); sensor white noise and common bias each with 5 cm standard deviation","registration covariance","No processor stated. Sec. IV-C reports that the 12 extra ICP registrations of the unscented transform take 6 s when run in parallel and the remaining steps less than 0.1 s; the 65-run Monte Carlo baseline is more than five times as demanding. Sec. VI calls the method real time; real-time use depends on implementation","https:\u002F\u002Fgithub.com\u002FCAOR-MINES-ParisTech\u002F3d-icp-cov","MIT",[44],{"relation":45,"title":8,"doi_or_url":46},"preprint","https:\u002F\u002Farxiv.org\u002Fabs\u002F1909.05722",{"id":5,"kind":48,"shortName":7,"title":8,"authors":49,"year":9,"venue":53,"venueType":54,"publisher":55,"volumeIssuePages":56,"doi":57,"arxivId":58,"url":46,"firstPublicDate":59,"publicationStatus":16,"metadataStatus":60,"fulltextStatus":15,"era":10,"classicReason":36,"codeUrl":41,"cluster":11,"topics":61,"mdpi":62,"verification":63,"label":6,"fulltextRoute":64,"versionRead":65,"addedByCensus":62},"component",[50,51,52],"Martin Brossard","Silvere Bonnabel","Axel Barrau","IEEE Robotics and Automation Letters","journal","IEEE","5(2), pp. 744-751","10.1109\u002Flra.2020.2965391","1909.05722","2019-09-12","metadata_verified",[11],false,"corrected","arXiv","arXiv 1909.05722v2 (27 Jan 2020, 8 pp., accepted RA-L preprint) read in full; IEEE RA-L 5(2):744-751 version of record (via NTU) compared: Tables I and II checked as rendered images and match v2",[67],{"category":68,"model":69,"canonical":69,"role":70,"dataset":71,"specs":72,"locator":73},"lidar","Hokuyo sensor (model number not given in the paper)","dataset sensor","Challenging data sets for point cloud registration algorithms (Pomerleau et al., IJRR 2012)","white noise and bias standard deviation taken as 5 cm, the mean value reported for this sensor by Pomerleau et al. (CARPI 2012)","Sec. IV-B",[],{"totalRows":76,"groupCount":77,"groups":78,"others":311},26,3,[79,204,285],{"slug":80,"group":81,"sourceId":5,"sourceLabel":6,"table":82,"selfRows":83,"metrics":84,"seqs":91,"entrants":110,"cells":120,"outcomes":197,"locators":199,"hardware":200,"wordings":201,"notes":202},"brossard2020icpcov-table-2","brossard2020icpcov:Table 2","Table 2",16,[85,89],{"label":86,"unit":87,"statistic":88,"alignment":37},"Mah. dist. trans.","unitless","mean",{"label":90,"unit":87,"statistic":88,"alignment":37},"Mah. dist. rot.",[92,96,98,100,102,104,106,108],{"dataset":93,"sequence":94,"environment":95},"Challenging data sets for point cloud registration (Pomerleau et al. 2012)","Apartment","per-sequence type not stated in the paper",{"dataset":93,"sequence":97,"environment":95},"Hauptgebaude",{"dataset":93,"sequence":99,"environment":95},"Stairs",{"dataset":93,"sequence":101,"environment":95},"Mountain",{"dataset":93,"sequence":103,"environment":95},"Gazebo summer",{"dataset":93,"sequence":105,"environment":95},"Gazebo winter",{"dataset":93,"sequence":107,"environment":95},"Wood summer",{"dataset":93,"sequence":109,"environment":95},"Wood winter",[111,115,118],{"name":112,"methodId":113,"linkable":114,"proposed":62,"self":62},"CELLO-3D","landry2019cello3d",true,{"name":116,"methodId":117,"linkable":62,"proposed":62,"self":62},"ini.+ICP (fusion without cross-covariance)",null,{"name":119,"methodId":5,"linkable":114,"proposed":114,"self":114},"proposed (full ML covariance, Eq. 15)",[121,125,128,130,131,133,134,135,136,138,139,141,142,144,145,147,148,150,152,154,156,158,160,162,164,165,167,168,170,171,173,174,176,178,180,182,184,185,186,187,188,189,190,191,192,193,195,196],[122,122,122,123,124,122,124,124,122],0,0.2,-1,[122,126,122,127,124,122,124,124,122],1,0.1,[122,122,126,129,124,122,124,124,122],0.3,[122,126,126,123,124,122,124,124,122],[122,122,132,127,124,122,124,124,122],2,[122,126,132,123,124,122,124,124,122],[122,122,77,117,122,122,124,124,122],[122,126,77,117,122,122,124,124,122],[122,122,137,123,124,122,124,124,122],4,[122,126,137,123,124,122,124,124,122],[122,122,140,127,124,122,124,124,122],5,[122,126,140,123,124,122,124,124,122],[122,122,143,127,124,122,124,124,122],6,[122,126,143,129,124,122,124,124,122],[122,122,146,127,124,122,124,124,122],7,[122,126,146,129,124,122,124,124,122],[126,122,122,149,124,122,124,124,122],3.5,[126,126,122,151,124,122,124,124,122],15,[126,122,126,153,124,122,124,124,122],1.9,[126,126,126,155,124,122,124,124,122],3.2,[126,122,132,157,124,122,124,124,122],1.1,[126,126,132,159,124,122,124,124,122],4.2,[126,122,77,161,124,122,124,124,122],1.5,[126,126,77,163,124,122,124,124,122],1.2,[126,122,137,157,124,122,124,124,122],[126,126,137,166,124,122,124,124,122],2.1,[126,122,140,153,124,122,124,124,122],[126,126,140,169,124,122,124,124,122],3.7,[126,122,143,161,124,122,124,124,122],[126,126,143,172,124,122,124,124,122],4.6,[126,122,146,163,124,122,124,124,122],[126,126,146,175,124,122,124,124,122],4.8,[132,122,122,177,124,122,124,124,122],2.3,[132,126,122,179,124,122,124,124,122],9.8,[132,122,126,181,124,122,124,124,122],1.8,[132,126,126,183,124,122,124,124,122],2.9,[132,122,132,157,124,122,124,124,122],[132,126,132,159,124,122,124,124,122],[132,122,77,163,124,122,124,124,122],[132,126,77,163,124,122,124,124,122],[132,122,137,126,124,122,124,124,122],[132,126,137,177,124,122,124,124,122],[132,122,140,181,124,122,124,124,122],[132,126,140,169,124,122,124,124,122],[132,122,143,161,124,122,124,124,122],[132,126,143,194,124,122,124,124,122],4.7,[132,122,146,163,124,122,124,124,122],[132,126,146,159,124,122,124,124,122],[198],"not_run (Mountain not considered in the CELLO-3D paper)",[82],[],[],[203],"Trajectory consistency: Mahalanobis distance of compounded ICP trajectories to ground truth, averaged over 40 initial trajectories per sequence; target 1, below 1 pessimistic; CELLO-3D reproduced from Landry et al. with a slightly different ICP setting",{"slug":205,"group":206,"sourceId":5,"sourceLabel":6,"table":207,"selfRows":208,"metrics":209,"seqs":226,"entrants":230,"cells":238,"outcomes":275,"locators":280,"hardware":281,"wordings":282,"notes":283},"brossard2020icpcov-table-1","brossard2020icpcov:Table 1","Table 1",8,[210,212,214,216,218,220,222,224],{"label":211,"unit":87,"statistic":88,"alignment":37},"NNE trans.",{"label":213,"unit":87,"statistic":88,"alignment":37},"NNE rot.",{"label":215,"unit":87,"statistic":88,"alignment":37},"KL div. trans.",{"label":217,"unit":87,"statistic":88,"alignment":37},"KL div. rot.",{"label":219,"unit":87,"statistic":88,"alignment":37},"NNE* trans. (robust)",{"label":221,"unit":87,"statistic":88,"alignment":37},"NNE* rot. (robust)",{"label":223,"unit":87,"statistic":88,"alignment":37},"KL div.* trans. (robust)",{"label":225,"unit":87,"statistic":88,"alignment":37},"KL div.* rot. (robust)",[227],{"dataset":93,"sequence":228,"environment":229},"average of 8 sequences","structured to unstructured, indoor to outdoor",[231,234,236],{"name":232,"methodId":233,"linkable":114,"proposed":62,"self":62},"Q_censi (closed-form, Censi 2007)","censi2007covariance",{"name":235,"methodId":117,"linkable":62,"proposed":62,"self":62},"Q_monte carlo (65 Monte Carlo ICP samples)",{"name":237,"methodId":5,"linkable":114,"proposed":114,"self":114},"proposed",[239,241,242,244,246,248,250,251,252,253,254,255,256,258,260,261,262,263,265,266,267,269,271,273],[122,122,122,240,122,122,124,124,122],1000,[122,126,122,240,122,122,124,124,122],[122,132,122,243,126,122,124,124,122],10000,[122,77,122,245,132,122,124,124,122],100000,[122,137,122,247,124,122,124,124,122],38,[122,140,122,249,77,122,124,124,122],100,[122,143,122,240,122,122,124,124,122],[122,146,122,245,132,122,124,124,122],[126,122,122,240,122,122,124,124,122],[126,126,122,249,77,122,124,124,122],[126,132,122,243,126,122,124,124,122],[126,77,122,243,126,122,124,124,122],[126,137,122,257,124,122,124,124,122],22,[126,140,122,259,124,122,124,124,122],20,[126,143,122,240,122,122,124,124,122],[126,146,122,240,122,122,124,124,122],[132,122,122,159,124,122,124,124,122],[132,126,122,264,124,122,124,124,122],34,[132,132,122,249,77,122,124,124,122],[132,77,122,249,77,122,124,124,122],[132,137,122,268,124,122,124,124,122],0.8,[132,140,122,270,124,122,124,124,122],3.8,[132,143,122,272,124,122,124,124,122],31,[132,146,122,274,124,122,124,124,122],98,[276,277,278,279],"printed only as order of magnitude 10^3","printed only as order of magnitude 10^4","printed only as order of magnitude 10^5","printed only as order of magnitude 10^2",[207],[],[],[284],"ICP covariance consistency averaged over 8 sequences, 1020 registrations x 1000 initializations; Q_ini 0.1 m and 10 deg; noise and bias SD 5 cm; NNE target 1; starred = robust statistics after removing extreme quantiles; baselines printed as orders of magnitude",{"slug":286,"group":287,"sourceId":5,"sourceLabel":6,"table":288,"selfRows":132,"metrics":289,"seqs":295,"entrants":298,"cells":300,"outcomes":303,"locators":305,"hardware":307,"wordings":308,"notes":309},"brossard2020icpcov-text-sec-iv-c","brossard2020icpcov:Text Sec. IV-C","Text Sec. IV-C",[290,293],{"label":291,"unit":292,"statistic":18,"alignment":18},"time for the 12 unscented-transform ICP registrations, computed in parallel","s",{"label":294,"unit":292,"statistic":18,"alignment":18},"time for the remaining steps of the algorithm",[296],{"dataset":93,"sequence":297,"environment":36},"per registration pair",[299],{"name":237,"methodId":5,"linkable":114,"proposed":114,"self":114},[301,302],[122,122,122,143,124,122,124,124,122],[122,126,122,127,122,122,124,124,122],[304],"stated as less than 0.1 s",[306],"Sec. IV-C",[],[],[310],"Execution time of the covariance computation per registration pair (Algorithm 1)",[],1790510664576]