[{"data":1,"prerenderedAt":213},["ShallowReactive",2],{"method-cai2021ikdtree":3},{"method":4,"reference":48,"equipment":66,"figures":82,"results":122},{"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":22,"limitations":25,"sensors":27,"platform":30,"estimator":35,"association":36,"timeModel":37,"deskew":37,"loopClosure":37,"globalOptimization":38,"mapRepresentation":39,"prior":38,"outputGeometry":40,"compute":41,"codeUrl":42,"codeLicense":43,"relatedVersions":44},"cai2021ikdtree","Cai et al., 2021","ikd-Tree","ikd-Tree: An Incremental K-D Tree for Robotic Applications",2021,"recent","C12","map_representation_or_reconstruction","ikd-Tree 讓 k-d 樹只以新進點增量更新，支援單點與方盒範圍的插入、重新插入與刪除（刪除採延遲標記），並在樹上同步降採樣：以邊長 L 的立方格劃分空間，每格只保留最接近格心的點。樹以類似 scapegoat 樹的準則監測平衡並局部重建，大型子樹可交由第二執行緒重建，以維持即時性。","An incremental k-d tree supporting point- and box-wise insert\u002Fdelete, on-tree voxel downsampling (keeping the point nearest each cell center), and partial rebuilding with optional parallel threads.","full_text_reviewed","preprint","main_body","未在營建場域驗證；唯一實測為香港大學本部大樓的戶外 LiDAR 慣性建圖（Fig. 7）。樹上降採樣在每個邊長 L 的立方格只保留最接近格心的點（FAST-LIO 測試中 L 為 0.2 m，Table II），因此決定 FAST-LIO2 類系統輸出地圖的點密度上限（推論）。",[20,21],"simulation","controlled_experiment",[23,24],"[\"On randomized data, incremental update time stayed around 1.6 ms while a static kd-tree rebuild grew linearly with points (Sec. V-A, Fig. 4).\", \"In FAST-LIO the per-scan fusion time stayed nearly constant around 1.6 ms (average of the last 100 scans), enabling mapping up to 100 Hz versus 10 Hz for the original static-tree FAST-LIO","the static tree exceeded 10 ms per scan from 366 s onward (Sec. V-B, Fig. 6(a)).\", \"Authors report about two orders of magnitude higher efficiency in their conclusion, while the abstract states ikd-Tree used only 4% of the static k-d tree running time (Sec. VI, Abstract).\"]",[26],"[\"Insertion of spatially compact data is slower because rebuilds are triggered more often (Sec. V-A, Fig. 5).\", \"k-nearest search was slightly slower than the PCL static k-d tree in the randomized test (Sec. V-A, Fig. 4(b)).\", \"Space complexity is O(n) but with a constant a few times larger than a static k-d tree (Sec. IV-B).\", \"Box-wise operations use boxes aligned with the data coordinate axes (Sec. III-C).\", \"(inference) The on-tree downsampling keeps one point per cell, so the stored map is a resolution-limited subset of raw measurements rather than all points.\"]",[28,29],"3D LiDAR","IMU (in the FAST-LIO application test)",[31,32,33,34],"[\"not_verified (outdoor Livox Avia data","Fig. 7 shows the mapping result of the Main Building, University of Hong Kong","onboard computer DJI Manifold 2-C4","carrier not stated)\"]","not_applicable (data structure; tested inside FAST-LIO)","exact (not approximate) k-nearest-neighbour search that uses per-node range bounds and lazy-label pushdown (Sec. III-E)","not_applicable","none","incremental k-d tree of map points with lazy-delete labels and on-tree voxel downsampling","downsampled point set stored in the tree","CPU only: randomized tests on a PC with Intel i7-10700 (2.90 GHz, only 2 threads running); FAST-LIO test on a DJI Manifold 2-C4 (quad-core Intel i7-8550U 1.8 GHz, 8 GB RAM); optional second thread rebuilds sub-trees larger than Nmax = 1500","https:\u002F\u002Fgithub.com\u002Fhku-mars\u002Fikd-Tree","GPL-2.0 (repository LICENSE)",[45],{"relation":46,"title":47,"doi_or_url":42},"code_release","hku-mars\u002Fikd-Tree",{"id":5,"kind":49,"shortName":7,"title":8,"authors":50,"year":9,"venue":54,"venueType":16,"publisher":55,"volumeIssuePages":56,"doi":57,"arxivId":58,"url":59,"firstPublicDate":60,"publicationStatus":16,"metadataStatus":61,"fulltextStatus":15,"era":10,"classicReason":37,"codeUrl":42,"cluster":11,"topics":62,"mdpi":63,"verification":64,"label":6,"fulltextRoute":55,"versionRead":65,"addedByCensus":63},"component",[51,52,53],"Yixi Cai","Wei Xu","Fu Zhang","arXiv preprint","arXiv","not_reported",null,"2102.10808","https:\u002F\u002Farxiv.org\u002Fabs\u002F2102.10808","2021-02-22","metadata_verified",[11],false,"corrected","arXiv 2102.10808v1 (22 Feb 2021, 8 pp., only version; no journal reference on the abs page)",[67,73,78],{"category":68,"model":69,"canonical":69,"role":70,"dataset":57,"specs":71,"locator":72},"lidar","Livox Avia","method input","70 degree FoV; frame rate 100 Hz","Sec. V-B",{"category":74,"model":75,"canonical":75,"role":76,"dataset":57,"specs":77,"locator":72},"compute","DJI Manifold 2-C4","compute for runtime","1.8 GHz quad-core Intel i7-8550U CPU, 8 GB RAM",{"category":74,"model":79,"canonical":79,"role":76,"dataset":57,"specs":80,"locator":81},"Intel i7-10700","PC, 2.90 GHz, only 2 threads running","Sec. V-A",[83,96,104,114],{"refId":5,"refLabel":6,"fig":84,"whatZh":85,"license":86,"licenseUrl":87,"sourceUrl":88,"src":89,"width":90,"height":91,"thumb":92,"thumbWidth":93,"thumbHeight":94,"modified":95},"Fig. 1","增量式 k-d 樹插入新點與局部重新平衡的立體示意","CC BY 4.0","https:\u002F\u002Fcreativecommons.org\u002Flicenses\u002Fby\u002F4.0\u002F","https:\u002F\u002Farxiv.org\u002Fhtml\u002F2102.10808v1\u002Fkd_tree_space.png","\u002Ffigure-files\u002Fcai2021ikdtree\u002Ffig-1.webp",1400,732,"\u002Ffigure-files\u002Fcai2021ikdtree\u002Ffig-1.thumb.webp",480,251,"resized to at most 1400 px wide and converted to WebP",{"refId":5,"refLabel":6,"fig":97,"whatZh":98,"license":86,"licenseUrl":87,"sourceUrl":99,"src":100,"width":90,"height":101,"thumb":102,"thumbWidth":93,"thumbHeight":103,"modified":95},"Fig. 2","樹上降採樣前後的點雲對照","https:\u002F\u002Farxiv.org\u002Fhtml\u002F2102.10808v1\u002Fdownsample.png","\u002Ffigure-files\u002Fcai2021ikdtree\u002Ffig-2.webp",691,"\u002Ffigure-files\u002Fcai2021ikdtree\u002Ffig-2.thumb.webp",237,{"refId":5,"refLabel":6,"fig":105,"whatZh":106,"license":86,"licenseUrl":87,"sourceUrl":107,"src":108,"width":109,"height":110,"thumb":111,"thumbWidth":93,"thumbHeight":112,"modified":113},"Fig. 6","FAST-LIO 中 ikd-Tree 與靜態 k-d 樹融合一筆新掃描的時間、時間分解與平衡指標","https:\u002F\u002Farxiv.org\u002Fhtml\u002F2102.10808v1\u002Ffastlio_exp_combine_1.png","\u002Ffigure-files\u002Fcai2021ikdtree\u002Ffig-6.webp",790,710,"\u002Ffigure-files\u002Fcai2021ikdtree\u002Ffig-6.thumb.webp",431,"converted to WebP",{"refId":5,"refLabel":6,"fig":115,"whatZh":116,"license":86,"licenseUrl":87,"sourceUrl":117,"src":118,"width":90,"height":119,"thumb":120,"thumbWidth":93,"thumbHeight":121,"modified":95},"Fig. 7","香港大學本部大樓 100 Hz 建圖成果與 LiDAR 軌跡","https:\u002F\u002Farxiv.org\u002Fhtml\u002F2102.10808v1\u002Fmapping_result_3.png","\u002Ffigure-files\u002Fcai2021ikdtree\u002Ffig-7.webp",519,"\u002Ffigure-files\u002Fcai2021ikdtree\u002Ffig-7.thumb.webp",178,{"totalRows":123,"groupCount":124,"groups":125,"others":212},4,2,[126,188],{"slug":127,"group":128,"sourceId":5,"sourceLabel":6,"table":129,"selfRows":130,"metrics":131,"seqs":141,"entrants":146,"cells":158,"outcomes":173,"locators":178,"hardware":181,"wordings":183,"notes":184},"cai2021ikdtree-text-sec-v-b","cai2021ikdtree:Text Sec. V-B","Text Sec. V-B",3,[132,136,138],{"label":133,"unit":134,"statistic":135,"alignment":56},"average running time of fusing one new LiDAR scan","ms","mean",{"label":137,"unit":134,"statistic":135,"alignment":56},"average time of incremental updates per scan",{"label":139,"unit":140,"statistic":56,"alignment":56},"mapping rate","Hz",[142],{"dataset":143,"sequence":144,"environment":145},"authors' outdoor recording","Main Building, University of Hong Kong","outdoor campus",[147,150,153,154,156],{"name":148,"methodId":5,"linkable":149,"proposed":149,"self":149},"FAST-LIO with ikd-Tree",true,{"name":151,"methodId":152,"linkable":149,"proposed":63,"self":63},"FAST-LIO with static K-D tree","fastlio2021",{"name":7,"methodId":5,"linkable":149,"proposed":149,"self":149},{"name":155,"methodId":57,"linkable":63,"proposed":63,"self":63},"static k-d tree",{"name":157,"methodId":152,"linkable":149,"proposed":63,"self":63},"original FAST-LIO with static k-d tree",[159,163,165,167,169,171],[160,160,160,161,160,160,160,162,160],0,1.6,-1,[164,160,160,57,164,160,160,162,160],1,[124,164,160,166,162,164,160,162,164],0.23,[130,164,160,168,162,164,160,162,164],5.71,[160,124,160,170,124,124,160,162,124],100,[123,124,160,172,130,160,162,162,124],10,[174,175,176,177],"stated as nearly constant around 1.6 ms","increases roughly linearly and exceeds 10 ms from 366 s onward","stated as enabling a mapping rate up to 100 Hz","value quoted from the original FAST-LIO work, not re-measured here",[72,179,180],"Sec. V-B, Fig. 6(b)","Sec. V-B, Fig. 7",[182],"DJI Manifold 2-C4 (Intel i7-8550U, 8 GB RAM)",[],[185,186,187],"FAST-LIO on a real outdoor scene with a Livox Avia; time to fuse one new LiDAR scan averaged over the most recent 100 scans; ikd-Tree replaces the static PCL k-d tree for build, update and query; downsampling cube 0.2 m","FAST-LIO on a real outdoor scene with a Livox Avia; breakdown of k-d tree related operations per scan","Achievable mapping rate stated by the authors",{"slug":189,"group":190,"sourceId":5,"sourceLabel":6,"table":191,"selfRows":164,"metrics":192,"seqs":195,"entrants":200,"cells":202,"outcomes":204,"locators":206,"hardware":207,"wordings":209,"notes":210},"cai2021ikdtree-text-sec-v-a","cai2021ikdtree:Text Sec. V-A","Text Sec. V-A",[193],{"label":194,"unit":134,"statistic":56,"alignment":56},"time for incremental updates per test operation, including re-building",[196],{"dataset":197,"sequence":198,"environment":199},"randomized synthetic points","incremental test (5,000 to about 200,000 points)","synthetic",[201],{"name":7,"methodId":5,"linkable":149,"proposed":149,"self":149},[203],[160,160,160,161,160,160,160,162,160],[205],"stated as remaining stably around 1.6 ms",[81],[208],"PC, Intel i7-10700 2.90 GHz, 2 threads",[],[211],"Randomized test: 5,000 initial points in a 10 m cube, 1,000 operations of 200-point insertions and 5-NN queries, periodic box-wise deletes and 2,000-point insertions; tree grows to about 200,000 points; static tree is the PCL k-d tree rebuilt each operation; no downsampling",[],1790510665100]