[{"data":1,"prerenderedAt":501},["ShallowReactive",2],{"method-lemon2026":3},{"method":4,"reference":60,"equipment":86,"figures":113,"results":114},{"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":24,"limitations":31,"sensors":38,"platform":40,"estimator":43,"association":44,"timeModel":45,"deskew":46,"loopClosure":47,"globalOptimization":48,"mapRepresentation":49,"prior":50,"outputGeometry":51,"compute":52,"codeUrl":53,"codeLicense":54,"relatedVersions":55},"lemon2026","Wang et al., 2026","LEMON-Mapping","LEMON-Mapping: Loop-Enhanced Large-Scale Multi-Session Point Cloud Merging and Optimization for Globally Consistent Mapping",2026,"recent","C06","offline_map_refinement","LEMON-Mapping 指出傳統多機位姿圖最佳化只把迴圈當作位姿節點間約束，忽略地圖幾何，導致重疊區發散與模糊。其框架包含：迴圈處理模組（剔除離群、分類迴圈並召回被誤刪的正確迴圈）；對多機地圖做空間 BA（孤立迴圈用 DBA、成群迴圈用 HBA）以消除重疊區不一致；再以兩步驟位姿圖最佳化，把 BA 精修後的局部精度傳遞到整張地圖。","LEMON-Mapping augments multi-session PGO with robust loop processing and spatial bundle adjustment in overlapping regions, then propagates BA-refined accuracy through a two-step PGO.","full_text_reviewed","peer_reviewed_published","main_body","無工地資料。公開資料含 S3E Tunnel、GEODE Tunnelingtunnel 與 Stairs 等隧道與樓梯序列；自行蒐集資料（Mid360 光達）含車庫（室內）、圖書館、庭院、實驗室與飛行場（Sec. VII-A、Table I 至 II）。地圖品質以 DJI L1 光達經 DJI Terra 處理的高精度點雲作為 MARS-LVIG 參考地圖，AWD 0.25 至 0.65 m、CD 0.43 至 1.11 m（Sec. VII-D、Table V）；S3E 平面厚度 0.08 m（Table VI）。這些誤差等級遠大於工地尺寸檢核常用的公差（推論）。",[20,21,22,23],"public_benchmark","underground_or_tunnel","independent_reference","cross_site",[25,26,27,28,29,30],"Merged all 10 multi-robot sequences of Table IV, whereas LAMM failed on 4 and DCL-SLAM on 6 (failure = ATE RMSE above 30 m) (Sec. VII-C, Table IV)","Map quality against DJI L1 reference maps on MARS-LVIG Island, Town and Airport: AWD 0.25 m to 0.65 m and CD 0.43 m to 1.11 m, lower than LAMM on every metric (Sec. VII-D, Table V)","S3E plane thickness 0.08 m and planarity 0.86 versus 0.12 m and 0.61 (DCL-SLAM) and 0.14 m and 0.59 (LAMM) (Table VI)","Loop outlier rejection F1 higher than best-tuned PCM and GNC on three of four S3E sequences and comparable on Tunnel (Table VIII)","Spatial BA gave the lowest z-drift and z-RMSE on all four self-collected single-robot scenes versus BALM2 and HBA (Table III)","100% merging success with 5, 10 and 20 sessions of R3LIVE data (Table X)",[32,33,34,35,36,37],"Information-matrix weights are heuristic; principled covariance estimation left to future work (Sec. VIII)","Runtime grows with the number of loop closures; multi-threaded parallel windows left to future work (Sec. VII-H, VIII)","Spatial BA without the last PGO can increase ATE because it breaks odometry continuity (Sec. VII-F, Table IX)","Loop rejection precision and recall are computed against the framework's own optimized trajectory with a 5 m threshold, not independent labels (Sec. VII-E)","Single-robot study has no ground truth; uses z-drift relative to the first frame and MME, and MME was not best on Garage or Laboratory (Sec. VII-B, Table III)","Hardware for runtime and memory results not reported (Sec. VII-H)",[39],"3D LiDAR: Velodyne (S3E), Avia and Ouster (GEODE), Avia (MARS-LVIG, R3LIVE), Mid360 (self-collected) (Tables I-II)",[41,42],"not described in the paper: public multi-session datasets S3E, GEODE, MARS-LVIG and R3LIVE (platforms described only in the dataset papers)","self-collected Livox Mid360 dataset (Garage, Library, Yard, Laboratory, Flying Arena); carrier not stated","loop processing (outlier rejection, classification, recall) then spatial BA (DBA for isolated loops, HBA for clustered loops) and two-step pose-graph optimization","RING++ provides all intra- and inter-robot loop candidates; statistical outlier removal, GICP initialized by RING++ and RANSAC correspondence rejection with an inlier count and GICP fitness test verify loops; BFS region growing labels loops as clustered or isolated; rejected loops are recalled if their poses are within 2 m after the first PGO; spatial BA windows are spherical radius searches in a pose kd-tree around each loop; RING++ similarity (inter-robot) and a registration-based minimum-eigenvalue test (intra-robot) select sparse BA constraints for the last PGO (Sec. IV, V, VI-B).","discrete poses","not_reported (inputs from LiDAR-inertial odometry)","robust loop outlier rejection with recall of wrongly removed loops","window-based spatial BA in overlaps plus two-step PGO propagating local accuracy","multi-session point-cloud submaps","multi-agent odometry and submaps; relative transforms between sessions unknown","merged globally consistent point-cloud map and trajectories","Offline, C++ with ROS; hardware not reported. Runtime is governed by the number of loop closures (Dormitory, 62 loops, about 70 s); within LEMON-Mapping the spatial HBA stage for clustered loops takes about 67% to 80% and cluster preprocessing (PCA reordering and GICP) about 17% to 28% of runtime, isolated DBA and PGO each under 5%; peak memory stays bounded and below the HBA baseline, whereas a full BALM2 pipeline needs 30 GiB to 50 GiB even after aggregating f frames per submap (e.g., f = 30) (Sec. VII-H, Fig. 18).",null,"not_applicable",[56],{"relation":57,"title":58,"doi_or_url":59},"preprint","arXiv 2505.10018 (v4, 2026-06-10)","https:\u002F\u002Farxiv.org\u002Fabs\u002F2505.10018",{"id":5,"kind":61,"shortName":7,"title":8,"authors":62,"year":9,"venue":73,"venueType":74,"publisher":75,"volumeIssuePages":76,"doi":77,"arxivId":78,"url":59,"firstPublicDate":79,"publicationStatus":16,"metadataStatus":80,"fulltextStatus":15,"era":10,"classicReason":54,"codeUrl":53,"cluster":11,"topics":81,"mdpi":82,"verification":83,"label":6,"fulltextRoute":84,"versionRead":85,"addedByCensus":82},"method",[63,64,65,66,67,68,69,70,71,72],"Lijie Wang","Xiaoyi Zhong","Ziyi Xu","Kaixin Chai","Anke Zhao","Tianyu Zhao","Changjian Jiang","Qianhao Wang","Xieyuanli Chen","Fei Gao","IEEE Transactions on Automation Science and Engineering","journal","IEEE","23:12318-12336","10.1109\u002Ftase.2026.3709653","2505.10018","2025-05-15","metadata_verified",[11],false,"confirmed","arXiv","arXiv 2505.10018v4 (2026-06-10), author version; IEEE T-ASE version of record (DOI 10.1109\u002FTASE.2026.3709653) not compared",[87,95,100,104,107],{"category":88,"model":89,"canonical":90,"role":91,"dataset":92,"specs":93,"locator":94},"lidar","Mid360","Livox MID-360","method input","self-collected (Garage, Library, Yard, Laboratory, Flying Arena)","not_reported","Table I",{"category":88,"model":96,"canonical":96,"role":97,"dataset":98,"specs":93,"locator":99},"Velodyne (model not named)","dataset sensor","S3E","Table II",{"category":88,"model":101,"canonical":102,"role":97,"dataset":103,"specs":93,"locator":99},"Avia","Livox Avia","GEODE (Inlandwaterways, Tunnelingtunnel); MARS-LVIG; R3LIVE",{"category":88,"model":105,"canonical":105,"role":97,"dataset":106,"specs":93,"locator":99},"Ouster (model not named)","GEODE (Stairs, Offroad)",{"category":88,"model":108,"canonical":108,"role":109,"dataset":110,"specs":111,"locator":112},"DJI L1 LiDAR sensor","reference or ground truth","MARS-LVIG","high-precision point cloud processed with DJI Terra","Sec. VII-D",[],{"totalRows":115,"groupCount":116,"groups":117,"others":484},50,7,[118,242,322,414],{"slug":119,"group":120,"sourceId":5,"sourceLabel":6,"table":121,"selfRows":122,"metrics":123,"seqs":135,"entrants":149,"cells":159,"outcomes":236,"locators":237,"hardware":238,"wordings":239,"notes":240},"lemon2026-table-iii","lemon2026:Table III","Table III",12,[124,129,132],{"label":125,"unit":126,"statistic":127,"alignment":128},"MME","unitless","mean","none",{"label":130,"unit":93,"statistic":127,"alignment":131},"z-Drift","first-pose",{"label":133,"unit":93,"statistic":134,"alignment":131},"z-RMSE","RMSE",[136,140,143,146],{"dataset":137,"sequence":138,"environment":139},"self-collected (Mid360)","Garage","indoor, 334 m",{"dataset":137,"sequence":141,"environment":142},"Library","outdoor, 519 m",{"dataset":137,"sequence":144,"environment":145},"Yard","indoor and outdoor, 232 m",{"dataset":137,"sequence":147,"environment":148},"Laboratory","outdoor, 98 m",[150,154,157],{"name":151,"methodId":152,"linkable":153,"proposed":82,"self":82},"HBA","hba2023",true,{"name":155,"methodId":156,"linkable":153,"proposed":82,"self":82},"BALM2","balm2_2023",{"name":158,"methodId":5,"linkable":153,"proposed":153,"self":153},"Ours (spatial BA)",[160,164,167,170,172,174,176,178,180,182,184,186,188,190,192,194,196,198,200,202,204,206,208,210,212,214,216,218,221,223,225,227,229,231,232,234],[161,161,161,162,163,161,163,163,161],0,-6.83,-1,[161,165,161,166,163,161,163,163,161],1,5.98,[161,168,161,169,163,161,163,163,161],2,7.14,[165,161,161,171,163,161,163,163,161],-6.93,[165,165,161,173,163,161,163,163,161],5.95,[165,168,161,175,163,161,163,163,161],7.17,[168,161,161,177,163,161,163,163,161],-6.86,[168,165,161,179,163,161,163,163,161],4.87,[168,168,161,181,163,161,163,163,161],5.35,[161,161,165,183,163,161,163,163,161],-5.95,[161,165,165,185,163,161,163,163,161],6.17,[161,168,165,187,163,161,163,163,161],6.78,[165,161,165,189,163,161,163,163,161],-6.2,[165,165,165,191,163,161,163,163,161],6.51,[165,168,165,193,163,161,163,163,161],7.31,[168,161,165,195,163,161,163,163,161],-6.23,[168,165,165,197,163,161,163,163,161],3.68,[168,168,165,199,163,161,163,163,161],4.53,[161,161,168,201,163,161,163,163,161],-6.39,[161,165,168,203,163,161,163,163,161],2.33,[161,168,168,205,163,161,163,163,161],2.7,[165,161,168,207,163,161,163,163,161],-5.93,[165,165,168,209,163,161,163,163,161],1.86,[165,168,168,211,163,161,163,163,161],2.12,[168,161,168,213,163,161,163,163,161],-6.4,[168,165,168,215,163,161,163,163,161],1.12,[168,168,168,217,163,161,163,163,161],1.25,[161,161,219,220,163,161,163,163,161],3,-6.26,[161,165,219,222,163,161,163,163,161],3.24,[161,168,219,224,163,161,163,163,161],3.72,[165,161,219,226,163,161,163,163,161],-6.04,[165,165,219,228,163,161,163,163,161],3.22,[165,168,219,230,163,161,163,163,161],3.7,[168,161,219,189,163,161,163,163,161],[168,165,219,233,163,161,163,163,161],3.17,[168,168,219,235,163,161,163,163,161],3.64,[],[121],[],[],[241],"Single-robot study on self-collected Mid360 data; spatial BA versus BALM2 (sliding window) and HBA, all on raw odometry without loop-based refinement; z-drift and z-RMSE relative to the z-value of the first frame; MME via MapEval (lower is better)",{"slug":243,"group":244,"sourceId":5,"sourceLabel":6,"table":245,"selfRows":122,"metrics":246,"seqs":255,"entrants":263,"cells":269,"outcomes":316,"locators":317,"hardware":318,"wordings":319,"notes":320},"lemon2026-table-v","lemon2026:Table V","Table V",[247,250,252,254],{"label":248,"unit":249,"statistic":93,"alignment":93},"AWD(m)","m",{"label":251,"unit":249,"statistic":93,"alignment":93},"CD(m)",{"label":253,"unit":126,"statistic":93,"alignment":93},"SCS",{"label":125,"unit":126,"statistic":93,"alignment":93},[256,259,261],{"dataset":110,"sequence":257,"environment":258},"Island","large-scale aerial scene over 100,000 m^2",{"dataset":110,"sequence":260,"environment":258},"Town",{"dataset":110,"sequence":262,"environment":258},"Airport",[264,267],{"name":265,"methodId":266,"linkable":153,"proposed":82,"self":82},"LAMM","lamm2025",{"name":268,"methodId":5,"linkable":153,"proposed":153,"self":153},"Ours",[270,272,274,276,278,280,282,284,286,288,290,292,294,296,298,300,302,304,306,308,310,312,313,314],[161,161,161,271,163,161,163,163,161],1.3,[161,165,161,273,163,161,163,163,161],6.44,[161,168,161,275,163,161,163,163,161],0.5,[161,219,161,277,163,161,163,163,161],-5.51,[165,161,161,279,163,161,163,163,161],0.3,[165,165,161,281,163,161,163,163,161],0.43,[165,168,161,283,163,161,163,163,161],0.49,[165,219,161,285,163,161,163,163,161],-6.13,[161,161,165,287,163,161,163,163,161],1.77,[161,165,165,289,163,161,163,163,161],10.64,[161,168,165,291,163,161,163,163,161],0.57,[161,219,165,293,163,161,163,163,161],-6.37,[165,161,165,295,163,161,163,163,161],0.65,[165,165,165,297,163,161,163,163,161],1.11,[165,168,165,299,163,161,163,163,161],0.47,[165,219,165,301,163,161,163,163,161],-6.52,[161,161,168,303,163,161,163,163,161],0.75,[161,165,168,305,163,161,163,163,161],36.05,[161,168,168,307,163,161,163,163,161],0.62,[161,219,168,309,163,161,163,163,161],-6.17,[165,161,168,311,163,161,163,163,161],0.25,[165,165,168,299,163,161,163,163,161],[165,168,168,291,163,161,163,163,161],[165,219,168,315,163,161,163,163,161],-6.28,[],[245],[],[],[321],"Mapping quality against the MARS-LVIG ground-truth map (DJI L1 LiDAR processed with DJI Terra); AWD average Wasserstein distance, CD Chamfer distance, SCS spatial consistency score, MME mean map entropy; all lower is better; metrics as defined in MapEval",{"slug":323,"group":324,"sourceId":5,"sourceLabel":6,"table":325,"selfRows":326,"metrics":327,"seqs":330,"entrants":349,"cells":354,"outcomes":407,"locators":409,"hardware":410,"wordings":411,"notes":412},"lemon2026-table-iv","lemon2026:Table IV","Table IV",10,[328],{"label":329,"unit":249,"statistic":134,"alignment":93},"RMSE of the ATE (m)",[331,334,336,337,339,342,344,346,347,348],{"dataset":98,"sequence":332,"environment":333},"Campus 3","multi-robot, see Table II of the paper",{"dataset":98,"sequence":335,"environment":333},"Dormitory",{"dataset":98,"sequence":141,"environment":333},{"dataset":98,"sequence":338,"environment":333},"Tunnel",{"dataset":340,"sequence":341,"environment":333},"GEODE","Inlandwaterways",{"dataset":340,"sequence":343,"environment":333},"Offroad",{"dataset":340,"sequence":345,"environment":333},"Tunnelingtunnel",{"dataset":110,"sequence":262,"environment":333},{"dataset":110,"sequence":260,"environment":333},{"dataset":110,"sequence":257,"environment":333},[350,352,353],{"name":351,"methodId":53,"linkable":82,"proposed":82,"self":82},"DCL-SLAM",{"name":265,"methodId":266,"linkable":153,"proposed":82,"self":82},{"name":268,"methodId":5,"linkable":153,"proposed":153,"self":153},[355,357,359,361,363,365,367,369,371,373,375,376,378,380,381,383,385,387,388,390,391,393,394,396,397,399,400,402,404,405],[161,161,161,356,163,161,163,163,161],17.89,[165,161,161,358,163,161,163,163,161],12.51,[168,161,161,360,163,161,163,163,161],3.51,[161,161,165,362,163,161,163,163,161],3.52,[165,161,165,364,163,161,163,163,161],28.67,[168,161,165,366,163,161,163,163,161],3.46,[161,161,168,368,163,161,163,163,161],6.37,[165,161,168,370,163,161,163,163,161],4.58,[168,161,168,372,163,161,163,163,161],1.48,[161,161,219,374,163,161,163,163,161],3.09,[165,161,219,53,161,161,163,163,161],[168,161,219,377,163,161,163,163,161],0.98,[161,161,379,53,161,161,163,163,161],4,[165,161,379,53,161,161,163,163,161],[168,161,379,382,163,161,163,163,161],7.44,[161,161,384,53,161,161,163,163,161],5,[165,161,384,386,163,161,163,163,161],25.45,[168,161,384,271,163,161,163,163,161],[161,161,389,53,161,161,163,163,161],6,[165,161,389,307,163,161,163,163,161],[168,161,389,392,163,161,163,163,161],0.16,[161,161,116,53,161,161,163,163,161],[165,161,116,395,163,161,163,163,161],8.41,[168,161,116,271,163,161,163,163,161],[161,161,398,53,161,161,163,163,161],8,[165,161,398,53,161,161,163,163,161],[168,161,398,401,163,161,163,163,161],1.49,[161,161,403,53,161,161,163,163,161],9,[165,161,403,53,161,161,163,163,161],[168,161,403,406,163,161,163,163,161],0.801,[408],"failed (RMSE above 30 m)",[325],[],[],[413],"Multi-robot localization; RMSE of ATE (m); failure (x) = any sequence with RMSE above 30 m; GEODE, MARS-LVIG and S3E sequences split into sessions with unknown relative transforms",{"slug":415,"group":416,"sourceId":5,"sourceLabel":6,"table":417,"selfRows":116,"metrics":418,"seqs":420,"entrants":432,"cells":439,"outcomes":478,"locators":479,"hardware":480,"wordings":481,"notes":482},"lemon2026-table-ix","lemon2026:Table IX","Table IX",[419],{"label":329,"unit":249,"statistic":134,"alignment":93},[421,424,425,426,427,429,431],{"dataset":98,"sequence":422,"environment":423},"Campus 1","multi-robot",{"dataset":98,"sequence":332,"environment":423},{"dataset":98,"sequence":335,"environment":423},{"dataset":98,"sequence":338,"environment":423},{"dataset":340,"sequence":428,"environment":423},"Stairs",{"dataset":110,"sequence":430,"environment":423},"Valley",{"dataset":110,"sequence":257,"environment":423},[433,435,437],{"name":434,"methodId":53,"linkable":82,"proposed":82,"self":82},"FPGO",{"name":436,"methodId":53,"linkable":82,"proposed":82,"self":82},"FPGO + BA",{"name":438,"methodId":5,"linkable":153,"proposed":153,"self":153},"LEMON Full",[440,442,444,446,448,450,451,453,455,456,458,460,461,463,465,467,469,471,473,475,477],[161,161,161,441,163,161,163,163,161],9.97,[165,161,161,443,163,161,163,163,161],10.05,[168,161,161,445,163,161,163,163,161],9.94,[161,161,165,447,163,161,163,163,161],3.82,[165,161,165,449,163,161,163,163,161],4.26,[168,161,165,360,163,161,163,163,161],[161,161,168,452,163,161,163,163,161],4.12,[165,161,168,454,163,161,163,163,161],4.4,[168,161,168,366,163,161,163,163,161],[161,161,219,457,163,161,163,163,161],1.29,[165,161,219,459,163,161,163,163,161],1.52,[168,161,219,377,163,161,163,163,161],[161,161,379,462,163,161,163,163,161],0.194,[165,161,379,464,163,161,163,163,161],0.191,[168,161,379,466,163,161,163,163,161],0.142,[161,161,384,468,163,161,163,163,161],7.84,[165,161,384,470,163,161,163,163,161],7.39,[168,161,384,472,163,161,163,163,161],6.76,[161,161,389,474,163,161,163,163,161],0.963,[165,161,389,476,163,161,163,163,161],1.076,[168,161,389,406,163,161,163,163,161],[],[417],[],[],[483],"Ablation of the map merging module: first PGO only, first PGO plus spatial BA, and full LEMON-Mapping; RMSE of ATE (m)",[485,490,496],{"group":486,"slug":487,"sourceLabel":6,"table":488,"selfRows":379,"datasets":489},"lemon2026:Table VIII","lemon2026-table-viii","Table VIII",[98],{"group":491,"slug":492,"sourceLabel":6,"table":493,"selfRows":219,"datasets":494},"lemon2026:Table X","lemon2026-table-x","Table X",[495],"R3LIVE",{"group":497,"slug":498,"sourceLabel":6,"table":499,"selfRows":168,"datasets":500},"lemon2026:Table VI","lemon2026-table-vi","Table VI",[98],1790510654936]