[{"data":1,"prerenderedAt":366},["ShallowReactive",2],{"method-lamp2_2022":3},{"method":4,"reference":70,"equipment":101,"figures":138,"results":139},{"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":23,"limitations":29,"sensors":34,"platform":38,"estimator":42,"association":43,"timeModel":44,"deskew":45,"loopClosure":46,"globalOptimization":47,"mapRepresentation":48,"prior":49,"outputGeometry":50,"compute":51,"codeUrl":52,"codeLicense":53,"relatedVersions":54},"lamp2_2022","Chang et al., 2022","LAMP 2.0","LAMP 2.0: A Robust Multi-Robot SLAM System for Operation in Challenging Large-Scale Underground Environments",2022,"recent","C11b","full_slam_with_global_correction","LAMP 2.0 是 CoSTAR 團隊為 DARPA 地下挑戰賽開發的集中式多機器人 LiDAR 位姿圖 SLAM。各機器人的前端介面可接不同里程計（LOCUS 或 Hovermap）與不同 LiDAR 配置，先以 HeRO 狀態估計去除掃描畸變、合併多顆 LiDAR，再用自適應體素濾波讓點數一致，並每約 2 m 或 30 度建立關鍵節點與對應掃描送到基地站。基地站的多機器人前端以自適應半徑產生迴圈閉合候選，依可觀測性、圖神經網路預測效益與 RSSI 排序，再以 TEASER++ 或 SAC-IA 初始對齊後用 GICP 精修；後端以 GNC 搭配 Levenberg-Marquardt 在 GTSAM 中做抗離群值的位姿圖最佳化。作者在煤礦、核電廠、SubT 決賽場地與石灰岩礦的四組資料上評估，並釋出含地面真值的資料集。","Centralized multi-robot LiDAR pose-graph SLAM for large underground sites: odometry-agnostic front-end interfaces, prioritized inter- and intra-robot loop closures computed with TEASER++ or SAC-IA plus GICP, and a GNC outlier-robust back-end in GTSAM; evaluated on four datasets from a coal mine, a power plant, the SubT Finals course and a limestone mine with released ground truth.","full_text_reviewed","peer_reviewed_published","supplementary","驗證場景是地下礦坑、廢棄核電廠與人工洞穴，屬地下基礎設施而非施工工地；釋出的位姿圖、關鍵掃描與以測量地圖產生的真值，可作為隧道與地下工程 SLAM 的大尺度參考資料（Sec. I、III-B）。",[20,21,22],"underground_or_tunnel","independent_reference","cross_site",[24,25,26,27,28],"Per-robot ATE below 2 m for trajectories up to 2.2 km (Table IV; Sec. III-D)","TEASER++ or SAC-IA initialization lowered false-positive rates and loop-closure pose errors compared with odometric initialization (Table II)","Fewer candidates but more verified and inlier loop closures than LAMP 1.0; GNC handled more than 80% outlier loop closures (Table III; Sec. III-C)","Handles heterogeneous robots with different lidar configurations and odometry sources in the same run (Sec. II-B, III-A)","Map errors below 4 m against surveyed maps in all four large environments (Fig. 4; Sec. III-D)",[30,31,32,33],"Centralized architecture may not scale to large robot teams; a distributed version is future work (Sec. IV)","Loop-closure recall in the wide KU limestone tunnels stays low across initializations (11.3% to 29.0%, the highest with GT initialization) (Table II)","Initial common frame relies on a calibration gate with reflective markers (Sec. II-A)","Map accuracy is shown only as colour-coded cloud-to-cloud error maps (Fig. 4), not tabulated",[35,36,37],"3D LiDARs (three Velodyne lidars on Husky, a single lidar on Spot; models not reported)","Hovermap payload on some robots","odometry input from LOCUS or Hovermap (front-end agnostic)",[39,40,41],"wheeled UGV (Husky)","legged (Spot)","multi-robot teams of up to four robots with a centralized base station","Centralized multi-robot pose-graph optimization in GTSAM with Levenberg-Marquardt and Graduated Non-Convexity (GNC) for outlier-robust inlier selection, optionally after Incremental Consistency Maximization (ICM); single-robot front-ends send sparse pose graphs (key nodes every about 2 m or 30 deg) with keyed scans (Sec. II-B, II-D)","Proximity-based loop-closure candidates with adaptive radius, prioritized by observability (ICP information-matrix eigenvalues), a GNN-predicted graph benefit and RSSI beacons; relative pose from TEASER++ or SAC-IA initialization refined by GICP, rejecting poor alignments (Sec. II-C)","discrete key nodes; scans de-skewed by the HeRO local state estimate (Sec. II-B)","motion distortion corrected with the Heterogeneous Robust Odometry (HeRO) local state estimate before merging multi-lidar scans (Sec. II-B)","intra- and inter-robot loop closures from the multi-robot front-end with two-stage registration (TEASER++ or SAC-IA, then GICP) (Sec. II-C)","centralized multi-robot pose-graph optimization with GNC (and ICM) outlier rejection in GTSAM (Sec. II-D)","pose graph with keyed scans (adaptive voxel-filtered point clouds); optimized global point-cloud map formed by transforming keyed scans with optimized poses (Sec. II-D)","common reference frame from a gate with three reflective plates of known coordinates at the entrance (Sec. II-A)","globally consistent multi-robot trajectories and point-cloud map; map compared with surveyed ground-truth map by cloud-to-cloud error (Fig. 4)","base station during SubT: AMD Ryzen Threadripper 3990x (64 cores); paper experiments: laptop Intel i7-8750H (12 cores), data played back in real time (Sec. III-A, III-D)","https:\u002F\u002Fgithub.com\u002FNeBula-Autonomy\u002FLAMP","MIT (LICENSE file on main branch checked 2026-09-25)",[55,59,63,66],{"relation":56,"title":57,"doi_or_url":58},"preprint","arXiv 2205.13135 v1 to v3","https:\u002F\u002Farxiv.org\u002Fabs\u002F2205.13135",{"relation":60,"title":61,"doi_or_url":62},"predecessor","LAMP: Large-scale autonomous mapping and positioning for exploration of perceptually-degraded subterranean environments, ICRA 2020, pp. 80-86 (ref. [8]; DOI not verified)","not_verified",{"relation":64,"title":65,"doi_or_url":52},"code_release","NeBula-Autonomy\u002FLAMP",{"relation":67,"title":68,"doi_or_url":69},"dataset_release","NeBula-Autonomy\u002Fnebula-multirobot-dataset (pose graphs, keyed scans, ground truth)","https:\u002F\u002Fgithub.com\u002FNeBula-Autonomy\u002Fnebula-multirobot-dataset",{"id":5,"kind":71,"shortName":7,"title":8,"authors":72,"year":9,"venue":85,"venueType":86,"publisher":87,"volumeIssuePages":88,"doi":89,"arxivId":90,"url":91,"firstPublicDate":92,"publicationStatus":16,"metadataStatus":93,"fulltextStatus":15,"era":10,"classicReason":94,"codeUrl":52,"cluster":11,"topics":95,"mdpi":96,"verification":97,"label":6,"fulltextRoute":98,"versionRead":99,"addedByCensus":100},"method",[73,74,75,76,77,78,79,80,81,82,83,84],"Yun Chang","Kamak Ebadi","Christopher E. Denniston","Muhammad Fadhil Ginting","Antoni Rosinol","Andrzej Reinke","Matteo Palieri","Jingnan Shi","Arghya Chatterjee","Benjamin Morrell","Ali-akbar Agha-mohammadi","Luca Carlone","IEEE Robotics and Automation Letters","journal","IEEE","7(4):9175-9182","10.1109\u002Flra.2022.3191204","2205.13135","https:\u002F\u002Fdoi.org\u002F10.1109\u002FLRA.2022.3191204","2022-05-26","metadata_verified","not_applicable",[11],false,"corrected","arXiv","arXiv v3 (2022-07-09), marked as accepted for publication at RA-L; version of record not compared",true,[102,109,112,117,121,127,130,135],{"category":103,"model":104,"canonical":104,"role":105,"dataset":106,"specs":107,"locator":108},"platform","Husky","method input","CoSTAR multi-robot datasets","wheeled platform equipped with three Velodyne lidars and a Hovermap","Sec. III-A",{"category":103,"model":110,"canonical":110,"role":105,"dataset":106,"specs":111,"locator":108},"Spot","quadruped platform equipped with either a single lidar or a Hovermap",{"category":113,"model":114,"canonical":114,"role":105,"dataset":106,"specs":115,"locator":116},"lidar","Velodyne lidars (three per Husky; model not reported)","merged after extrinsic calibration and adaptive voxelization","Sec. II-B, III-A",{"category":118,"model":119,"canonical":119,"role":105,"dataset":106,"specs":120,"locator":116},"mobile_scanner_device","Hovermap","provides odometry and point clouds as an alternative front-end",{"category":122,"model":123,"canonical":123,"role":124,"dataset":125,"specs":126,"locator":108},"compute","AMD Ryzen Threadripper 3990x (64 cores)","compute for runtime",null,"portable base-station workstation during the SubT Challenge",{"category":122,"model":128,"canonical":128,"role":124,"dataset":125,"specs":129,"locator":108},"laptop with Intel i7-8750H (12 cores)","runs the experiments reported in the paper",{"category":131,"model":132,"canonical":132,"role":105,"dataset":125,"specs":133,"locator":134},"other","three reflective plates on an entrance gate","fiducial markers with known 3D coordinates for initial pose calibration","Sec. II-A",{"category":113,"model":136,"canonical":136,"role":105,"dataset":106,"specs":137,"locator":108},"single lidar on Spot (model and manufacturer not reported)","one of the two Spot sensor configurations (the other is a Hovermap); point clouds differ in size and density from the Husky three-lidar setup",[],{"totalRows":140,"groupCount":141,"groups":142,"others":365},38,2,[143,288],{"slug":144,"group":145,"sourceId":5,"sourceLabel":6,"table":146,"selfRows":147,"metrics":148,"seqs":153,"entrants":188,"cells":194,"outcomes":282,"locators":283,"hardware":284,"wordings":285,"notes":286},"lamp2-2022-table-iv","lamp2_2022:Table IV","Table IV",26,[149],{"label":150,"unit":151,"statistic":152,"alignment":152},"ATE [m] (text: average trajectory error)","m","not_reported",[154,158,160,164,166,168,172,174,176,178,182,184,186],{"dataset":155,"sequence":156,"environment":157},"CoSTAR multi-robot dataset: Tunnel","husky3 (traversed 1194 m)","NIOSH Safety Research Coal Mine, Pittsburgh (narrow, mostly featureless tunnels)",{"dataset":155,"sequence":159,"environment":157},"husky4 (traversed 1362 m)",{"dataset":161,"sequence":162,"environment":163},"CoSTAR multi-robot dataset: Urban","husky1 (traversed 612 m)","Satsop abandoned nuclear power plant, Elma (two floors, open areas, small rooms, stairs)",{"dataset":161,"sequence":165,"environment":163},"husky4 (traversed 416 m)",{"dataset":161,"sequence":167,"environment":163},"spot1 (traversed 502 m)",{"dataset":169,"sequence":170,"environment":171},"CoSTAR multi-robot dataset: KU","husky1 (traversed 2204 m)","Kentucky Underground Storage limestone mine, Wilmore (10-20 m wide tunnels)",{"dataset":169,"sequence":173,"environment":171},"husky2 (traversed 1526 m)",{"dataset":169,"sequence":175,"environment":171},"husky3 (traversed 1678 m)",{"dataset":169,"sequence":177,"environment":171},"husky4 (traversed 896 m)",{"dataset":179,"sequence":180,"environment":181},"CoSTAR multi-robot dataset: Final","husky3 (traversed 72 m)","DARPA SubT Finals course, Louisville Mega Cavern (tunnel, cave and urban-like)",{"dataset":179,"sequence":183,"environment":181},"spot1 (traversed 430 m)",{"dataset":179,"sequence":185,"environment":181},"spot3 (traversed 484 m)",{"dataset":179,"sequence":187,"environment":181},"spot4 (traversed 238 m)",[189,190,192],{"name":7,"methodId":5,"linkable":100,"proposed":100,"self":100},{"name":191,"methodId":5,"linkable":100,"proposed":100,"self":100},"LAMP 2.0 single robot (no inter-robot loop closures)",{"name":193,"methodId":125,"linkable":96,"proposed":96,"self":96},"LAMP 1.0",[195,199,202,204,206,208,210,212,214,216,219,220,222,225,227,229,232,234,236,239,241,243,245,247,249,252,254,256,259,260,262,265,267,268,271,273,275,278,280],[196,196,196,197,198,196,198,198,196],0,0.65,-1,[200,196,196,201,198,196,198,198,196],1,0.83,[141,196,196,203,198,196,198,198,196],1.07,[196,196,200,205,198,196,198,198,196],0.72,[200,196,200,207,198,196,198,198,196],0.63,[141,196,200,209,198,196,198,198,196],1.44,[196,196,141,211,198,196,198,198,196],0.79,[200,196,141,213,198,196,198,198,196],0.87,[141,196,141,215,198,196,198,198,196],0.95,[196,196,217,218,198,196,198,198,196],3,0.76,[200,196,217,211,198,196,198,198,196],[141,196,217,221,198,196,198,198,196],0.78,[196,196,223,224,198,196,198,198,196],4,1.31,[200,196,223,226,198,196,198,198,196],1.46,[141,196,223,228,198,196,198,198,196],0.99,[196,196,230,231,198,196,198,198,196],5,1.01,[200,196,230,233,198,196,198,198,196],0.9,[141,196,230,235,198,196,198,198,196],5.34,[196,196,237,238,198,196,198,198,196],6,0.71,[200,196,237,240,198,196,198,198,196],0.75,[141,196,237,242,198,196,198,198,196],2.85,[196,196,244,224,198,196,198,198,196],7,[200,196,244,246,198,196,198,198,196],1.33,[141,196,244,248,198,196,198,198,196],2.11,[196,196,250,251,198,196,198,198,196],8,0.69,[200,196,250,253,198,196,198,198,196],0.86,[141,196,250,255,198,196,198,198,196],5.49,[196,196,257,258,198,196,198,198,196],9,0.16,[200,196,257,258,198,196,198,198,196],[141,196,257,261,198,196,198,198,196],0.56,[196,196,263,264,198,196,198,198,196],10,0.2,[200,196,263,266,198,196,198,198,196],0.37,[141,196,263,266,198,196,198,198,196],[196,196,269,270,198,196,198,198,196],11,0.21,[200,196,269,272,198,196,198,198,196],0.38,[141,196,269,274,198,196,198,198,196],0.55,[196,196,276,277,198,196,198,198,196],12,0.15,[200,196,276,279,198,196,198,198,196],0.23,[141,196,276,281,198,196,198,198,196],0.62,[],[146],[],[],[287],"End-to-end system evaluation: data played back in real time to the base station (about 1 h per run), same odometry input for all variants; ground-truth trajectories from scan-to-map localization in surveyed global maps; per-robot ATE",{"slug":289,"group":290,"sourceId":5,"sourceLabel":6,"table":291,"selfRows":276,"metrics":292,"seqs":300,"entrants":309,"cells":312,"outcomes":359,"locators":360,"hardware":361,"wordings":362,"notes":363},"lamp2-2022-table-iii","lamp2_2022:Table III","Table III",[293,296,298],{"label":294,"unit":295,"statistic":152,"alignment":94},"# Generated (loop-closure candidates)","count",{"label":297,"unit":295,"statistic":152,"alignment":94},"# Verified (passed ICP)",{"label":299,"unit":295,"statistic":152,"alignment":94},"# Inliers (ICM or GNC)",[301,303,305,307],{"dataset":155,"sequence":302,"environment":157},"Tunnel",{"dataset":161,"sequence":304,"environment":163},"Urban",{"dataset":179,"sequence":306,"environment":181},"Final",{"dataset":169,"sequence":308,"environment":171},"KU",[310,311],{"name":193,"methodId":125,"linkable":96,"proposed":96,"self":96},{"name":7,"methodId":5,"linkable":100,"proposed":100,"self":100},[313,315,317,319,321,323,325,327,329,331,333,335,337,339,341,342,344,346,348,350,352,353,355,357],[196,196,196,314,198,196,198,198,196],22206,[196,200,196,316,198,196,198,198,196],3032,[196,141,196,318,198,196,198,198,196],52,[200,196,196,320,198,196,198,198,196],9755,[200,200,196,322,198,196,198,198,196],5656,[200,141,196,324,198,196,198,198,196],1645,[196,196,200,326,198,196,198,198,196],17742,[196,200,200,328,198,196,198,198,196],497,[196,141,200,330,198,196,198,198,196],237,[200,196,200,332,198,196,198,198,196],5356,[200,200,200,334,198,196,198,198,196],1505,[200,141,200,336,198,196,198,198,196],284,[196,196,141,338,198,196,198,198,196],12559,[196,200,141,340,198,196,198,198,196],681,[196,141,141,237,198,196,198,198,196],[200,196,141,343,198,196,198,198,196],3856,[200,200,141,345,198,196,198,198,196],1484,[200,141,141,347,198,196,198,198,196],393,[196,196,217,349,198,196,198,198,196],22616,[196,200,217,351,198,196,198,198,196],14,[196,141,217,269,198,196,198,198,196],[200,196,217,354,198,196,198,198,196],22344,[200,200,217,356,198,196,198,198,196],885,[200,141,217,358,198,196,198,198,196],197,[],[291],[],[],[364],"Number of loop closures at each stage of the multi-robot front-end and back-end; LAMP 1.0 uses fixed-radius candidates, odometric ICP initialization and ICM; LAMP 2.0 adds adaptive radius, prioritization, SAC-IA or TEASER++ initialization and GNC",[],1790510661546]