[{"data":1,"prerenderedAt":336},["ShallowReactive",2],{"method-pronto2020":3},{"method":4,"reference":60,"equipment":83,"figures":151,"results":190},{"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":39,"estimator":44,"association":45,"timeModel":46,"deskew":47,"loopClosure":48,"globalOptimization":49,"mapRepresentation":50,"prior":51,"outputGeometry":52,"compute":53,"codeUrl":54,"codeLicense":55,"relatedVersions":56},"pronto2020","Camurri et al., 2020","Pronto","Pronto: A Multi-Sensor State Estimator for Legged Robots in Real-World Scenarios",2020,"recent","C07","odometry","Pronto 是為腿式機器人設計的模組化擴展卡爾曼濾波器：以 IMU 作為高頻過程模型，先融合腿部運動學與接觸偵測得到的速度，再把延遲且低頻的視覺里程計與 LiDAR 點雲配準結果，以鬆耦合的位姿修正方式插入約 10 秒的量測歷史中重新傳播。這樣可在控制迴路中提供 250 至 1000 Hz 的低延遲狀態估測，同時利用外感測器抑制長時間漂移。","Modular EKF for legged robots that runs an IMU process model with leg-odometry velocity updates at control rate and inserts delayed, low-rate visual odometry and ICP-based LiDAR corrections into a stored measurement history, giving low-latency, low-drift estimates for closed-loop locomotion.","full_text_reviewed","peer_reviewed_published","supplementary","腿式機器人（如 ANYmal、Spot）已用於工地巡檢與掃描，本文提供其底層狀態估測的完整設計與長時間實測。ANYmal 實驗在消防訓練場的工業設施內進行，有水面反光、低光照與崎嶇地形，並以 Leica TS16 全測站作為真值，與施工現場的條件部分相近（推論）。但本文只評估位姿，不產生或評估點雲地圖。",[20,21,22],"controlled_experiment","independent_reference","cross_site",[24,25,26,27,28],"Validated on four legged robots over more than 2 h and 1.37 km, including closed-loop control during dynamic gaits (Sec. 7; Table 2)","RPE over 10 m of 0.016 to 0.033 m in the smaller humanoid and HyQ experiments (Table 2)","On Valkyrie the translation error averaged 1.6 cm and never exceeded 7.5 cm, while the estimator without LiDAR drifted without bound (Sec. 7.3)","On ANYmal in an industrial fire-training site, fusing VO and AICP reduced RPE by 60 % compared with IMU and leg odometry only; the pose was within 30 cm of ground truth after 240 m (Sec. 7.5)","Runs on laptop-class CPUs without GPU and is released as open-source ROS packages (Sec. 5)",[30,31,32,33],"No triaging when exteroceptive sources disagree; measurement confidence is a fixed covariance (Sec. 8)","Loosely coupled design; the authors name tightly coupled joint optimization as future work (Sec. 8)","No loop closure or global map; ground truth was not available for all experiments (Table 2)","The Atlas and HyQ outdoor accuracy values are bounds from indirect evaluation (Table 2 footnotes)",[35,36,37,38],"IMU (KVH 1750, KVH 1775, 3DM-GX4-25 or Xsens MTi-100 depending on robot)","joint encoders and force or torque sensing for leg odometry","stereo camera (Carnegie Robotics Multisense SL) or RGB-D camera (Intel RealSense D435) for FOVIS visual odometry","LiDAR (Hokuyo UTM-30LX-EW spinning in the Multisense SL, or Velodyne VLP-16 on ANYmal) for AICP registration",[40,41,42,43],"Atlas humanoid","Valkyrie humanoid","HyQ quadruped","ANYmal quadruped","modular extended Kalman filter with an IMU process model on the real-time control computer; leg-odometry velocity updates; zero-velocity gyro bias update when stationary at least 400 ms; loosely coupled pose or position corrections from FOVIS visual odometry and from AICP LiDAR registration applied through a measurement history of typically 10 s to handle their latency (Secs. 4-5)","leg odometry from kinematics with contact detection (Schmitt trigger on humanoids, probabilistic ground-reaction-force contact on quadrupeds); FOVIS feature-based stereo or RGB-D visual odometry; AICP auto-tuned ICP with planar pre-filtering and overlap-based outlier rejection against a reference cloud (Sec. 4)","discrete EKF at the IMU rate (250 to 1000 Hz) with delayed asynchronous exteroceptive updates at 1 to 15 Hz inserted into the stored history and re-propagated (Sec. 5.1; Fig. 6)","not described in the paper; AICP registers accumulated point clouds","no","none; the filter marginalizes previous states (Sec. 9)","AICP reference point cloud updated after a travelled distance; no global map is maintained by the estimator","robot kinematic model; fixed measurement covariances per module; for HyQ outdoor ground truth an ICP prior map was used for evaluation only","base pose and velocity at control rate for closed-loop locomotion","consumer-grade processors (e.g., laptop-class Intel i7), no GPU; proprioceptive thread on the control computer, VO and LiDAR on a perception computer (Sec. 5)","https:\u002F\u002Fgithub.com\u002Fori-drs\u002Fpronto","LGPL-2.1 (GitHub license metadata for ori-drs\u002Fpronto)",[57],{"relation":58,"title":59,"doi_or_url":54},"code_release","ori-drs\u002Fpronto (core EKF and leg odometry), with fovis_ros and aicp_mapping modules",{"id":5,"kind":61,"shortName":7,"title":8,"authors":62,"year":9,"venue":67,"venueType":68,"publisher":69,"volumeIssuePages":70,"doi":71,"arxivId":72,"url":73,"firstPublicDate":74,"publicationStatus":16,"metadataStatus":75,"fulltextStatus":15,"era":10,"classicReason":76,"codeUrl":54,"cluster":11,"topics":77,"mdpi":78,"verification":79,"label":6,"fulltextRoute":80,"versionRead":81,"addedByCensus":82},"method",[63,64,65,66],"Marco Camurri","Milad Ramezani","Simona Nobili","Maurice Fallon","Frontiers in Robotics and AI","journal","Frontiers Media SA","7, article 68","10.3389\u002Ffrobt.2020.00068",null,"https:\u002F\u002Fdoi.org\u002F10.3389\u002Ffrobt.2020.00068","2020-06-05","metadata_verified","not_applicable",[11],false,"corrected","publisher open access (OpenAlex oa_url)","Publisher version of record (Frontiers HTML and PDF, open access, CC BY 4.0)",true,[84,91,97,102,106,110,114,119,125,130,136,141,145,148],{"category":85,"model":86,"canonical":86,"role":87,"dataset":88,"specs":89,"locator":90},"imu","KVH 1750","method input","Atlas (DRC Finals)","333 Hz; initial bias 0.5 deg\u002Fh and 0.5 mg; bias stability 0.05 deg\u002Fh and 0.05 mg","Table 1",{"category":92,"model":93,"canonical":93,"role":87,"dataset":94,"specs":95,"locator":96},"stereo_camera","Multisense SL","Atlas, Valkyrie, HyQ","tri-modal ruggedized sensor; stereo 10 Hz, 1024 x 1024 px, 80 x 80 deg FoV (CMV4000 imager) on Atlas; used on Atlas, Valkyrie and HyQ","Table 1; Sec. 6",{"category":98,"model":99,"canonical":100,"role":87,"dataset":94,"specs":101,"locator":90},"lidar","UTM-30LX-EW (spinning, in Multisense SL)","Hokuyo UTM-30LX","40 Hz; full-rotation FoV 220 x 180 deg (Atlas entry); produces sparse clouds that AICP accumulates",{"category":85,"model":103,"canonical":103,"role":87,"dataset":104,"specs":105,"locator":90},"3DM-GX4-25","Valkyrie","500 Hz",{"category":107,"model":108,"canonical":108,"role":87,"dataset":104,"specs":109,"locator":90},"other","Omega85","foot force and torque sensing for contact detection",{"category":85,"model":111,"canonical":111,"role":87,"dataset":112,"specs":113,"locator":90},"KVH 1775","HyQ","1000 Hz; initial bias 0.5 deg\u002Fh and 0.5 mg; bias stability 0.05 deg\u002Fh and 0.05 mg",{"category":85,"model":115,"canonical":116,"role":87,"dataset":117,"specs":118,"locator":90},"MTi-100","Xsens MTi-100","ANYmal","400 Hz; initial bias 0.2 deg\u002Fs and 5 mg; bias stability 10 deg\u002Fh and 15 mg",{"category":120,"model":121,"canonical":122,"role":87,"dataset":117,"specs":123,"locator":124},"rgbd","RealSense D435","Intel RealSense D435","30 Hz, 848 x 480 for VO; a second downward-facing D435 also mounted","Table 1; Sec. 7.5",{"category":98,"model":126,"canonical":127,"role":87,"dataset":117,"specs":128,"locator":129},"VLP-16","Velodyne VLP-16","mounted on top of ANYmal for localization and global mapping; used by AICP","Sec. 6.4",{"category":107,"model":131,"canonical":131,"role":132,"dataset":133,"specs":134,"locator":135},"Vicon","reference or ground truth","Valkyrie, HyQ","100 Hz, millimetre-accurate; Valkyrie and HyQ indoor experiments","Sec. 7.1.1",{"category":137,"model":138,"canonical":138,"role":132,"dataset":139,"specs":140,"locator":135},"total_station","TS16","ANYmal (Fire Service College)","laser tracking system following a reflective prism with millimetre accuracy; ANYmal experiment",{"category":142,"model":143,"canonical":143,"role":87,"dataset":112,"specs":144,"locator":90},"wheel_or_leg_odometry","AEDA3300-BE1","joint encoders, 1,000 Hz, resolution \u003C0.0045 deg; leg odometry input",{"category":107,"model":146,"canonical":146,"role":87,"dataset":112,"specs":147,"locator":90},"Burster 8417","force sensing, 1,000 Hz, resolution \u003C25 N",{"category":142,"model":149,"canonical":149,"role":87,"dataset":117,"specs":150,"locator":90},"ANYdrive","joint encoders (resolution \u003C0.025 deg) and torque at 400 Hz; leg odometry input",[152,165,173,182],{"refId":5,"refLabel":6,"fig":153,"whatZh":154,"license":155,"licenseUrl":156,"sourceUrl":157,"src":158,"width":159,"height":160,"thumb":161,"thumbWidth":162,"thumbHeight":163,"modified":164},"Figure 5","系統方塊圖：IMU 過程模型與腿部里程計在控制電腦即時執行，視覺里程計與 LiDAR 配準在感知電腦執行後回饋修正","CC BY 4.0","http:\u002F\u002Fcreativecommons.org\u002Flicenses\u002Fby\u002F4.0\u002F","https:\u002F\u002Fwww.frontiersin.org\u002Ffiles\u002FArticles\u002F520900\u002Ffrobt-07-00068-HTML\u002Fimage_m\u002Ffrobt-07-00068-g005.jpg","\u002Ffigure-files\u002Fpronto2020\u002Ffigure-5.webp",536,311,"\u002Ffigure-files\u002Fpronto2020\u002Ffigure-5.thumb.webp",480,279,"converted to WebP",{"refId":5,"refLabel":6,"fig":166,"whatZh":167,"license":155,"licenseUrl":156,"sourceUrl":168,"src":169,"width":159,"height":170,"thumb":171,"thumbWidth":162,"thumbHeight":172,"modified":164},"Figure 7","Atlas 在 DRC 決賽資料上的 AICP 表現：上方為 206 片點雲對齊結果的上視圖，左為含人群的原始點雲，右為濾除後點雲；下方比較未套用修正時閥門在各片點雲中位置不一致，以及定位成功後估測一致的結果","https:\u002F\u002Fwww.frontiersin.org\u002Ffiles\u002FArticles\u002F520900\u002Ffrobt-07-00068-HTML\u002Fimage_m\u002Ffrobt-07-00068-g007.jpg","\u002Ffigure-files\u002Fpronto2020\u002Ffigure-7.webp",372,"\u002Ffigure-files\u002Fpronto2020\u002Ffigure-7.thumb.webp",333,{"refId":5,"refLabel":6,"fig":174,"whatZh":175,"license":155,"licenseUrl":156,"sourceUrl":176,"src":177,"width":178,"height":179,"thumb":180,"thumbWidth":162,"thumbHeight":181,"modified":164},"Figure 9","HyQ 室內重複性測試、室外昏暗工業區探索測試，以及不同感測組合估測軌跡的比較","https:\u002F\u002Fwww.frontiersin.org\u002Ffiles\u002FArticles\u002F520900\u002Ffrobt-07-00068-HTML\u002Fimage_m\u002Ffrobt-07-00068-g009.jpg","\u002Ffigure-files\u002Fpronto2020\u002Ffigure-9.webp",1084,317,"\u002Ffigure-files\u002Fpronto2020\u002Ffigure-9.thumb.webp",140,{"refId":5,"refLabel":6,"fig":183,"whatZh":184,"license":155,"licenseUrl":156,"sourceUrl":185,"src":186,"width":178,"height":187,"thumb":188,"thumbWidth":162,"thumbHeight":189,"modified":164},"Figure 12","ANYmal 在消防訓練場的實驗：場景照片、機上相機畫面（水面反光），以及估測軌跡與真值比較","https:\u002F\u002Fwww.frontiersin.org\u002Ffiles\u002FArticles\u002F520900\u002Ffrobt-07-00068-HTML\u002Fimage_m\u002Ffrobt-07-00068-g012.jpg","\u002Ffigure-files\u002Fpronto2020\u002Ffigure-12.webp",384,"\u002Ffigure-files\u002Fpronto2020\u002Ffigure-12.thumb.webp",170,{"totalRows":191,"groupCount":192,"groups":193,"others":335},14,2,[194,271],{"slug":195,"group":196,"sourceId":5,"sourceLabel":6,"table":197,"selfRows":198,"metrics":199,"seqs":209,"entrants":232,"cells":241,"outcomes":263,"locators":266,"hardware":267,"wordings":268,"notes":269},"pronto2020-table-2","pronto2020:Table 2","Table 2",8,[200,205,207],{"label":201,"unit":202,"statistic":203,"alignment":204},"RPE \u003C=0.03* (footnote: by evaluation of the ground truth point cloud) [m], over 10 m","m","mean","not_reported",{"label":206,"unit":202,"statistic":203,"alignment":204},"RPE [m], over 10 m",{"label":208,"unit":202,"statistic":203,"alignment":204},"RPE \u003C=0.03** (footnote: by accuracy in returning to the initial position) [m], over 10 m",[210,214,217,220,223,226,229],{"dataset":211,"sequence":212,"environment":213},"Pronto experiments (authors)","Exp. 1 (Atlas)","DARPA Robotics Challenge Finals course, 1236 s, 16 m, 154 m2, no GT",{"dataset":211,"sequence":215,"environment":216},"Exp. 2 (Valkyrie)","flat-ground repeated walking, 341 s, 12 m, 78 m2, Vicon GT",{"dataset":211,"sequence":218,"environment":219},"Exp. 3 (Valkyrie)","stair climbing, 50 s, 2.5 m, 78 m2, Vicon GT",{"dataset":211,"sequence":221,"environment":222},"Exp. 4 (HyQ)","laboratory trotting, 1740 s, 400 m, 7.5 m2, Vicon GT",{"dataset":211,"sequence":224,"environment":225},"Exp. 5 (HyQ)","poorly lit industrial area, 1740 s, 400 m, 9 m2, no GT",{"dataset":211,"sequence":227,"environment":228},"Exp. 6 (HyQ)","poorly lit industrial area with ramps and rock beds, 2640 s, 300 m, 100 m2, GT available",{"dataset":211,"sequence":230,"environment":231},"Exp. 7 (ANYmal)","Fire Service College industrial training site, 1996 s, 240 m, 1381 m2, Leica TS16 GT",[233,235,237,239],{"name":234,"methodId":5,"linkable":82,"proposed":82,"self":82},"Pronto (IMU+LO+AICP, online, control loop)",{"name":236,"methodId":5,"linkable":82,"proposed":82,"self":82},"Pronto (IMU+LO+VO+AICP, online, control loop)",{"name":238,"methodId":5,"linkable":82,"proposed":82,"self":82},"Pronto (IMU+LO+VO+AICP, offline)",{"name":240,"methodId":5,"linkable":82,"proposed":82,"self":82},"Pronto (IMU+LO only (without VO and AICP), offline)",[242,246,249,250,253,255,258,261],[243,243,243,244,243,243,245,245,243],0,0.03,-1,[243,247,247,248,245,243,245,245,243],1,0.016,[243,247,192,248,245,243,245,245,243],[247,247,251,252,245,243,245,245,243],3,0.027,[247,192,254,244,247,243,245,245,243],4,[247,247,256,257,245,243,245,245,243],5,0.033,[192,247,259,260,245,243,245,245,243],6,0.34,[251,247,259,262,245,243,245,245,243],0.83,[264,265],"upper bound as printed '\u003C=0.03*'","upper bound as printed '\u003C=0.03**'",[197],[],[],[270],"Summary of experiments; RPE = translational part of relative pose error evaluated over 10 m distance; OL = online, CL = used in the control loop; GT = ground truth available (Vicon for Valkyrie and HyQ indoor, Leica TS16 for ANYmal)",{"slug":272,"group":273,"sourceId":5,"sourceLabel":6,"table":274,"selfRows":259,"metrics":275,"seqs":289,"entrants":301,"cells":310,"outcomes":322,"locators":326,"hardware":331,"wordings":332,"notes":333},"pronto2020-sec-7-text","pronto2020:Sec. 7 text","Sec. 7 text",[276,278,281,284,286],{"label":277,"unit":202,"statistic":203,"alignment":204},"error in translation, average",{"label":279,"unit":202,"statistic":280,"alignment":204},"error in translation, never exceeded","max",{"label":282,"unit":202,"statistic":283,"alignment":204},"average 3D median translation error","median",{"label":285,"unit":202,"statistic":204,"alignment":204},"pose estimate distance from ground truth after travelling 240 m",{"label":287,"unit":288,"statistic":204,"alignment":204},"drift as percentage of travelled distance (Pronto without AICP)","%",[290,293,296,298],{"dataset":211,"sequence":291,"environment":292},"Exp. 2 (Valkyrie walk to target)","flat-ground repeated walking, Vicon GT",{"dataset":211,"sequence":294,"environment":295},"Exp. 6c-6e (HyQ)","poorly lit industrial area",{"dataset":211,"sequence":230,"environment":297},"Fire Service College industrial training site",{"dataset":211,"sequence":299,"environment":300},"preliminary indoor tests (Atlas)","indoor",[302,304,306,308],{"name":303,"methodId":5,"linkable":82,"proposed":82,"self":82},"Pronto (IMU+LO+AICP)",{"name":305,"methodId":5,"linkable":82,"proposed":82,"self":82},"Pronto (IMU+LO+AICP+VO)",{"name":307,"methodId":5,"linkable":82,"proposed":82,"self":82},"Pronto (IMU+LO+VO+AICP)",{"name":309,"methodId":5,"linkable":82,"proposed":82,"self":82},"Pronto (IMU+LO, without AICP)",[311,312,314,316,318,320],[243,243,243,248,245,243,245,245,243],[243,247,243,313,245,243,245,245,243],0.075,[243,192,247,315,243,247,245,245,243],0.049,[247,192,247,317,245,247,245,245,243],0.032,[192,251,192,319,247,192,245,245,243],0.3,[251,254,251,321,192,251,245,245,243],1.67,[323,324,325],"approximate as printed ('~4.9 cm')","upper bound as printed ('\u003C30 cm')","approximate as printed ('~1.67 %')",[327,328,329,330],"Sec. 7.3","Sec. 7.4; Fig. 11","Sec. 7.5","Sec. 7.2",[],[],[334],"Values stated in the text of Section 7",[],1790510662096]