SLAM 點雲圖鑑Point clouds for construction
M-01 · 2020

LIC-Fusion 2.0

LIC-Fusion 2.0 將光達處理改為滑動視窗內的平面特徵追蹤：以 IMU 做運動補償後擷取低曲率平面點，跨多次掃描追蹤並初始化平面，且考慮幀間轉換不確定性來剔除錯誤匹配。論文同時分析光達慣性子系統在平面特徵下的可觀測性（observability），指出某些運動會使時空校正退化，並以蒙地卡羅模擬驗證估計一致性。

里程計近十年已讀全文同儕審查已出版LiDAR、視覺與慣性融合

閱讀版本
arXiv 2008.07196 v1 (2020-08-17), all 8 pages; IEEE version of record (IROS 2020, pp. 5112-5119) obtained through NTU institutional access and compared: besides the IEEE header, the version of record carries many minor wording edits (for example 'LiDAR-IMU' renamed 'IMU-LiDAR' throughout, a reworded abstract sentence, 'our own multi-sensor platform', citation style), with no change to the method, equations or any table value
全文取得途徑
arXiv 預印本
本頁內容

Adds sliding-window plane-feature tracking with uncertainty-aware outlier rejection and an observability analysis of degenerate motions for spatio-temporal calibration to the LIC-Fusion filter.

技術屬性

欄位內容為文獻擷取紀錄的原文用語（英文），以原文為據；「未查證」表示本研究尚未讀到該資訊，不代表該方法不具備此能力。

感測輸入3D LiDAR (Velodyne VLP-16, 10 Hz)、IMU (Xsens, model not reported, 400 Hz)、monocular global-shutter camera (model not reported, 20 Hz, 1920x1200)
原文測試平台simulation、未查證 (sensor suite carried indoors)
狀態估計sliding-window MSCKF-type filter with online spatio-temporal calibration; visual pipeline based on OpenVINS
資料關聯IMU-motion-compensated low-curvature planar points tracked across the sliding window as plane features, with an outlier test accounting for inter-frame transformation uncertainty; sparse visual features
時間表示discrete cloned poses; all time offsets estimated online from zero initial guess
去畸變IMU-propagated poses are buffered and interpolated to each LiDAR point time (SO(3) interpolation for orientation, linear for position); all points are transformed to the pose at the sweep start time (Sec. III-A)
迴圈閉合none reported
全域最佳化none
地圖表示sliding-window plane features and SLAM plane landmarks; no global map reported
先驗資訊none
可輸出幾何trajectory only (no map product reported)
計算需求IMU-camera subsystem 16.8 ms and LiDAR-IMU subsystem 40.2 ms average on i7-8086K for one sequence (Sec. VI-B)

使用設備

原文使用的感測器、運算硬體與載具（equipment）。型號保留原文寫法，連結到設備頁中同一型號的歸併名稱；角色依原文用途分為方法輸入、資料集感測器、執行運算平台、參考或真值量測（reference or ground truth）與比較對象設備。

類別型號（原文寫法）角色資料集原文規格出處
LiDARVelodyne VLP-16方法輸入未標示10 Hz (Table IV)(Zuo et al., 2020, Sec. VI; Fig. 4; Table IV)
慣性量測單元（IMU）Xsens IMU (model not reported)方法輸入未標示400 Hz (Table IV)(Zuo et al., 2020, Sec. VI; Fig. 4; Table IV)
相機global-shutter monocular camera (model not reported)方法輸入未標示20 Hz, 1920x1200 (Table IV)(Zuo et al., 2020, Sec. VI; Fig. 4; Table IV)
運算硬體Intel i7-8086k [email protected]執行運算平台未標示desktop computer (Sec. VI-B)(Zuo et al., 2020, Sec. VI-B)
其他VICON motion capture system (model not reported)參考或真值量測Vicon Room sequences原文未報告(Zuo et al., 2020, Sec. VI-B)

作者報告的優勢與限制

優勢

  • Observability analysis identifies degenerate motions for LiDAR-IMU calibration with plane features, validated in Monte-Carlo simulation (Sec. IV, Table I, Table III, Fig. 3)
  • Lowest average ATE on the six Vicon-room sequences (2.410 deg / 0.113 m, five runs each) against OpenVINS-IC, Proposed-LI, LOAM, LIO-MAP and LIC-Fusion (Table VI); no failure on the seven teaching-building sequences, where the authors claim higher accuracy on most sequences (Sec. VI-A, Table V)
  • With online calibration from perturbed initial values, simulated ATE and NEES stay close to the true-calibration case (0.129 deg / 0.021 m vs 0.118 deg / 0.020 m) (Table III)

限制

  • The LiDAR-IMU subsystem (Proposed-LI) exceeds 30 m final drift on teaching-building Seq 3 and Seq 6, where long corridors show only parallel planes (Sec. VI-A, Table V)
  • Start-to-end drift is not uniformly smallest: LIO-MAP has smaller drift in every printed component on Seq 1 and Seq 7, and LOAM on Seq 2 (Table V)
  • Evaluation limited to indoor trajectories of about 34 to 237 m with start-to-end drift or Vicon ATE; no map accuracy assessment (Sec. VI, Tables V-VI)
  • Edge-feature tracking left for future work (Sec. VII)

營建工程相關證據

室內測試於浙江大學教學大樓（完工建物）與 Vicon 室；報告長走廊僅見平行平面時的退化，對工地走廊類場景具參考意義（推論），但無工地驗證。

原文驗證環境：模擬、受控實驗、已完工建築

報告的性能數據

以下是原文作者報告的性能數值（author-reported results），不是本研究重新量測的結果。每張圖只並列同一個比較組（comparison group，同一張表、同一組實驗設定）內的方法；不同比較組之間的數值不可直接比較，也不構成排名。

本方法共出現在 5 個比較組，合計 45 筆紀錄。以下列出本方法紀錄最多的 4 組，其餘 1 組列在最後，並連到性能比較頁。

Zuo et al., 2020 · Table III 本方法 16 筆

資料集與序列simulation (virtual room) · 12 Monte-Carlo runs

表格設定（擷取紀錄原文）：Monte-Carlo simulation, 12 runs averaged; virtual room with structural planes; 'true' starts from ground-truth calibration, 'bad' from perturbed calibration; IC = IMU-camera subsystem only (Table II settings: LiDAR 7 Hz, camera 10 Hz, IMU 200 Hz, LiDAR point noise 0.03 m) (Zuo et al., 2020, Table III)

ATE (deg), averaged over 12 runs，simulation (virtual room) · 12 Monte-Carlo runs

只並列這張表在相同設定下報告的方法；以「本方法：」開頭者為本頁方法。失敗、未執行與未報告以標記呈現，不是 0。

這些是 Zuo et al., 2020 在此表設定下報告的數值（author-reported results），只能在同一個比較組內對照，不代表方法在其他資料或設定下的表現。

統計量：原文未報告；對齊方式：原文未報告；單位：deg；場景：simulated indoor room

資料來源作者報告值(Zuo et al., 2020, Table III)

數值與出處
方法（原文寫法）報告值出處
LIC-Fusion 2.0 (true w/ calib)本方法原文提出0.118 deg(Zuo et al., 2020, Table III)
LIC-Fusion 2.0 (bad w/ calib)本方法原文提出0.129 deg(Zuo et al., 2020, Table III)
LIC-Fusion 2.0 (bad w/o calib)本方法原文提出0.148 deg(Zuo et al., 2020, Table III)
LIC-Fusion 2.0 (true w/o calib)本方法原文提出0.122 deg(Zuo et al., 2020, Table III)
IMU-camera subsystem (IC true w/o calib)0.159 deg(Zuo et al., 2020, Table III)

Zuo et al., 2020 · Table VI 本方法 14 筆

表格設定（擷取紀錄原文）：Averaged ATE of 5 runs on 6 Vicon-room sequences (cluttered room, Vicon ground truth), orientation (deg) and position (m); ATE computed following Zhang and Scaramuzza [23]; '-' = translational error above 20 m. The Average column is printed by the authors (for LIO-MAP and LIC-Fusion it averages only the sequences that did not fail). (Zuo et al., 2020, Table VI)

averaged ATE, orientation (deg)，Vicon Room sequences (authors' data) · Seq 1 (42.62 m)

只並列這張表在相同設定下報告的方法；以「本方法：」開頭者為本頁方法。失敗、未執行與未報告以標記呈現，不是 0。

  • 失敗

這些是 Zuo et al., 2020 在此表設定下報告的數值（author-reported results），只能在同一個比較組內對照，不代表方法在其他資料或設定下的表現。

統計量：原文未報告；對齊方式：原文未報告；單位：deg；場景：indoor Vicon motion-capture room (cluttered)

資料來源作者報告值(Zuo et al., 2020, Table VI)

數值與出處
方法（原文寫法）報告值出處
LIC-Fusion 2.0本方法原文提出2.537 deg(Zuo et al., 2020, Table VI)
OpenVINS-IC2.625 deg(Zuo et al., 2020, Table VI)
Proposed-LI2.333 deg(Zuo et al., 2020, Table VI)
LOAM5.88 deg(Zuo et al., 2020, Table VI)
LIO-MAP無數值失敗註記（擷取紀錄）：failed(Zuo et al., 2020, Table VI)
LIC-Fusion2.345 deg(Zuo et al., 2020, Table VI)

Zuo et al., 2020 · Table V 本方法 7 筆

資料集與序列Teaching Building sequences (authors' data) · Seq 1 (about 108 m)

表格設定（擷取紀錄原文）：Averaged start-to-end drift of 5 runs on 7 teaching-building sequences (Zhejiang University) that start and end at the same position; each cell printed as a three-component vector in metres without axis labels; '-' = severe failure with final drift norm larger than 30 m. No scalar is printed, so value is null and the printed vector is kept in metric_as_written. (Zuo et al., 2020, Table V)

averaged start-to-end drift error vector as printed: (0.213, 0.074, 0.338) m，Teaching Building sequences (authors' data) · Seq 1 (about 108 m)

這張表在此指標與資料序列只列出本方法一筆，沒有可並列的其他方法，因此不畫圖，數值與出處見下表。這是 Zuo et al., 2020 在此表設定下報告的數值（author-reported results），不代表方法在其他資料或設定下的表現。

統計量：平均值（mean）；對齊方式：未對齊；單位：m；場景：indoor teaching building (completed building), corridors and stairs

數值與出處
方法（原文寫法）報告值出處
LIC-Fusion 2.0本方法原文提出無數值未報告註記（擷取紀錄）：other: three-component vector printed, no scalar reported(Zuo et al., 2020, Table V)

Lv et al., 2023 · Table VII 本方法 6 筆

指標APE (RMSE, meter)

表格設定（擷取紀錄原文）：Self-collected Vicon Room dataset (indoor, handheld random motion, same rig as YQ: 16-beam LiDAR, camera, IMU); motion-capture ground truth; APE RMSE in metres; all without loop closure (Lv et al., 2023, Table VII)

APE (RMSE, meter)，CLIC Vicon Room dataset (authors) · Seq1 (43 m)

只並列這張表在相同設定下報告的方法；以「本方法：」開頭者為本頁方法。失敗、未執行與未報告以標記呈現，不是 0。

這些是 Lv et al., 2023 在此表設定下報告的數值（author-reported results），只能在同一個比較組內對照，不代表方法在其他資料或設定下的表現。

統計量：均方根誤差（RMSE）；對齊方式：原文未報告；單位：m；場景：indoor motion-capture room, handheld

資料來源作者報告值(Lv et al., 2023, Table VII)

數值與出處
方法（原文寫法）報告值出處
LVI-SAM (w/o loop)0.393 m(Lv et al., 2023, Table VII; Sec. VI-C)
LIC-Fusion 2.0 (w/o loop)本方法0.033 m(Lv et al., 2023, Table VII; Sec. VI-C)
CLIC (w/o loop)原文提出0.08 m(Lv et al., 2023, Table VII; Sec. VI-C)

其他比較組

列出其餘 1 個比較組

來源

  • Zuo et al., 2020

    Xingxing Zuo, Yulin Yang, Patrick Geneva, Jiajun Lv, Yong Liu, Guoquan Huang, Marc Pollefeys（2020）LIC-Fusion 2.0: LiDAR-Inertial-Camera Odometry with Sliding-Window Plane-Feature Tracking2020 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), pp. 5112-5119

    DOI 10.1109/iros45743.2020.9340704arXiv 2008.07196

    同儕審查已出版已讀全文近十年

回到方法圖鑑

SLAM 點雲圖鑑

本站為綜述論文《SLAM-Based Point Clouds for Construction: A Review》研究過程的互動說明文件。互動圖為教學模擬，不代表任何特定系統的實測表現；方法描述、數值與結論以所引原始文獻為準。

內容基準日 2026-09-25 · 研究與網站製作使用 AI 輔助工具，見 證據標記與方法

選擇開啟Esc關閉