LIO-SAM
LIO-SAM 把 LiDAR 慣性里程計建構在因子圖(factor graph)上,以 iSAM2 增量最佳化 IMU 預積分、LiDAR 里程計、GNSS 與迴圈閉合四種因子,形成緊耦合(tightly-coupled)系統。IMU 積分的運動用來對點雲去畸變並提供掃描配準初值,LiDAR 里程計結果再回饋估計 IMU 偏差。為維持即時性,新關鍵影格只與固定數量的近期子關鍵影格(sub-keyframes)組成的局部體素地圖配準,而非與整張全域地圖配準;迴圈以歐氏距離搜尋候選並以掃描配準建立約束。
本頁內容
LIO-SAM fuses IMU preintegration, LOAM-style keyframe scan matching against a sliding local map, optional GPS and loop-closure factors in an iSAM2 factor graph, using the IMU for de-skewing and initial guesses.
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)、IMU (MicroStrain 3DM-GX5-25)、GNSS (Reach M, optional) |
|---|---|
| 原文測試平台 | handheld、wheeled UGV (Clearpath Jackal)、boat (Duffy 21) |
| 狀態估計 | factor graph smoothing with iSAM2: IMU preintegration, lidar odometry, GPS and loop-closure factors; scan matching solved by Gauss-Newton (Sec. III) |
| 資料關聯 | LOAM-style edge/planar features matched to a local voxel map built from n=25 recent sub-keyframes (point-to-edge / point-to-plane) (Sec. III-C) |
| 時間表示 | discrete keyframe states (keyframe added at 1 m or 10 deg pose change) (Sec. III-A, III-C) |
| 去畸變 | IMU-estimated nonlinear motion de-skews each scan and gives the initial guess for scan matching (Sec. I, abstract) |
| 迴圈閉合 | Euclidean-distance candidate search within 15 m, scan matching of the new keyframe to +/-12 sub-keyframes around the candidate, added as a loop factor; compatible with descriptor-based place recognition (Sec. III-E) |
| 全域最佳化 | incremental factor-graph optimization (iSAM2) including GPS factors when estimated position covariance exceeds GPS covariance (Sec. III-D) |
| 地圖表示 | keyframe edge/planar feature clouds; local voxel maps downsampled at 0.2 m (edge) and 0.4 m (planar) (Sec. III-C) |
| 先驗資訊 | GNSS (optional); no prior map |
| 可輸出幾何 | global feature map assembled from keyframe edge and planar feature clouds plus the keyframe trajectory; lidar frames between keyframes (1 m or 10 deg pose change) are discarded; maps are shown aligned with Google Earth imagery (Figs. 4 to 7); dense map export not described |
| 計算需求 | CPU only, Intel i7-10710U laptop; mapping runtime per scan 41.9-100.5 ms across datasets; up to 13x real-time playback in stress tests (Table IV) |
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 |
|---|---|---|---|---|---|
| LiDAR | Velodyne VLP-16 | 方法輸入 | 未標示 | 10 Hz rotation rate in the runtime discussion | (Shan et al., 2020, Sec. IV; Sec. IV-F) |
| 慣性量測單元(IMU) | MicroStrain 3DM-GX5-25 | 方法輸入 | 未標示 | 原文未報告 | (Shan et al., 2020, Sec. IV) |
| GNSS 接收器 | Reach M | 方法輸入 | 未標示 | optional GPS factor; also the RMSE reference in the Park dataset | (Shan et al., 2020, Sec. IV; Sec. IV-F) |
| 載具平台 | custom-built handheld device | 方法輸入 | 未標示 | Rotation, Walking and Campus datasets on the MIT campus | (Shan et al., 2020, Sec. IV; Fig. 2a) |
| 載具平台 | Clearpath Jackal | 方法輸入 | 未標示 | UGV without suspension; Park dataset on a forested hiking trail | (Shan et al., 2020, Sec. IV; Sec. IV-D; Fig. 2b) |
| 載具平台 | Duffy 21 | 方法輸入 | 未標示 | electric boat; Amsterdam canal dataset (about 3 h) | (Shan et al., 2020, Sec. IV; Sec. IV-E; Fig. 2c) |
| 運算硬體 | laptop with Intel i7-10710U | 執行運算平台 | 未標示 | CPU only, no parallel computing; ROS in Ubuntu | (Shan et al., 2020, Sec. IV) |
作者報告的優勢與限制
優勢
- End-to-end translation error when returning to start: 0.12 m (Campus), 0.04 m (Park), 0.17 m (Amsterdam), with LOAM and LIOM failing or drifting in the same tests (Sec. IV-C to IV-E, Table II)
- Park dataset RMSE w.r.t. GPS (x-y only) 0.96 m with loop closure and partial GPS use (Sec. IV-F, Table III)
- Runtime depends mainly on feature map density rather than graph size (Sec. IV-F)
限制
- Without absolute measurements or loop closures, lidar-inertial odometry still drifts over long durations (Sec. III-D, V)
- GPS elevation was very inaccurate (altitude errors approaching 100 m without loop closures) (Sec. III-E)
- Sunlight-induced false lidar returns and degenerate scenes under bridges caused other methods to fail in the Amsterdam test (Sec. IV-E)
- Accuracy evidence is limited to end-to-start return errors and an x-y RMSE against partially used GPS; no independent survey-grade trajectory or map reference (Sec. IV-C to IV-F)
- Follow-up work states LIO-SAM needs nine-axis IMU readings for de-skewing and front-end odometry and could not be run on a six-axis-IMU dataset (Li et al., 2021b, Sec. 2, 5.2)
營建工程相關證據
原論文在 MIT 校園手持、公園 UGV 與阿姆斯特丹運河船載資料中測試,未含營建工地。Feng 等人(Feng et al., 2025)在施工中醫院門診大樓(RS-Helios-16P 與九軸 IMU 的施工機器人平台)以預設參數測試,作者報告 LIO-SAM 在十種方法中實際工地 APE RMSE 最低(2.26 m,Sec. 5.3, Table 4);在模擬工地中,作者指出其前視圖仍可見垂直漂移造成的結構傾斜(Sec. 5.2, Fig. 13),並於結論指出緊耦合 IMU、地面約束與迴圈閉合不能完全消除垂直漂移。論文未說明實際工地 APE 所用參考軌跡的來源(全文僅描述 Gazebo 模擬的真實軌跡外掛),故實際工地 APE 只能視為作者報告值,不能當作已驗證的幾何精度。
原文驗證環境:跨場域、獨立參考量測
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 75 個比較組,合計 582 筆紀錄。以下列出本方法紀錄最多的 4 組,其餘 71 組列在最後,並連到性能比較頁。
Wu et al., 2024a · Table I 本方法 32 筆
表格設定(擷取紀錄原文):Default parameters for FAST-LIO2 and LIO-SAM, LIO-SAM loop closure disabled; one LIO-EKF configuration for all data; KITTI relative errors and ATE; LIO-SAM not run on Newer College (needs IMU attitude) (Wu et al., 2024a, Table I)
Avg. tra. (KITTI relative translation error),UrbanNav · 20210517
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。
按 Tab 進入圖表後,用上下方向鍵逐一瀏覽各類別,Esc 關閉提示框;也可開啟表格檢視閱讀全部數值。
這些是 Wu et al., 2024a 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。
資料來源作者報告值(Wu et al., 2024a, Table I)
| 方法(原文寫法) | 報告值 | 出處 |
|---|---|---|
| FAST-LIO2 | 4.11% | (Wu et al., 2024a, Table I) |
| LIO-SAM本方法 | 3.18% | (Wu et al., 2024a, Table I) |
| LIO-EKF原文提出 | 3.2% | (Wu et al., 2024a, Table I) |
Lin & Zhang, 2024 · Table III 本方法 26 筆
表格設定(擷取紀錄原文):VoR Table III: absolute position error (APE, m) with standard deviation on NCLT (front-facing camera and 3D LiDAR, Segway robot), computed on the odometry output at LiDAR input for every method; loop closure of LIO-SAM and LVI-SAM deactivated; photometric calibration disabled for R3LIVE++ (unavailable for NCLT); '-' = failed midway, excluded from the averages; Our_LIO column omitted. The text says 2012-03-17 and 2012-08-04 were excluded for a 100 ms LiDAR-IMU timestamp delay, yet both appear in the 25-row table. (Lin & Zhang, 2024, Table III)
APE (m),NCLT · 2012-01-08 (6495.7 m, 01:25:35)
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。
按 Tab 進入圖表後,用上下方向鍵逐一瀏覽各類別,Esc 關閉提示框;也可開啟表格檢視閱讀全部數值。
這些是 Lin & Zhang, 2024 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。
資料來源作者報告值(Lin & Zhang, 2024, Table III)
| 方法(原文寫法) | 報告值 | 出處 |
|---|---|---|
| Our (R3LIVE++)原文提出 | 10.8 m原文指標寫法:APE (m), printed with STD 2.7 m | (Lin & Zhang, 2024, VoR Table III) |
| R2LIVE | 22.4 m原文指標寫法:APE (m), printed with STD 3.5 m | (Lin & Zhang, 2024, VoR Table III) |
| LVI-SAM | 23.4 m原文指標寫法:APE (m), printed with STD 3.7 m | (Lin & Zhang, 2024, VoR Table III) |
| FAST-LIVO | 13.4 m原文指標寫法:APE (m), printed with STD 2.9 m | (Lin & Zhang, 2024, VoR Table III) |
| Fast-LIO2 | 18.5 m原文指標寫法:APE (m), printed with STD 3.3 m | (Lin & Zhang, 2024, VoR Table III) |
| LIO-SAM本方法 | 21.7 m原文指標寫法:APE (m), printed with STD 3.6 m | (Lin & Zhang, 2024, VoR Table III) |
Huang et al., 2024b · Table II 本方法 22 筆
指標translation RMSE
表格設定(擷取紀錄原文):M2DGR; translation RMSE of ATE; loop closure disabled; map and scan downsampling 0.4 m; first and last 100 s of street07 and street10 discarded (RTK instability); LOG-C ablation column not stored (Huang et al., 2024b, Table II)
translation RMSE,M2DGR · gate01 (172 s)
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。
按 Tab 進入圖表後,用上下方向鍵逐一瀏覽各類別,Esc 關閉提示框;也可開啟表格檢視閱讀全部數值。
這些是 Huang et al., 2024b 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。
資料來源作者報告值(Huang et al., 2024b, Table II)
| 方法(原文寫法) | 報告值 | 出處 |
|---|---|---|
| LOG-LIO原文提出 | 0.097 m | (Huang et al., 2024b, Table II) |
| FAST-LIO2 | 0.091 m | (Huang et al., 2024b, Table II) |
| LIO-SAM本方法 | 0.122 m | (Huang et al., 2024b, Table II) |
Koide et al., 2024 · Table V 本方法 20 筆
表格設定(擷取紀錄原文):Multi-Camera Newer College (Ouster OS0-128, Alphasense Core); translational ATE; unlabeled rows are the no-loop-closure variant of the method printed in the next row (checked against the PDF layout and the ablation baseline in Table VIII) (Koide et al., 2024, Table V)
Absolute Trajectory Error [m] (no loop closure),Multi-Camera Newer College · quad-easy
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。
按 Tab 進入圖表後,用上下方向鍵逐一瀏覽各類別,Esc 關閉提示框;也可開啟表格檢視閱讀全部數值。
這些是 Koide et al., 2024 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。
資料來源作者報告值(Koide et al., 2024, Table V)
| 方法(原文寫法) | 報告值 | 出處 |
|---|---|---|
| LINS [3] | 0.16 m | (Koide et al., 2024, Table V) |
| LIO-SAM [10] (without loop closure; unlabeled row above LIO-SAM)本方法 | 0.086 m | (Koide et al., 2024, Table V) |
| FAST-LIO2 [5] | 0.068 m | (Koide et al., 2024, Table V) |
| CLINS [11] (without loop closure; unlabeled row above CLINS) | 0.197 m | (Koide et al., 2024, Table V) |
| DLO [14] | 0.08 m | (Koide et al., 2024, Table V) |
| GLIM (odometry, without loop closure; unlabeled row above GLIM)原文提出 | 0.07 m | (Koide et al., 2024, Table V) |
其他比較組
列出其餘 71 個比較組
- Huang et al., 2024b · Table III
- Bai et al., 2022 · Table II
- Nguyen et al., 2023 · Table I
- Jiao et al., 2022 · Table IV
- Stührenberg & Smarsly, 2025 · Table 5
- Stührenberg & Smarsly, 2025 · Table 6
- Palieri et al., 2021 · Table II
- Liu et al., 2026 · Table 2 (full SLAM with LC)
- Liu et al., 2026 · Table 2 (odometry without LC)
- Lv et al., 2021 · Table II
- Xu et al., 2022 · Table IV
- Shan et al., 2021 · Table II
- Yarovoi & Cho, 2024 · Table 2
- Yarovoi & Cho, 2024 · Table 4
- Pfreundschuh et al., 2024 · Table II
- Wang et al., 2023b · Table III
- Shan et al., 2020 · Table IV
- Stührenberg & Smarsly, 2025 · Table 7
- Lv et al., 2023 · Table III
- Shan et al., 2020 · Table II
- Nguyen et al., 2022a · Table 4
- Lv et al., 2021 · Table IV
- Pfreundschuh et al., 2024 · Table I
- Chen et al., 2023 · Table II
- Zhao et al., 2021 · Table II
- Ramezani et al., 2022 · Table II
- Xu et al., 2022 · Table V
- He et al., 2023a · Table 6
- Wu et al., 2024b · Table I
- Yan et al., 2026a · Table 1
- Lv et al., 2023 · Table IV
- Chen et al., 2023 · Table I
- Liu et al., 2024 · Table I
- Liu et al., 2024 · Table II
- Li et al., 2021b · Table 1
- Nguyen et al., 2023 · Table III
- Zheng & Zhu, 2024 · Table III
- Feng et al., 2025 · Table 3
- Feng et al., 2025 · Table 4
- Koide et al., 2024 · Table I
- He et al., 2023a · Table 5
- Nguyen et al., 2023 · Table II
- Lv et al., 2023 · Table V
- Wang et al., 2023b · Table IV
- Wang et al., 2023b · Table V
- Qian et al., 2021 · Table III
- Zhao et al., 2021 · Table III
- Frosi & Matteucci, 2022 · Table II
- Frosi & Matteucci, 2022 · Table III
- Frosi & Matteucci, 2022 · Table IV
- Chen et al., 2026 · Table 2
- Tang et al., 2023 · Table I
- Tang et al., 2023 · Table II
- Liu et al., 2023b · Table VIII
- Lee et al., 2024b · Table 4
- Wu et al., 2024a · Table II
- Shan et al., 2020 · Table III
- Palieri et al., 2021 · Table III
- Qian et al., 2021 · Table IV
- Yarovoi & Cho, 2024 · Table 3
- Ebadi et al., 2024 · Text Sec. IV-B
- He et al., 2023a · Table 7
- Lin & Zhang, 2024 · Text Sec.VI-E
- Qian et al., 2021 · Table V
- Ramezani et al., 2022 · Text Sec.VI-C
- Ghadimzadeh Alamdari et al., 2025 · Table 3
- Ghadimzadeh Alamdari et al., 2025 · Text Sec.7.1.2
- Koide et al., 2024 · Table X
- Liu et al., 2023b · Table IV
- Jiao et al., 2022 · Text Sec.V
- Trzeciak et al., 2023 · Text Practical application
來源
Shan et al., 2020
(2020)LIO-SAM: Tightly-coupled Lidar Inertial Odometry via Smoothing and Mapping2020 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), pp. 5135-5142
DOI 10.1109/iros45743.2020.9341176arXiv 2007.00258程式碼
同儕審查已出版已讀全文近十年查證後修正
相關版本
- 預印本:arXiv 2007.00258 (v1 2020-07-01, v3 2020-07-14) https://arxiv.org/abs/2007.00258
- 程式碼釋出:TixiaoShan/LIO-SAM https://github.com/TixiaoShan/LIO-SAM
程式碼:https://github.com/TixiaoShan/LIO-SAM(授權:BSD 3-Clause (LICENSE file))。有公開程式碼不等於已被重現,也不代表目前版本與論文版本相同。