FAST-LIVO
FAST-LIVO 的 LIO 與 VIO 皆採直接法(direct method):光達原始點以點到平面殘差配準到地圖,視覺部分則把影像小區塊(patch)附掛在光達地圖點上,直接以稀疏光度誤差對齊新影像,不擷取、不三角化視覺特徵。兩者在 ESIKF 中緊密耦合,並以遮蔽與深度不連續檢測剔除不穩定地圖點,因此計算成本低且可在 ARM 處理器上即時執行。
本頁內容
A sparse-direct LIVO that reuses LiDAR map points as visual landmarks carrying image patches, fusing point-to-plane and photometric residuals in an ESIKF at low computational cost.
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 (Ouster OS1-16 in NTU-VIRAL; Livox Avia in private data)、IMU、camera |
|---|---|
| 原文測試平台 | UAV (NTU-VIRAL)、custom sensor platform with Livox Avia, two industrial cameras and an onboard DJI Manifold-2C (carrying mode not stated in the paper) |
| 狀態估計 | error-state iterated Kalman filter fusing LiDAR and visual updates (LIO adapted from FAST-LIO2) |
| 資料關聯 | raw LiDAR points with frame-to-map point-to-plane residual; sparse-direct frame-to-map alignment of 8x8 image patches attached to LiDAR map points, with occlusion and depth-discontinuity outlier rejection |
| 時間表示 | discrete poses |
| 去畸變 | backward propagation as in FAST-LIO2 (Sec. III) |
| 迴圈閉合 | none |
| 全域最佳化 | none |
| 地圖表示 | LiDAR global map adopted from FAST-LIO2 (all past points in an ikd-Tree with internal downsampling) plus a separate visual global map of previously observed LiDAR points in equal-size hash-indexed voxels, each point storing several 8x8 patch pyramids with their camera poses |
| 先驗資訊 | Time offsets among LiDAR, IMU and camera assumed known (calibrated or synchronized in advance) and extrinsics pre-calibrated; the private rig is hardware-synchronized by STM32 timers at 10 Hz |
| 可輸出幾何 | real-time dense RGB-colored point cloud (Sec. VI-B4, Fig. 7) |
| 計算需求 | Mean times: VIO 10.23 ms and LIO 26.52 ms on a desktop Intel i7 (text: 8-core i7-10700U), VIO 13.82 ms and LIO 51.51 ms on Qualcomm RB5 (Kryo585); whole system 36.75 ms per LiDAR and image frame versus 45.16 ms plus 59.27 ms for R2LIVE |
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 | OS1 gen1 (16-channel) | 資料集感測器 | NTU-VIRAL | horizontal 16-channel spinning LiDAR | (Zheng et al., 2022, Sec. VI-A) |
| LiDAR | Livox Avia | 方法輸入 | FAST-LIVO private dataset | solid-state LiDAR with built-in IMU | (Zheng et al., 2022, Sec. VI-B1; Fig. 4) |
| 慣性量測單元(IMU) | OS1 internal IMU | 資料集感測器 | NTU-VIRAL | 原文未報告 | (Zheng et al., 2022, Sec. VI-A) |
| 慣性量測單元(IMU) | Livox Avia built-in IMU | 方法輸入 | FAST-LIVO private dataset | 原文未報告 | (Zheng et al., 2022, Fig. 4) |
| GNSS 接收器 | GPS receiver (labelled in Fig. 4) | 資料集感測器 | FAST-LIVO private dataset | 原文未報告 (no GNSS use described) | (Zheng et al., 2022, Fig. 4) |
| 相機 | left camera (model not stated) | 資料集感測器 | NTU-VIRAL | 原文未報告 | (Zheng et al., 2022, Sec. VI-A) |
| 相機 | MV-CA013-21UC | 方法輸入 | FAST-LIVO private dataset | two industrial cameras (left and right) | (Zheng et al., 2022, Sec. VI-B1; Fig. 4) |
| 載具平台 | UAV | 資料集感測器 | NTU-VIRAL | aerial platform of NTU-VIRAL (fast UAV motion noted) | (Zheng et al., 2022, Sec. VI-A) |
| 運算硬體 | DJI manifold-2c歸入:DJI Manifold 2C | 資料集感測器 | FAST-LIVO private dataset | onboard computer of the data rig; Intel i7-8550u CPU and 8 GB RAM | (Zheng et al., 2022, Sec. VI-B1) |
| 運算硬體 | desktop PC with 8-core Intel Core i7-10700U (as written)歸入:desktop PC with 8-core Intel Core i7-10700U | 執行運算平台 | 未標示 | 原文未報告 | (Zheng et al., 2022, Sec. VI-C) |
| 運算硬體 | RB5 with Qualcomm Kryo585 CPU | 執行運算平台 | 未標示 | embedded ARM platform | (Zheng et al., 2022, Sec. VI-C; Table III) |
| 其他 | STM32 synchronized timers | 資料集感測器 | FAST-LIVO private dataset | hardware trigger at 10 Hz for all sensors (PWM to LiDAR and camera) | (Zheng et al., 2022, Sec. VI-B1; Fig. 4) |
作者報告的優勢與限制
優勢
- Lower computation than feature-based LIV systems
- runs on Intel and ARM processors (abstract
- Table III) | Supports both spinning and solid-state LiDARs (Sec. I) | End-to-end error of 0.04 m over a 79.52 m indoor-outdoor path with texture-less wall and aggressive motions (Sec. VI-B3) | End-to-end drift of 0.05 m while moving along a wall about 30 m long, where FAST-LIO2 and SVO2.0 drifted (qualitative comparison) (Sec. VI-B2, Fig. 5)
限制
- Assumes all pixels of a patch share one depth, which reduces affine warping accuracy (reported by successor Zheng et al., 2025, Sec. I)
- No exposure-time handling and no treatment of LiDAR close-range blind zones (Zheng et al., 2025, Sec. I)
- No loop closure (inference from system design)
營建工程相關證據
未報告營建工地或隧道驗證;公開資料為 NTU-VIRAL 無人機序列,另有校園私人資料。
原文驗證環境:公開基準
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 20 個比較組,合計 158 筆紀錄。以下列出本方法紀錄最多的 4 組,其餘 16 組列在最後,並連到性能比較頁。
Lin & Zhang, 2024 · Table III 本方法 26 筆
指標APE (m)
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本方法 | 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 | 21.7 m原文指標寫法:APE (m), printed with STD 3.6 m | (Lin & Zhang, 2024, VoR Table III) |
Yuan et al., 2024 · Table III 本方法 18 筆
指標RMSE of ATE
表格設定(擷取紀錄原文):RMSE of ATE of LIO output versus LiDAR-assisted VIO output within each framework on NTU-VIRAL; 'x' = drifted halfway (Yuan et al., 2024, Table III)
RMSE of ATE,NTU-VIRAL · eee_01
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 Yuan 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(Yuan et al., 2024, Table III)
| 方法(原文寫法) | 報告值 | 出處 |
|---|---|---|
| R3Live (LIO module) | 1.69 m | (Yuan et al., 2024, Table III) |
| R3Live(V) (LiDAR-assisted VIO module) | 1.71 m | (Yuan et al., 2024, Table III) |
| Fast-LIVO (LIO)本方法 | 0.28 m | (Yuan et al., 2024, Table III) |
| Fast-LIVO(V) (LiDAR-assisted VIO)本方法 | 0.3 m | (Yuan et al., 2024, Table III) |
| Ours (SR-LIVO LIO)原文提出 | 0.21 m | (Yuan et al., 2024, Table III) |
| Ours(V) (authors' R3Live-like LiDAR-assisted VIO module, ablation) | 0.24 m | (Yuan et al., 2024, Table III) |
Zheng et al., 2025 · Table II 本方法 17 筆
資料集與序列NTU-VIRAL, Hilti'22, Hilti'23 · Average (25 sequences)
表格設定(擷取紀錄原文):Absolute translational error RMSE on Hilti'22 and Hilti'23 (handheld: PandarXT-32, BMI085; robot: BPearl, MTi-670; front camera), scored through the official Hilti website because ground truth (MoCap or total station) is not public; loop closure of LVI-SAM removed; ablation columns kept only in the Average row; NTU-VIRAL rows omitted for the row cap; Average over all 25 sequences (NTU-VIRAL, Hilti'22, Hilti'23; how failed runs enter the average is not stated) (Zheng et al., 2025, Table II)
absolute translational errors (RMSE), Average row,NTU-VIRAL, Hilti'22, Hilti'23 · Average (25 sequences)
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 Zheng et al., 2025 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(Zheng et al., 2025, Table II)
| 方法(原文寫法) | 報告值 | 出處 |
|---|---|---|
| SDV-LOAM | 7.416 m | (Zheng et al., 2025, Table II (Average row); text Sec. IX-B states 0.044 m for Ours) |
| Our LIO | 0.097 m | (Zheng et al., 2025, Table II (Average row); text Sec. IX-B states 0.044 m for Ours) |
| FAST-LIO2 | 0.151 m | (Zheng et al., 2025, Table II (Average row); text Sec. IX-B states 0.044 m for Ours) |
| R3LIVE | 0.278 m | (Zheng et al., 2025, Table II (Average row); text Sec. IX-B states 0.044 m for Ours) |
| LVI-SAM | 1.928 m | (Zheng et al., 2025, Table II (Average row); text Sec. IX-B states 0.044 m for Ours) |
| FAST-LIVO本方法 | 0.137 m | (Zheng et al., 2025, Table II (Average row); text Sec. IX-B states 0.044 m for Ours) |
| Ours原文提出 | 0.045 m | (Zheng et al., 2025, Table II (Average row); text Sec. IX-B states 0.044 m for Ours) |
| Ours (w/o expo) | 0.051 m | (Zheng et al., 2025, Table II (Average row); text Sec. IX-B states 0.044 m for Ours) |
| Ours (w normal) | 0.044 m | (Zheng et al., 2025, Table II (Average row); text Sec. IX-B states 0.044 m for Ours) |
| Ours (w/o update) | 0.089 m | (Zheng et al., 2025, Table II (Average row); text Sec. IX-B states 0.044 m for Ours) |
Zhou et al., 2025 · Table I 本方法 17 筆
表格設定(擷取紀錄原文):ATE RMSE on 16 Hilti'22 and Hilti'23 sequences computed with the official Hilti evaluation tools; parameters of all methods tuned by the authors; x = system totally failed (Zhou et al., 2025, Table I)
ATE (RMSE),Hilti'22 · Construction Ground
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
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 Zhou et al., 2025 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(Zhou et al., 2025, Table I)
| 方法(原文寫法) | 報告值 | 出處 |
|---|---|---|
| Ours原文提出 | 0.01 m | (Zhou et al., 2025, Table I) |
| FAST-LIVO2 | 0.01 m | (Zhou et al., 2025, Table I) |
| FAST-LIO2 | 0.013 m | (Zhou et al., 2025, Table I) |
| FAST-LIVO本方法 | 0.022 m | (Zhou et al., 2025, Table I) |
| R3LIVE | 0.021 m | (Zhou et al., 2025, Table I) |
| SDV-LOAM | 25.121 m | (Zhou et al., 2025, Table I) |
| LVI-SAM | 無數值失敗註記(擷取紀錄):failed (x) | (Zhou et al., 2025, Table I) |
其他比較組
列出其餘 16 個比較組
- Chen et al., 2025b · Table 5
- Zheng et al., 2022 · Table III
- Lang et al., 2023 · Table III
- Zheng et al., 2022 · Table II
- Yuan et al., 2024 · Table I
- Li et al., 2026b · Text Sec.3.1
- Zheng et al., 2025 · Table III
- Lang et al., 2023 · Table IV
- Boche et al., 2025 · Table IX
- Boche et al., 2025 · Table VI
- Chen et al., 2025b · Text Sec.5.2
- Zheng et al., 2022 · Text Sec.VI-B2
- Zheng et al., 2022 · Text Sec.VI-B3
- Zheng et al., 2022 · Text Sec.VI-C
- Ghadimzadeh Alamdari et al., 2025 · Table 3
- Koide et al., 2024 · Table X
來源
Zheng et al., 2022
(2022)FAST-LIVO: Fast and Tightly-coupled Sparse-Direct LiDAR-Inertial-Visual Odometry2022 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), pp. 4003-4009
DOI 10.1109/iros47612.2022.9981107arXiv 2203.00893程式碼
同儕審查已出版已讀全文近十年
相關版本
- 預印本:FAST-LIVO arXiv v1 (2022-03-02) https://arxiv.org/abs/2203.00893
- 期刊延伸版:FAST-LIVO2: Fast, Direct LiDAR-Inertial-Visual Odometry (states it is developed based on FAST-LIVO) 10.1109/TRO.2024.3502198
- 程式碼釋出:hku-mars/FAST-LIVO https://github.com/hku-mars/FAST-LIVO
程式碼:https://github.com/hku-mars/FAST-LIVO(授權:GPL-2.0 (LICENSE file checked))。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。