Tunnel localizability with LiDAR and UWB
作者把 LiDAR 在先驗地圖中定位的問題寫成一組點落在局部平面上的約束，計算量測距離對位置與姿態擾動的敏感度，分別堆疊成代表力的矩陣 F 與代表力矩的矩陣 T，並把特徵分解後各軸上累積的「虛擬力與力矩」大小定義為可定位性；這個觀點類比於操作力學中無摩擦的力封閉。UWB 測距只提供指向錨點方向的一個力，可補足隧道長軸方向的退化。定位部分以誤差狀態卡爾曼濾波融合 IMU、LiDAR 對地圖的位姿量測，以及由高斯粒子濾波把 UWB 距離轉成的位置量測。作者在卡內基美隆大學 35 m 長的 Smith Hall 隧道中，以改裝的 DJI M100 無人機定性比較有無 UWB 的定位結果。
本頁內容
Localizability model that measures how sensitive LiDAR ranges are to pose perturbations (virtual forces and torques, analogous to force closure) and shows that a single UWB range complements LiDAR along a straight tunnel; an ESKF fuses IMU, scan-to-map LiDAR poses and particle-filtered UWB positions, demonstrated qualitatively on a DJI M100 in a 35 m tunnel.
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 | rotating 2D LiDAR (Hokuyo UTM-30LX-EW on a motor rotating 180 deg/s)、IMU (Microstrain, 100 Hz)、UWB ranging (Pozyx target board on the robot, one anchor in the tunnel) |
|---|---|
| 原文測試平台 | UAV (customized DJI Matrice 100, manually flown at about 0.7 m/s) |
| 狀態估計 | Error-state Kalman filter: IMU propagation, 6D pose measurements from matching laser scans to the prior map, and UWB ranges converted to 3D position measurements by a Gaussian particle filter and inverse Kalman update (Sec. III-B) |
| 資料關聯 | laser scans matched to the prior map (details deferred to earlier work); localizability model uses plane normals fitted to 20 nearest neighbours of sampled points (Sec. III, IV-B) |
| 時間表示 | discrete filter updates at sensor-specific rates |
| 去畸變 | laser scans projected into the robot body frame using motor encoder angles (Sec. IV-A); motion distortion handling not described |
| 迴圈閉合 | none (localization in a prior map) |
| 全域最佳化 | none |
| 地圖表示 | prior point-cloud map of the tunnel built by aligning multiple local scans with ICP (Sec. IV-A, Fig. 8) |
| 先驗資訊 | prior tunnel map and the surveyed UWB anchor position in that map |
| 可輸出幾何 | robot trajectory in the prior map; reconstructed map assembled from scans with estimated poses (qualitative) |
| 計算需求 | DJI Manifold on-board computer (2.32 GHz); runtime not reported |
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 | Hokuyo UTM-30LX-EW (rotating)歸入：Hokuyo UTM-30LX | 方法輸入 | 未標示 | 40 Hz, 30 m range; mounted on a motor rotating continuously at 180 deg/s; scans projected with encoder angles | (Zhen & Scherer, 2019, Sec. IV-A) |
| 慣性量測單元（IMU） | Microstrain IMU (model not reported)歸入：MicroStrain IMU (model not reported) | 方法輸入 | 未標示 | 100 Hz | (Zhen & Scherer, 2019, Sec. IV-A) |
| UWB 測距 | Pozyx UWB target board | 方法輸入 | 未標示 | 100 Hz, 100 m range with clear line of sight; one anchor board placed on the tunnel floor at a position measured in the prior map | (Zhen & Scherer, 2019, Sec. IV-A, IV-C; Fig. 5) |
| 載具平台 | customized DJI Matrice 100 quadrotor | 方法輸入 | 未標示 | manually flown from the map origin to the far end of the tunnel at about 0.7 m/s; GPS, compass and gimbal camera not used | (Zhen & Scherer, 2019, Sec. IV-A, IV-C) |
| 運算硬體 | DJI Manifold computer | 執行運算平台 | 未標示 | 2.32 GHz | (Zhen & Scherer, 2019, Sec. IV-A) |
作者報告的優勢與限制
優勢
- Localizability metric with a physical interpretation (accumulated virtual forces and torques) that identifies degenerate directions in a prior map (Sec. III-A)
- Model predicts the tunnel's weak x-direction and roll constraints and shows that UWB compensates the x-direction (Sec. IV-B, Fig. 6)
- With one UWB anchor fused, the UAV localized throughout the tunnel flight, whereas LiDAR-only localization drifted shortly after take-off (Sec. IV-C)
限制
- Localization accuracy was assessed only qualitatively through the reconstructed map, as no motion capture was available (Sec. IV-C)
- Localizability in x drops near the single UWB anchor, which is a singular point; more anchors would be needed (Sec. IV-B)
- How to combine localizability of several sensor modalities remains unclear (Sec. V)
- Requires a prior map and known UWB anchor positions, an overhead for exploration (Sec. V)
營建工程相關證據
以隧道巡檢機器人為應用背景，驗證在一段 35 m x 2.4 m x 2.5 m、兩側有管線的校園設施隧道中進行；提供的是幾何退化的量化模型與 UWB 補強概念，沒有軌跡或地圖精度數值（Sec. IV）。
原文驗證環境：地下或隧道
報告的性能數據
性能數據仍在分批查證，目前尚未收錄此方法的報告值。
來源
Zhen & Scherer, 2019
（2019）Estimating the Localizability in Tunnel-like Environments using LiDAR and UWB2019 International Conference on Robotics and Automation (ICRA), pp. 4903-4908
同儕審查已出版已讀全文近十年查證後修正