Minimal C laser SLAM using a 1 cm grid of 'hole' functions around obstacle hits, scan-to-map scoring by summing map values at scan endpoints, and Monte Carlo search or particle filtering for pose estimation.

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。

tinySLAM (CoreSLAM) 的技術屬性
感測輸入2D laser scanner (Hokuyo URG-04LX)、wheel odometry (two free odometry wheels with 2000-point encoders)、GPS and compass optionally fused in the particle filter (the compass without good success)
原文測試平台wheeled UGV (MinesRover, six-wheel rocker-bogie)
狀態估計stand-alone: simple Monte Carlo search of the pose that best matches the scan to the map; or particle filter in which the scan-to-map distance is each particle's likelihood, with a slippage model (10% of particles stay in place with high noise) (Sec. IV)
資料關聯no explicit correspondences: the scan-to-map distance sums the map values under the transformed scan endpoints (Algorithm 2)
時間表示discrete poses
去畸變each scan corrected with a constant longitudinal and rotational speed during the sweep (Sec. III)
迴圈閉合none explicit
全域最佳化none
地圖表示2D grid of 2048 x 2048 16-bit cells at 1 cm per cell; each obstacle hit is drawn as a 'hole' function with its tip at the obstacle using a modified Bresenham ray update and an integration-speed (quality) parameter (Sec. IV; Algorithms 1, 3, 4)
先驗資訊none (a full prior map can be loaded for relocalization in the particle-filter version)
可輸出幾何2D grid map and robot trajectory
計算需求particle filter run on a desktop PC over a wireless link; the robot's QWERK module (ARM9 with FPGA, 200 MHz) handles sensors and actuators; SLAM runtime not reported (Sec. III; Fig. 3)

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。

原文使用的設備
類別型號(原文寫法)角色資料集原文規格出處
LiDARHokuyo URG-04LX方法輸入未標示10 Hz horizontal scan over about 240 deg; maximum range limited to 4 m (another sentence states 5 m); USB(Steux & Hamzaoui, 2010, Sec. III; Fig. 1 caption ('HOKUYO URG-04'))
輪式或腿式里程計2000 points encoders方法輸入未標示on the two free-rotating odometry wheels(Steux & Hamzaoui, 2010, Fig. 2 caption)
載具平台MinesRover方法輸入未標示six wheels: four driving and steering, two free odometry wheels; rocker-bogie; 14.8 V 4.1 Ah LiPo; four 45 W motors; top speed 3 m/s(Steux & Hamzaoui, 2010, Sec. III; Figs. 1-2)

作者報告的優勢與限制

優勢

限制

營建工程相關證據

未在營建場域驗證;實驗只在 Mines ParisTech 機電實驗室進行。其極簡程式與低成本短距雷射的組合,對教學或低成本 2D 平面建圖有參考價值,但缺乏定量精度證據,不宜作為工程量測依據(推論)。

原文驗證環境:已完工建築

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

  • Steux & Hamzaoui, 2010

    Bruno Steux, Oussama El Hamzaoui(2010)tinySLAM: A SLAM algorithm in less than 200 lines C-language program2010 11th International Conference on Control, Automation, Robotics and Vision (ICARCV 2010), Singapore, pp. 1975-1979

    同儕審查已出版已讀全文經典

回到方法圖鑑

選擇開啟Esc關閉