整合之字引導自動環境探索策略之虛擬柱狀障礙物建圖架構於戶外移動機器人覆蓋式任務

Virtual Pole-Like Obstacle Concept Based Mapping Framework Integrating Zigzag-Driven Autonomous Exploration for Outdoor Mobile Robot Coverage Task

莊佳芸(Chia-Yun Chuang)

指導教授 ： 連豊力

國立臺灣大學/電機資訊學院/電機工程學研究所/碩士(2022年)

https://doi.org/10.6342/NTU202201894

摘要

機器人要在未知環境中執行覆蓋任務，它必須能夠識別障礙物的位置並在該區域內安全導航而不會發生碰撞。此類的應用廣泛，舉凡從一般家庭中的掃地機器人、除草型機器人，到水下探索，因此逐漸被重視。
本論文提出了用於室外移動機器人覆蓋任務的建圖架構，其基於虛擬柱狀障礙物概念，並整合之字形引導自主探索，試圖在合理的時間於給定區域內探索並建立地圖。我們將機器人割草任務分為兩個獨立的階段，學習階段和割草階段。在學習階段，割草機估計負邊界，即地標的割草邊界，並定義正邊界，即期望的割草範圍。在割草階段，使用從學習階段生成的地標圖估計割草機的位置。
我們特別感興趣的是確保地標估測在學習階段不會失敗並且機器人能自主在未知環境中安全的導航。主要脈絡為建立佔用柵格圖並將特徵提取到固定的全局坐標中。全局特徵可提供路徑規劃器障礙物信息，以避免碰撞，或提供操作員直觀的視覺化。這些地圖是即時建立和更新的，並且在機器人上的單板電腦完成運算。
受人為手動遙控探勘的啟發，我們提出了具有最佳視點控制器的目標引導探索。探索由人工勢場中的重力驅動，而高層級決策由有限狀態機處理。我們系統的可行性通過各種模擬和割草機的實驗得到驗證。在模擬中，分析了各種樹配置的性能以及與基於前沿探索策略的比較。在實驗中，評估建圖結果和覆蓋任務的框架。從定量和定性的比較表明，所提出的方法較適合割草機的工作場域。

關鍵字

柵格地圖 ； 深度影像同步定位與地圖建構 ； 即時系統 ； 覆蓋式路徑規劃 ； 機器人視覺 ； 柱狀特徵 ； 自動探索

並列摘要

For a mobile robot to execute coverage tasks in unknown environments, it must be able to identify locations of obstacles and to safely navigate in the area without collisions. With its wide applications, from sweeping robots and mowers in daily life to underwater exploring and so on, this technology has received extensive attention.
This thesis presents a mapping framework for outdoor mobile robot coverage task which bases on virtual pole-like obstacle concept and integrates zigzag-driven autonomous exploration. It attempts to explore and map a given area in a reasonable amount of time. We divide the robotic mowing task into two separate phases, a learning phase and a mowing phase. During the learning phase, the mower estimates the negative boundary, which is the mowing border of landmarks, and defines the positive boundary, which is the desired mowing range. During the mowing phase, the landmark map generated from the learning phase is used to estimate the location of the mower.
Worth paying attention to our work is ensuring that the mapping framework does not fail and that robots can navigate autonomously and safely in unknown environments during learning phase. The main idea is to create an occupancy grid map and extract features into a fixed global coordinate. The global features can be used as obstacle information for path planners to avoid collisions or intuitive visualization for a human operator. The maps are created and updated in real-time and are built onboard in low-power platform of the robot.
Motivated by manually guiding exploration, we present goal-driven exploration with an optimal viewpoint controller. Exploration is guided by gravity in the artificial potential field and high-level mission decisions are handled by a finite-state machine. The capability of our system is verified by miscellaneous simulations and field experiments with a real mower. In the simulation, the performances of various tree configurations and the comparison with frontier-based method are analyzed. In the physical world, mapping results and the framework for coverage tasks are evaluated. Both quantitatively and qualitatively comparisons of mapping results and exploration performances show that the proposed method suits the mower’s work field well.

並列關鍵字

Occupancy map ； depth image SLAM system ； real-time system ； coverage path planning ； robot vision ； cylinder-like features ； autonomous exploration

參考文獻

Cesar Cadena, Luca Carlone, Henry Carrillo, Yasir Latif, Davide Scaramuzza, Jose Neira, Ian Reid, and John J. Leonard, “Past, Present, and Future of Simultaneous Localization and Mapping: Toward the Robust-Perception Age,” IEEE Transactions on Robotics, Vol. 32, No. 6, pp. 1309-1332, Dec. 2016.

Jian Tang, Yuwei Chen, Antero Kukko, Harri Kaartinen, Anttoni Jaakkola, Ehsan Khoramshahi, Teemu Hakala, Juha Hyyppä, Markus Holopainen, and Hannu Hyyppä, “SLAM-Aided Stem Mapping for Forest Inventory with Small-Footprint Mobile LiDAR,” Forests, Vol 6, No. 12, pp. 4588-4606, Dec. 2015.

Albert E. Patterson, Yang Yuan, and William R. Norris, “Development of User-Integrated Semi-Autonomous Lawn Mowing Systems: A Systems Engineering Perspective and Proposed Architecture,” AgriEngineering, Vol 1, No. 1, pp. 453-474, Sep. 2019.

Mathieu Labbé and François Michaud, “RTAB-Map as an open-source lidar and visual simultaneous localization and mapping library for large-scale and long-term online operation,” Journal of Field Robotics, Vol. 36, No. 2, pp. 416–446, Mar. 2019.

Giorgio Grisetti, Cyrill Stachniss, and Wolfram Burgard, “Improved Techniques for Grid Mapping With Rao-Blackwellized Particle Filters,” IEEE Transactions on Robotics, Vol. 23, No. 1, pp. 34–46, Feb. 2007.