行動機器人在動態環境之路徑規畫

本文之主要目的在設計與建立一機器人的自動導航系統，使其能在充滿行人的室內工作。為了降低機器人對人類活動的干擾，並提高行人與機器人本身的安全性，本文開發一套預測式的路徑規畫系統。本文提出一目標導向的行人運動模型，透過估計行人的行進目標預測其未來之軌跡。首先將環境中已知行人軌跡的起迄點進行群聚，即可得到數個可能之目標。再對於每個可能的目標，使用NF1演算法推估行人理想的行進方向，並使用位能場模型表示行人與行人以及機器人間的相互影響。比較推估與觀測的行人行為，即可估計行人的行進目標，進而預測行人未來的軌跡。經實驗證實，本文所提出之運動模型可有效估計行人目標並預測行人路徑。本文進而提出Predictive Anytime RRT 路徑規畫演算法，利用上述的預測模型，在狀態 – 時間空間中搜尋機器人可行的路徑。當行進路線將遭受阻礙時，此演算法可找出令機器人在某段時間下原地等待的路徑。此外，利用改良的距離量度標準提升效率，在複雜的地圖下速度則可達RRT-Blossom的30倍。實驗分為模擬與實作。模擬部分建立一多功能的軟體平台，使用行為庫模擬行人的動態，並物理引擎模擬機器人的運動，再以立體影像呈現路徑規畫與執行結果。在實作上，整合了使用雷射感測器之同步定位地圖建置與追蹤系統。整體系統可在室內環境中進行即時導航。

關鍵字

行人預測；路徑規畫；運動規畫；機器人

並列摘要

The main objective of this thesis is to develop an autonomous navigation system for a mobile robot, which operates in indoor environments among moving people. To reduce distraction to human activities, and to ensure safety, a path planner which predicts human motion is developed. A goal-directed model of pedestrian motion is proposed. Pedestrians are assumed to be moving toward a set of possible destinations, which are extracted from human trajectories collected in the environment. Human motion is then modeled to follow a navigation function to each goal, and interaction between people is modeled with an interaction force model. The probability a new person is going toward each destination is estimated using the motion model. And given that, the future positions of the person can be predicted. The model is shown to capture typical pedestrian motion faithfully. The thesis further develops the Predictive Anytime Rapidly-Exploring Random Tree (PARRT) path planner to find the path of a mobile robot in state-time space. In dynamic environments the algorithm is able to plan in real time. Moreover, with the help of an improved distance metric the planner is faster than RRT-Blossom for 30 times in complex maps. A software platform is developed for both simulation and for real-world navigation, where environment and planning results are visualized in 3D. In real-world implementation a simultaneous localization and mapping (SLAM) with moving object tracking (MOT) module, a global planner using Probabilistic Roadmap (PRM), and a motor control module are integrated. In our experiments, the system is able to navigate in indoor environments in real time.

並列關鍵字

motion prediction ； pedestrian model ； motion planning ； path planning ； rapidly-exploring random tree ； RRT ； robotics

參考文獻

[1] N. M. Amato, O. B. Bayazit, L. K. Dale, C. V. Jones, and D. Vallejo, “OBPRM: An Obstacle-Based PRM for 3D Workspaces,” Proc. Int. Workshop on Algorithmic Foundations of Robotics (WAFR), 1998.

[4] J. Barraquand and J.-C. Latombe, “Robot Motion Planning: A Distributed Representation Approach,” The International Journal of Robotics Research, Vol. 10, No. 6, pp. 628–649, Dec. 1991.

[7] J. van den Berg and M. Overmars, “Planning the Shortest Safe Path amidst Unpredictably Moving Obstacles,” Proc. Workshop on Algorithmic Foundations of Robotics, July 2006.

[10] J. Borenstein and Y. Koren, "The Vector Field Histogram – Fast Obstacle-Avoidance for Mobile Robots," IEEE Journal of Robotics and Automation, Vol. 7, No. 3, pp. 278–288, June 1991.

[11] O. Brock and O. Khatib, “High-Speed Navigation Using the Global Dynamic Window Approach,” IEEE Int. Conf. on Robotics and Automation, Detroit, MI, USA, pp. 341–346, 1999.

國際替代計量

行動機器人在動態環境之路徑規畫

全文下載

主題瀏覽