隨機散佈障礙環境下動態路徑規劃–結合GVD、D* Lite、與SVM之研究

Dynamic Path Planning under Randomly Distributed Obstacle Configuration – a Study Based on GVD, D* Lite and SVM

蕭孟華(Meng-Hua Shiao)

指導教授 ： 楊智旭

共同指導教授 ： 楊棧雲

淡江大學/工學院/機械與機電工程學系碩士班/碩士(2011年)

https://doi.org/10.6846/TKU.2011.00023

摘要

本論文針對隨機散佈障礙環境下，安全平滑之動態路徑更新，結合非格點式之拓樸地圖，建立快速之路徑再規劃方法，以符合動態路徑規劃之需求，基於非格點式拓樸地圖，重新檢驗、修改廣義Voronoi結構劃分(GVT, Generalized Voronoi Tessellation)與D* Lite演算法(D* Lite Algorithm)之理論，確立一個時間精省具有再規劃能力之前處理器，再以支持向量機(SVM, Support Vector Machine)構築一個具最大邊限的安全平滑路徑。
本研究援引之廣義Voronoi結構劃分與D* Lite演算法皆為計算時間考量下一時之選，銜接非格點式之拓樸地圖，環環相扣，奠定一計算精省之再規劃方法，具有實際應用之潛力，再佐以支持向量機之平滑後處理，確可達於隨機散佈障礙環境下安全平滑之動態路徑更新的目的。

關鍵字

路徑規劃 ； 廣義Voronoi結構劃分 ； D* Lite演算法 ； 支持向量機 ； 平滑 ； 安全

並列摘要

In real applications, a path planning under dynamic environment change is a practical and important topic for the grounded mobile robot. The study elementarily employs the methods of generalized Voronoi tessellation (GVT) and D* Lite shortest path algorithm for the topic. With the excellence of less computation, the topology of scatted waypoint configuration for mapping the terrain is first chosen for the real-time replanning request. The elementary methods are hence theoretically modified to fit adequately the waypoint topology. The modification seeks a swift solution to respond a replanning request corresponding to a waypoint topology change. The sub-sequential processes: waypoint topology change, GVT, and D* Lite form a pre-processing planner. It can be applied standalone to provide an optimized piecewise linear path for robot reaction. With an additional support vector machine (SVM) post-processing smoother, it can also provide a wide-margin safe and smooth path. The smooth path is optimized, too. The solid development based on the waypoint topology, instead of the quantized grid-map of previous researchers, is the most important contribution of the study. With small scaled experiments, evidential results show consistently the model is promising for a great improvement in applications and achieve the original goal of the study.

並列關鍵字

Path Planning ； Generalized Voronoi Tessellation ； D* Lite ； SVM ； Smooth ； Safe

參考文獻

[5] 馬志豪，基於分類之避障路徑規劃與實現，碩士論文，淡江大學機械與機電學系，民國九十九年

[4] 周峰毅，安全平滑之機器人路徑規劃—基於大邊限支向機的研究，碩士論文，淡江大學機械與機電學系，民國九十六年

[1] A. Zelinsky and I. Dowson, “Continuous Smooth Path Execution for an Autonomous Guided Vehicle,” TENCON '92. Technology Enabling Tomorrow : Computers, Communications and Automation Towards the 21st Century, 1992 IEEE Region 10 International Conference, 871-875 vol. 2, Melbourne. 11-13 Nov., Australia, 1992.

[2] T. Fraichard and M. Ahuactzin, “Smooth Path Planning for Cars,” ICRA 2001: 3722-3727, Proceedings of the 2001 IEEE International Conference on Robotics and Automation, ICRA 2001, May 21-26, Seoul, Korea, 2001.

[7] V. N. Vapnik, “Statistical Learning Theory,” John Wiley & Sons, New York, 1998.