基於微分平法之 Hybrid A* 路徑規劃與無人載具控制系統實現

Abstract

本研究旨在設計阿克曼(Ackermann)無人載具於戶外環境中的即時避障路徑 規劃與控制系統。首先提出一套改良式 Hybrid A* 演算法，透過遞迴搜尋與優化 節點擴展方式，提升搜尋效率與可行路徑品質，使其能有效產生符合載具轉向條 件的最短無碰撞路徑。接著，本研究進一步設計一套結合改良式 Hybrid A* 概念 與微分平(Differential Flatness, D.F.)理論之路徑規劃演算法，透過 D.F. 將阿克曼 載具模型從高維空間映射至低維空間輸出，並引入最佳化約束，得到滿足阿克曼 載具之非完整約束(Non-holonomic Constraint)之路徑，並進一步結合光達等感測器即時偵測障礙物，透過模糊控制器追蹤路徑，實現具備即時避障能力的無人阿克曼載具自動駕駛系統。

This study aims to develop a real-time obstacle avoidance path planning and control system for Ackermann unmanned vehicles in outdoor environments. First, an improved Hybrid A* algorithm is proposed to enhance search efficiency and path quality through recursive search and optimized node expansion, producing the shortest collision-free paths that satisfy vehicle steering constraints. Next, a path planning method combining improved Hybrid A* and Differential Flatness (D.F.) is designed. D.F. maps the Ackermann vehicle model from a high-dimensional space to a low-dimensional trajectory output, and optimal constraints are introduced to generate paths that meet non-holonomic constraints. Real-time obstacle detection using LiDAR and other sensors is also integrated, and a fuzzy controller is employed for path tracking, achieving a real-time obstacle avoidance autonomous Ackermann vehicle system.