符合阿克曼轉向模型之汽車避障策略

Abstract

本研究旨在透過阿克曼轉向幾何（Ackermann steering geometry）模型找到能夠成功閃避載具航行路徑上障礙物的臨界轉彎點位置，並將其概念引入現有的避障方法中。傳統的A*演算法雖能生成地圖中的最短路徑，但該路徑僅存在位置資訊，並未考慮載具所受之非完整約束，因此載具追蹤其生成軌跡行駛轉彎時仍存在碰撞風險。本研究的目標為調整A*演算法與速度障區法，透過載具與障礙物的位置、姿態等訊息，找到載具在符合阿克曼轉向模型之非完整約束條件下，恰能成功閃避障礙物的轉彎點位置，並將其運用於動態調整障礙物半徑，使得載具航行生成路徑時，不因轉向限制存在碰撞風險。

This study aims to identify the critical turning positions that enable a vehicle to successfully avoid obstacles in its path by using the Ackermann steering geometry model and to incorporate this concept into existing obstacle avoidance methods. While the traditional A* algorithm can generate the shortest path on a map, this path only includes positional information and does not consider the nonholonomic constraints imposed on the vehicle. Consequently, there remains a risk of collision when the vehicle follows this path through turns. The goal of this study is to adjust the A* algorithm and the Velocity Obstacle (VO) method by using the position and orientation information of both the vehicle and obstacles. This will help in identifying the turning positions that allow the vehicle to avoid obstacles while adhering to the nonholonomic constraints of the Ackermann steering model. Additionally, the approach dynamically adjusts the obstacle radius to ensure that the vehicle-generated path does not pose a collision risk due to steering limitations.