倒車入庫之最優路徑規劃及控制演算法

Abstract

目前自動停車結果的好壞，仰賴載具起始姿態與複雜的即時路徑規劃，並結合高精度感測器進行碰撞檢測。本研究以微分平坦理論進行基於阿克曼轉向機構之倒車入庫最優路徑規劃及控制。微分平坦算法具有低運算複雜度，可以進行即時路徑規劃同時設計出系統之控制。路徑規劃中，根據當前姿態決定姿態調整策略及專家停車策略，使得汽車可以根據起始姿態與車格之位置決定合適的停車策略。另外，停車前，採用路徑預規劃進行碰撞檢測，使得倒車入庫更加安全可靠。本研究以阿克曼轉向理論為基礎，設計出符合阿克曼轉向約束條件之四輪載具控制策略，並透過不斷迭代求解全局最優之控制策略。

Currently, the effectiveness of automatic parking depends on the initial posture of the vehicle, complex real-time path planning, and collision detection using high-precision sensors. This research employs differential flatness theory for optimal path planning and control of reverse parking based on the Ackermann steering mechanism. The differential flatness algorithm has low computational complexity, allowing for real-time path planning and simultaneous control system design. During path planning, the posture adjustment strategy and expert parking strategy are determined based on the current posture, enabling the car to decide the appropriate parking strategy based on the initial posture and the position of the parking space. Additionally, before parking, path pre-planning is used for collision detection, making reverse parking safer and more reliable. This study is based on Ackermann steering theory, designing a four-wheel vehicle control strategy that meets the Ackermann steering constraints, and iteratively solving for the globally optimal control strategy.