前置雷達電動車之自優化適應性巡航控制器

Abstract

適應性巡航控制(Adaptive Cruise Control, ACC)系統是實現自動駕駛的一項關鍵技術，車輛搭載ACC系統可以代替駕駛操作反覆的車速控制，因此可以減輕駕駛的注意力和視力負荷，然而現有的ACC系統存在下列幾個問題：1. 無法適用於各種駕駛情境；2. 車間距離過大無法達到提升道路吞吐量的目的，且容易面臨車輛插隊(Cut-in)的情境；3. 頻繁的加減速導致消耗額外的能量。 本論文以使用輪轂直流無刷馬達為動力的電動車為研究對象，設想毫米波雷達、動力總成、車輛等規格和參數齊備，用適應最佳控制演算法結合模糊PID控制器構成ACC系統的自優化適應性巡航控制器，以和前車的間隔時間(Headway time)和碰撞時間的倒數(Inverse Time To Collision, ITTC)做為模糊邏輯的前件部輸入，適應最佳控制演算法以優化跟車性能和能源效率為目標，透過不同的試驗行駛自動優化各模糊規則之下的PID控制器參數。 本論文用電腦模擬驗證ACC系統的自優化適應性巡航控制器，透過US06行車型態及多種情境的訓練自動優化各模糊規則的PID參數，訓練完成的控制器在UDDS、HWFET、US06、LA92行車型態，以及車輛插隊、停再開(Stop & Go)、追上前車時進行減速並跟隨前車的模擬情境下都能達成目標，且結果顯示使用自優化適應性巡航控制器和模糊邏輯控制器相比，皆能以較小的累計誤差及較高的能源效率完成任務。

Adaptive cruise control (ACC) system is a key component of automotive autopilot. It assists the driver with automatic speed control that lessens driver’s burden in attention and vision. However, several shortcomings appear in the existing ACC system. First, the system is not applicable to many driving scenarios that the driver may encounter. Second, a large headway distance is necessary that diminishes road throughput and may incur cut-in situations. Third, frequent acceleration and deceleration makes the vehicle consume more energy. This thesis presents the self-optimizing adaptive cruise controller that can maintain vehicle speed under various driving resistance, follow leading car even in stop-and-go, and prevent from any collision while a car cut-in. This innovative design is developed on an in-wheel motor-powered electric vehicle that has a front millimeter-wave radar to detect leading car’s relative speed and distance. The self-optimizing adaptive cruise controller consists of a fuzzy PID controller and the adaptive optimal control (AOC) algorithm. Particularly, premise inputs are headway time and inverse time-to-collision (ITTC). Fuzzification actually divides the operating points of the vehicle system into several linear regions, each associated with a PID control law. The AOC algorithm is dedicated to adjust the PID parameters for achieving better cruising performance and energy efficiency. On a simulation system, the proposed design is examined in driving cycles such as UDDS, HWFET, US06, and LA92, including scenarios such as cut-in, stop & go, and car following. The self-optimizing adaptive cruise controller succeeds in every driving cycle and scenario, and outperforms a fuzzy logic controller in term of accumulated error and energy efficiency.