以適應性模糊類神經網路為基礎之縱向車隊強健控制

Abstract

在本文中，我們提出以三種不同的縱向車隊控制器，包括模糊控制器，滑動模式控制器，以及以滑動模式觀測器為基礎的滑動模式控制器，並搭配非直接適應性模糊類神經網路近似器及 概念來滿足系統的強健性及表現。對於一列直線行進的車隊，已知隨著前車動態的變化，譬如由低速前進到高速前進，車隊中車輛間的相對距離也將隨之變動，此一車距的改變將被用來改變車隊中後方車子的行進行為。我們的控制策略是施加合理的油門控制力迫使車隊中每部車子之間相對於前方車輛保持適當的安全距離，並以舒適的加減速追隨前方的車子的行進行為。在此假設在車輛行進的過程中，車間相對位置是可量測的。 在縱向車隊行進控制上，模糊控制器可呈現良好的控制效果，而滑動模式控制器提供了更穩定且可信賴的控制效果。在引入模糊類神經網路近似器及 概念後維持了系統的強健性及改善了系統控制輸入的切跳現象。為了更符合真實車隊系統的效益，我們假設僅能量測到車間的相對距離。在此限制下，以滑動模式觀測器為基礎的滑動模式控制器亦保證對前車位置的追蹤是全域穩定的。由模擬的結果，可以證明三種控制器的正確性和穩定性。

In this Thesis, three different controllers for the longitudinal car following system, including a fuzzy logic controller, a sliding mode controller , and a sliding-observer-based sliding mode controller with indirect adaptive fuzzy neural network approximator and performance, are proposed. For a longitudinal control of a platoon of a vehicles in a straight line, as the speed of the preceding vehicle increases or decreases, the relative distance of vehicles changes. This will be used to act the throttle ( or brake) of the follow vehicle. The main control strategy is to force the follow vehicle tracking the lead one with a safety distance. We assume that the relative distance is measurable and measured by the follow vehicle. The fuzzy logic controller for a longitudinal car-following system provides a good performance for the follow vehicle to track the lead one. The sliding mode controller provides more stable and reliable performance. After associating with the fuzzy neural network approximator and performance, the controller performs as well as the sliding mode controller and meanwhile smooth the control actions. Under the assumption that only the relative distance is measurable, the sliding observer is combined with the former controller. It also guarantees the overall system is globally stable. Simulation results will show the validity and effectiveness of the proposed controllers.