先進車輛控制及安全系統之設計與模擬---子計畫四---高可靠度跟車系統之設計(I)

Abstract

自從車輛被發明之後就一直扮演著人類日常生活中一項不可或缺的交通工具，然而近幾年來隨著車輛數量的大幅成長，駕駛員所可能遇到的行車疲勞之負擔與行車安全之威脅也相對地越來越高，車輛的行車安全考量問題就顯得日益重要，所以本計畫積極地尋求如何將先進的科技運用在車輛上以提昇車輛行駛時的安全性，以求能夠減少事故以及傷亡的發生。 本子計畫著重於車輛自動跟車之控制系統設計工作，期望可以藉由無線通訊技術與影像處理技術，不斷地提供受控車輛即時之車前資訊，藉由本子計畫所開發設計之控制法則讓所有受控車輛都可以保持固定車速與車距，如此將能夠提高道路運輸效率並降低交通事故發生。另外通常一個控制系統是由多個傳感器、天線、微處理器和數字式信號處理器組成，感測器能向司機或車用電腦提供車輛前、後、兩側及鄰邊車道可能發生意外事件的即時信息，為了確保整個自動跟車系統之控制性能與安全考量，所以本計畫所設計開發之自動跟車控制系統將必須能夠解決時間延遲問題且具有故障檢錯及自我回復等能力。本計畫提出一個為期三年的計畫，其主要研究內容與進行步驟描述如下： 第一年： 第一年度之研究計畫正在執行中，目前已建立整個車隊之系統模擬環境，並利用模糊控制理論成功設計一智慧型自動跟車控制系統，整個模糊控制器不需知道前導車之速度與加速度等資訊，只需知道該車與前車之車速與加速度資訊即可達到十分良好之控制響應，接著將繼續朝向模糊類神經網路控制架構，設計開發具有自我自我學習能力之新型自動跟車控制系統，藉此達到本年度之計畫目標。 第二年： 通常之控制系統是由多個傳感器來向司機提供車輛附近可能發生意外事件的即時信息，所以第二年除將繼續開發改善智慧型自動跟車控制法則外，也將增加考慮各種跟車模式，如合併、分離、切換車道等等問題。並利用回饋型模糊類神經網路來解決降低因致動器與感測器所造成時間延遲對系統的影響。 第三年： 先進車安全控制系統除了要求控制性能外，系統的安全性和可靠度的要求也愈來愈高，所以故障檢錯及自我回復系統扮演一個極重要的角色，當安裝於車上之各種不同功能之感測器失效或車輛部分零件故障時，如何利用小波類神經網路設計一具有故障檢錯及自我回復能力的自動跟車控制系統確保整個行車之安全性，即為第三年度主要研究工作。 本計畫所開發之全部控制法則因為不需事先知道受控系統之數學動態模式，其控制參數依據李亞普諾夫穩定定理推導出來的適應性法則來調整，故其系統控制性能得以確保，其系統穩定性亦可證明。所以如飛行系統、機電整合系統、高溫熔爐系統、電力系統、核能反應系統及化學程序控制系統等等均將可適用。

Transportation technology is one of the most influential areas on the human life. Therefore, researchers have been involving in wide scope of related research activities aiming to enhance efficiency, comfort, and safety of transportation systems. Due to the ever growing number of vehicles on the roads, urban highways are congested and need additional capacity. Though building new roadways sounds like the prominent solution at the first glance, in practice it causes number of other unsolvable problems mainly due to scarce of suitable lands. To this end, to increase the traffic capacity by automating the traffic flow has been identified as a smarter option. Moreover, due to ever increasing traffic on the roads, the tendency that the drivers face tough, complicated situations is increasing. Safety issues due to driver weariness when driving for long hours is another concern. Therefore, built-in driver support systems to assist the drivers in hazardous situations are of utmost importance. An automated vehicle following control system that enables the vehicle to control the engine torque by its own keeping a prescribed safe distance from the preceding vehicle is proposed. With the inclusion of complementary driver support systems, the overall control system can mitigate driver』s work load and guarantee much improved safety. Such a car-following collision prevention system usually includes radar and vision sensors, micro-processors, antenna, and digital signal processors (DSPs). This project will last three years. Now, in the first year, we have successfully developed a platoon simulator considering of various road conditions, and successfully developed a fuzzy controller to achieve a favorable tracking performance. The fuzzy controller does not use any communication links between the vehicles in the platoon. In the future, we will utilize a fuzzy neural networks based approach to solve intelligent automated car-following control problem. An on-line tuning algorithm is derived in the Lyapunov sense and the stability will be guaranteed even in the face of various road conditions. Since the automated car-following control problem accounts for lots of efforts that involve sensor system, antenna hardware programming and design associated with microprocessors and DSPs, in the second year, we will continue our focus mainly on the intelligent car-following control problem. Moreover, the problems of lane changing, merging, and leaving the platoon will be addressed. A novel intelligent automated car-following scheme using recurrent fuzzy neural networks will be developed to cope with the inherent time delay problem caused by communication and actuator delays. System safety and reliability are the other most demanding issues to be addressed. In the third year, the purpose of this project will be to develop systematic design methodologies in order to achieve the designed systems to simultaneously deliver desired performance, fault detection ability, and recognition capability. Extensive simulations of automatic steering and throttle/brake performances will be carried out to demonstrate the effectiveness of the developed tools and controllers. Since the control algorithms would be model-free techniques and are tuned on-line using Lyapunov-based adaptation laws, not only the control accuracy but also the system stability can be guaranteed. The wide range of applications of the developed technology will also include flight control, power systems, nuclear power industry, process control, and transportation systems.