本研究所探討的對象為由一般八輪甲車所衍生發展而具備兩棲功能的浮游甲車，由於滿足浮游性能之兩棲甲車，其重量、重心位置及動力系統等可能與原型甲車相異，故本研究對於此一兩棲甲車的陸地翻覆安全性能做深入的分析及探討。文中探討車輛翻覆之安全控制，以主動轉向（Active Steering）及差動剎車（Differential Braking）作為控制輸入，採用閉迴路跟隨模型架構（ModelFollowing Control Structure），以理想之虛擬車輛模型之穩態響應，作為實車追隨之目標。本研究以LQR 二次線性最佳化控制法則為基礎，再使用粒子群演算法（Particle Swarm Optimization），分別以搜尋LQR 之權重矩陣Q 間接求得增益矩陣K，及直接搜尋系統參數所得增益矩陣K 作比較，來求解控制系統目標函數之最小值。最後利用Matlab®/Simulink® 軟體結合TruckSim® 卡車模擬軟體，以驗證控制成效。模擬結果顯示以直接搜尋法求得之系統最佳增益值較佳，能有效提高車輛高速行駛時之操控穩定性。最後，藉由灰理論來偵測翻覆控制的啟動時機，能讓翻覆控制系統提早介入示警，未來在實車應用上能更有效率且安全地達到翻覆控制目的，有效降低行車危安風險。

The object of this research is an amphibious eight-wheeled armored vehicle which is an alternative model developed from a typical eight-wheeled armored vehicle. Since amphibious armored vehicles that have floating capabilities may have different weights, center of gravity and power systems in comparison to the original armored vehicle. Thus, this study explores the rollover safety of the amphibious vehicles on the ground. In this paper, the control of vehicle overturning safety was assessed with active steering and differential braking as control inputs. A model following control structure was adopted in which the steady-state response of an ideal virtual vehicle model was used as the target to be followed by the real vehicle. Based on the LQR optimization control law and a particle swarm optimization (PSO) method, a better performance of the control system was chosen by comparing the following two optimization approaches: Minimizing the objective function of the control system to determine the feedback gain matrix K by indirectly searching the weighting matrix Q of the LQR control, and directly searching the system parameters by the PSO method, respectively. Matlab®/Simulink® software was used in combination with simulation software TruckSim® to verify the control effectiveness. Simulation results showed that the feedback gain obtained by the direct search method, which can effectively improve the stability of the vehicle during high-speed turning, is superior to that of the indirect search one. Finally, grey theory was used to detect the triggering time of the rollover control system in order to activate the control system in advance. As a result, the control system may be applied to real vehicles to efficiently and safely achieve anti-rollover control and reduce the risk of traffic accidents.