應用於室內巡邏之自主式輪型機器人適應性動態控制器設計與實現

Abstract

本文提出一種混合式智慧型演算法以實現輪型機器人於已知室內環境的巡邏任務。首先，我們藉由D* Lite演算法決定輪型機器人從起點到終點的最佳路徑，再由所提出的混合式智慧型控制器使得輪型機器人可以有效地達成路徑追蹤的目的。此智慧型控制器可分成兩個部分，分別為運動學控制器以及TSK模糊控制器。在運動學控制器方面，利用倒階控制法設計的控制器可以保證輪型機器人的位置誤差與角度誤差將有效地趨近於零。此外，我們利用運動學控制器所得到的速度及角速度當作TSK模糊控制器的參考訊號。藉由TSK模糊控制器，輪型機器人的速度誤差與角速度誤差將會趨近於零。透過李普洛夫理論證明了此兩個控制器是漸近穩定。本研究先透過電腦模擬證實了該方案的有效性與可行性。最後，透過實驗驗證了該方法可以取得良好的效果。This thesis presents a hybrid intelligent algorithm to implement patrol tasks of a wheeled mobile robot (WMR) in a known indoor environment. First, we use D* Lite algorithm to determine the optimal path between the initial position and the destination. By using the proposed hybrid intelligent controller, we can efficiently accomplish the purpose of trajectory-tracking for a WMR. The intelligent controller is divided in two parts, which are kinematic controller and TSK fuzzy controller. The kinematic controller is designed by using a backstepping method. By using the kinematic controller, the position error and the angle error of the WMR efficiently converge to zero. Moreover, we use the linear and angular velocities obtained by the kinematic controller as the reference signals of the TSK fuzzy controller. Then the velocity error and the angular velocity error of the WMR converge to zero by using the TSK fuzzy controller. The asymptotic stability of kinematic and dynamic controllers is proven by Lyapunov theory. The effectiveness and the feasibility of the proposed scheme are verified by simulation results. Finally, experiments show that the proposed method can achieve good results.