本論文是利用模糊滑動控制器(Fuzzy Sliding-Mode Controller, FSMC)和自適應調變器(Adaptive Tuner) ，設計一個自適應模糊滑動控制器（Adaptive Fuzzy Sliding-Mode Controller, AFSMC），作為輪型機器人之行進軌跡追蹤使用，使其能夠追隨規劃之路徑。在FSMC的設計上，以機器人之x, y座標軸之位置誤差所構成的PD滑動函數作為模糊控制器的輸入變數，以左右輪馬達的轉速作為控制器的輸出變數，在模糊集合的設計上，是以正規值加上修正值作為後件部的模糊歸屬函數；在自適應調變器的設計上，則是以Lyapunov 穩定定理推導修正量的調變演算法則。
為驗證輪型機器人行進蹤跡控制之結果，分成模擬與實驗兩部分呈現，分述如下：(1)模擬部分：首先設定幾種不同路徑，以所提之AFSMC來計算輪型機器人的左輪、右輪的速度與轉彎角度，進行路徑追蹤。(2)實驗部分：利用Matlab撰寫AFSMC的演算法之後，計算並送出左、右輪之速度，接著利用Labview啟動輪型機器人（FESTO Robotino），讀取並送出速度值，進行路徑追蹤。
與其他控制方法比較，本論文所提出之AFSMC具有下列特點，(1)模糊規則庫的結構精簡，使用較少的模糊規則；(2)後件部採用單值模糊歸屬函數，可減輕系統的抖動現象：(3)參數調變方面，只有一個參數需要調整，可降低CPU的運算量。從模擬與實驗結果可看出，本論文所提出之輪型機器人之行進蹤跡控制架構具有不錯的追蹤表現。
In this thesis, an adaptive fuzzy sliding-mode controller (AFSMC) is developed for wheeled robot to track planned path. The proposed control scheme contains a fuzzy sliding mode controller (FSMC) and an adaptive tuner. In the FSMC, the robot's x, y position errors are utilized to form the PD sliding surfaces and the sliding surfaces are treated as the input variable of the fuzzy inference system. While the left and right wheel motor speeds are designated as the output variables. The singleton fuzzy membership functions are adopted as the consequent part of the fuzzy system. In the adaptive tuner, the adaptive law is derived from Lyapunov stability theorem to tune the parameter.
To verify the tracking performance of the wheeled robot, some simulations and experiments are provided. (1) Simulation part: 3 different paths and 2 road conditions are set firstly. Then, the proposed AFSMC is applied to calculate the right and left speed of the wheeled robot. (2) Experiment part: The AFSMC algorithm is written by Matlab and the results are sent to Labview. Then the wheeled robot (FESTO Robotino) is activated by Labview via WiFi, and the commands are sent to motors to track the planned path.
The proposed AFSMC has some salient features: (1) Compact fuzzy rule base: 3 fuzzy rules are used; (2) Chattering phenomenon is reduced: the singleton fuzzy membership functions are used in the consequent part of a fuzzy rule; (3) CPU computation load is reduced: only one parameter needs to be adjusted. From the simulation and experimental results, it can be seen that the proposed tracking controller possesses robust and stable performance for wheeled robot path tracking.
