ANFIS應用於四足機器人跨越非連續河石地形路徑規劃之研究

Abstract

本文研究目的為使用適應性網路模糊推論系統(ANFIS)演算法，應用於四足機器人跨越非平坦河石地形(River-stone Terrain)之路徑規劃，並搭配靜態ZMP穩定性分析，使機器人隨著不同地形做不同姿態調整，提高人工智慧步伐規劃的層次。 程式設計部分，首先針對本實驗室四足機器人(NC_F4-09)架構中四組空間三自由度連桿手臂進行規劃研究，過程由簡入深，首先從平面三自由度連桿手臂開始做起，接著應用至空間三自由度連桿手臂，最後則是延伸至四足機器人做路徑規劃，處理平衡問題時使用ZMP理論，當行走時投影至地面的ZMP點，如果落在腳底支撐面上，行走過程中就能維持穩定。 本研究成果還包括以MATLAB為基礎所建立的步伐規劃自動產生的圖控式介面系統(GUI)，有助於簡化四足機器人跨越河石地形的步伐規劃與實際控制驗正。 實驗硬體方面主要採用壓克力板組成架構形狀，再加裝伺服馬達、7.2V充電電池和馬達控制器(SSC-32)，經過測試調整後，完成四足機器人的實驗模組(NC_F4-09)，四足機器人的步態規劃過程為透過ANFIS演算法，將MATLAB程式計算出的各軸運動數據，利用RS232傳輸線由電腦傳送至SSC-32中，讓四足機器人隨著不同使用者輸入的地形產生相對應的步伐控制數據，實際去跨越非連續或非平坦的河石地形。

The study proposes an innovative method which combines Adaptive Neuro-Fuzzy Inference Systems(ANFIS) with Zero-Moment-Point(ZMP) algorithms and applies in path planning for quadruped robot walking through a discontinuous River-stone terrain.Therefore,the robot can change its walking figure for adapting to different terrain like humans. Because the quadruped robot can be regarded as combinations of four 3-DOF manipulators and the ANFIS algorithm can also be regarded as combinations of Neural-Network and Fuzzy inference system,this thesis study is planned and proceeded from simple to difficult case.Firstly, we study the case of 3-DOF manipulator in a plane based on Fuzzy theorem .Second we compare the case of 3-DOF spatial manipulator based on both Neural-Network and ANFIS theorems.Finally this thesis proposes the ANFIS and ZMP methods to apply to the gait planning of 12-DOF quadruped robot. This research deals with the stability control algorithms of quadruped robot by ZMP method,when the robot walks through discontinuous terrain. If its projection of ZMP on the ground is inside the triangle foot print area,the robot will keep stable on walking. There are many experimental instruments were adoptive in the study,such as acrylic sheet,servo motors and SSC-32 servo controller. This gait planner transfers the angle datas made by ANFIS programs to the SSC-32 board by RS232 connected cable,and let the robot pass through the discontinuous River-stone successfully and stably.