Design an Intelligent Balanced Control of Quadruped Legs Based on Adaptive Neuro-Fuzzy Inference System (ANFIS)

摘要

This study discusses the 12 DoF stability control algorithm for Quadruped robot legs to adjust the balance on irregular terrain. The source of the instability is the irregularity of the ground surface and external forces. Therefore, dynamic stability criteria are needed to plan the robot's movement and restore balance for the movement of a four-legged robot with a dynamic gait over an irregular terrain. The novelty of this study is the use of 12 ANFIS at once to manage the 12 DoF of each leg, which are grouped into four sections, and each section consists of 3 ANFIS. The ANFIS method is used as an algorithm to move the 12-DoF robot legs by training some robot leg movement data based on the slope angle of the surface. The results of training with the ANFIS method can be optimal if the number of rules is close to the given training data. From 29 body tilt angle position data and 12-DOF robot legs, good results will be obtained if the 5x5 number membership function is used for each input which will produce 25 ANFIS rules and combined using the Gaussian type so that it can produce RMSE = 0.068233. The next research is to develop reliable methods such as Zero Moment Point (ZMP) combined with the BPNN or ANFIS methods so that it is expected to get a reliable robot body balance.