INVERSE KINEMATIC CONTROL IN A ROBOTIC ARM USING ARTIFICIAL NEURO-FUZZY INTERFACE SYSTEM

Abstract

Degree of Freedom) of robot is increased it becomes a difficult task to find the solution through inverse kinematics. The work presented here is about the solution of inverse kinematics based on adaptive network learning. The mapping of Cartesian space to joint space is done using ANFIS (Adaptive Neuro-Fuzzy Inference System), which uses hybrid learning algorithm to update parameter values in membership function. The trained ANFIS network is then used as a part of a larger control system to control the robotic arm. For testing purposes an experimental setup is prepared for SCORBOT-ER Vplus robotic arm with five DOF and vision system with single overhead stationary camera. Using the camera images are captured for different types of objects and using Brown's distortion model pixel coordinates to Cartesian coordinates are generated for Centroid of each object. The Cartesian coordinates are given as inputs to the ANFIS model while the output of the ANFIS model is in the form of joint angles. It is then converted to encoder counts and then injected to the robot controller. The controller uses ACL (Advance Control Language) to control the servo motors connected to the each joint of robotic arm, the communication between computer and controller is done using serial port RS232 with MATLAB platform. According to the value of joint angles the robotic arm moves to the center of each object and picks it up and places it in a suitable