Designing a fuzzy logic controller for a quadruped robot using human expertise extraction

Abstract

Many actions have been taken by researchers to design an appropriate controller for a quadruped robot. These efforts can be divided into two main categories: analytical and intelligent methods. The analytical method is based on the motion equations of the robot which are mostly complicated. These complication leads to difficulty in implementation. In contrast, there exists intelligent method which is based on the learning algorithms. Despite lots of improvements in the controlling of the quadruped robots, their ability of walking is still far less from the four legged animals' capability. Many researchers believe that four legged animals owe this priority to the “learning” attribute. Therefore In this paper, we used human learning ability to propose a linguistic controller. First a human without having any information about the dynamics of a quadruped robot, simulated by Webots software, tried to control it by means of a keyboard. After he gained enough experience to control the robot properly, we extracted his knowledge as 28 fuzzy rules. Fuzzy controllers can control the quadruped robot with lower cost and better quality. The extracted rules fed to the robot and the results determined that the applied fuzzy controller can appropriately control the walking of the robot.