INVESTIGATION OF THE IMPLEMENTATION OF LEADER-FOLLOWER FORMATION CONTROL TO BIPEDAL LOCOMOTION

Abstract

Robot that uses bipedal locomotion such as humanoid robot has a unique appeal in which the robot can perform many duties/works that cannot be done by wheeled robots due to spatial and environmental constraints. Related literature were about biped robots and robot arm that uses the concept of central pattern generator (CPG). Biped robots that use CPG (technically a neural oscillators) do not need any mathematical model of the robot itself. The controller “exploits” the dynamics of the robot to achieve an efficient way to drive the robot’s joints. The driving frequency from this type of controller is thus heavily influenced by the dynamics of the robot it controls. One disadvantage of this method is that we cannot adjust the walking/movement parameter easily. Inspired by the idea of central pattern generator where one central “brain” controls the movement of the joints of the robot and the fact that the system uses neural oscillators reach some kind of “consensus”, we are interested to study the feasibility of the implementation of leader – follower formation control based on consensus algorithm to bipedal locomotion system. The method studied in this report is to force the dynamic model of the double pendulum to match the kinematic model of the double integrator agent that uses consensus algorithm in forming formation. The walking/movement parameters of the robot using the proposed method are defined by one central “brain”, and then the joints are working in consensus way to achieve the target joints’ trajectories specified by the “brain”. The study also concludes the notion of stability of the system driven by this controller which strongly related to the consensus ability in the context of Multi – Agent System.