Distributed Formation Control using Consensus for Multiple Humanoid Robots

Abstract

Formation control in multi-robot is one of the problems widely studied and variety of methods have been developed with their own advantages and disadvantages. Distributed control has the advantage in that it does not require centralized coordination, allowing rapid restructuring in the event of failure of one of the agents. Another advantages include saving computation power, amount of data transmitted over communication, and immunity to delay. In this research a distributed formation controller is designed and implemented in humanoid robots. The distributed controller is designed consisting of layer for tracking, layer for constructing the formation, and consensus layer that keeps the error between each agent can be agreed upon so that a reference can be achieved together. To solve the problem of bad odometry suffered by legged mobile robot caused by the slip between the legs with a floor, visual odometry is employed using the ceiling camera. The control system is simulated using GNU-Octave software and implemented on humanoid robots NAO from Aldebaran. Experiment demonstrates that the distributed controllers are able to closely follow the desired reference.