Distributed formation control of non-holonomic mobile robots

Implementation and testing in Robot Operating System

More Info
expand_more

Abstract

The use of distributed multi-agent formation control for non-holonomic mobile robots shows great promise in practical environments. In this thesis, a Robot Operating System (ROS) package is developed in which various formation control algorithms are implemented for use on a group of skid-steer drive mobile robots. These formation control algorithms are tested in a simulated laboratory environment. For this, first seven distinct ways to adapt holonomic
(omnidirectional) formation control algorithms to be used on skid-steer drive, differential drive or unicycle model mobile robots are presented. These seven adaptations can also be used to categorise existing non-holonomic formation control algorithms for unicycle model mobile robots. These adaptations are used to implement three formation control algorithms in a ROS package. Furthermore, obstacle avoidance and goal reaching is added to the package. Control inputs are generated from objects detected in the local coordinate frame of the robots, by means of the 2D LIght Detection And Ranging (LIDAR) and an open source obstacle detection package. The formation control package and the implemented formation control algorithms are tested on a group of skid-steer drive mobile robots, the Husarion ROSbot, in the Gazebo simulation environment. Simulations show the effectiveness of the developed formation control algorithms. Various conditions regarding control frequency and LIDAR noise levels are evaluated, and finally, a number of practical recommendations for the use of formation control algorithms on non-holonomic mobile robots are made.