Agile Experimentation of Robot Swarms in Large Scale

Abstract

This chapter aims to present a new ROS package to automate experimentation with multiple mobile robots. A robot swarm is a specific system that requires a complicated setup and has a high cost with regard to experimentation. Virtual environments can be used to expedite testing; however, these also are very laborious. This package is a tool set to easily configure the experimentation environment for swarm tasks with the most popular perception systems and absolute or relative localization references. The user specifies in the tool only the number of robots required, sensors, and functions without having to configure each robot individually for the simulation. This chapter presents two examples of the developed package to show how the new package simplifies working with swarms of robots and focuses on the application rather than the required configuration. An example of a formation maintained through the fuzzy approach is developed to demonstrate the potential of the proposed package. The approach is based on a leader agent that executes autonomous navigation through a LIDAR perception system, and follower agents that are responsible for maintaining the formation based on the leader. A fuzzy intelligent behavior commands the dynamic formation adaptation of the robot swarm as it attempts to overcome obstacles. Finally, the computational cost is evaluated to allow readers to estimate the computational resources necessary to perform practical experimentation.