Project SMURFS – A Society of Multiple Robots

A reconfigurable robot society built from scratch with LEGO Mindstorms was presented at the IJCAI 2009. This paper describes the robots as well as the algorithms developed over the last few months. A leader selection as well as a formation for area coverage algorithm were tested and presented at the conference. For localization a birds-eyeview camera system based on reacTIvision fiducial markers is used. Introduction and Background This paper describes the project SMURFS, which was our contribution to Robotics Workshop and the Eighteenth Annual Robot Exhibition at the International Joint Conference (IJCAI) 2009 in Pasadena, CA. The project uses a reconfigurable robot society built from scratch with LEGO Mindstorms over the last 7 months. Separately a cooperation algorithm, in its simplest form, controlling the robots formations, was implemented and tested with a simulator, also developed over the last few months. A machine learning approach and a vectorbased approach were evaluated in the simulator and the vector-based approach was then also tested with the reconfigurable robots. Reconfigurable Robots Reconfigurable multi-robot societies is a young area of robotics that promises versatility, robustness and low cost as it relies on a society of multiple robots that will be able to perform an immeasurable amount of different tasks, tasks that not even were thought of at design time. In order to create a new reconfigurable multi robot society from scratch that could be developed fast and with low costs, it was opted to create a prototype from LEGO Mindstorms NXT equipment, complemented with some extra electronics developed to expand the capabilities of the existing LEGO Mindstorms NXT system. Reconfigurable robot systems will have a great impact, especially in the field of space exploration, where the Copyright © 2009, Association for the Advancement of Artificial Intelligence (www.aaai.org). All rights reserved. lower mass needs and the flexibility to use the system in many different ways (even not yet foreseen tasks) will be appreciated. Cooperative Behavior Cooperation is not just discussed in robotics but also in political and other human sciences. Robot cooperation takes ideas from biology and tries to extend the behaviorbased control to get simple cooperation (Arai et al., 2002). The use of cooperative robotics reaches technological constraints, even more than regular robotics, because of the need to cope with multiple, autonomous entities. At the same time it is highly inter-disciplinary and draws influences from many other fields of research, e.g. artificial intelligence, as well as biology. Examples of cooperating in nature (e.g. bees and ants) show possibilities for simple robots to work together to solve a very complex task. In contrast to the low level control of the robot (e.g. motion), cooperation can be seen as high level control, involving task and motion planning, task sharing, and formations, which can be seen as the simplest form of cooperation between autonomous robots.