— This paper focuses on the development of wandering robot formations and shows the cooperation and coordination among the robot teammates to maintain the achieved formation regardless the complexity of the environment. To address these issues, this paper proposes a new behavior based robot architecture. This architecture is based on a novel technique for location determination using local sensing. The proposed architecture is implemented using the well known robot simulator Webots. Experiments for many difficult tasks such as the passage through narrow corridors, obstacle avoidance, swerving with large angles, and switching between different formations have been conducted. These Experiments prove the efficiency of the proposed controller. The obtained results show that the constructed formations are more stable and accurate even in cluttered and uncluttered environments. Keywords— Local sensing, Multi"robot coordination, Robot formations. I. INTRODUCTION his paper addresses the problem of achieving a global behavior by a group of mobile robots. This problem has become of paramount interest nowadays as a result of its various important applications. Mobile sensing networks, cooperative transportation and mine field exploration are some examples of tasks which can be assigned to a robot group. Robot formations comprise an essential part for most of these applications. The paperaim is to concentrate on developing an intelligent controller for robots having very simple design so that they could exhibit a complex global behavior. The robots are aimed to organize themselves to take different geometric formations. Also, they should be capable of dealing with difficult situations which they may face in their environment. The robot team should be able to avoid obstacles, pass through narrow corridors (in fish or bird swarm fashion), swerve in large angles and switch between different formations. Exhibiting all these capabilities using very simple hardware is really a well respected challenge. Manuscript submitted December, 14, 2008 All authors are with the Electronics and Communication Engineering Department, University of Helwan, Cairo, Egypt. (corresponding author e" mail: hosam_ib_ali@ yahoo.com). The rest of the paper is organized as follows. Section 2 gives a review of related work. Section 3 illustrates the hardware platform for the robots. Section 4 presents the proposed intelligent controller. Section 5 describes the experimental work. Section 6 gives a discussion and the conclusions. II. RELATED WORK Various approaches have been proposed for the problem of emerging a global behavior for a group of robots [1"3]. In [4], a group of simulated robots are used to perform some robot formations such as circles and simple polygons. In this work, robots have a global knowledge about the other robots’ positions. Each robot orients itself to the furthest and nearest robot. In [5], a group of mobile robot motion was also considered. They used the matrix formation performing a right turn as an example. In [6], a formation is defined by a so" called virtual structure (VS). The algorithm assumed that all robots have a global knowledge; it iteratively fit the VS to the current robot positions, displaced the VS in some desired direction, and updated the robots’ positions. [7"8] face parts of the controller needed for robot formations, their work depend on vision and complicated manipulations. The work in [9] tried to find a general algorithm for robot formations using local sensing. Their approach was based on gaining other robots information via a camera mounted on each robot. The algorithm depended on constructing an ordered robots chain using their ID’s. Then using simple rules, each robot could compute the distance and angle to be maintained with a friend robot. Using a camera as a local sensor complicated the controller which may be simplified if any other simpler hardware could be used. The work in [10] used the so"called local template by which robots gradually construct the desired formation. It verified the proposed approach for chain formation only. The main point achieved was the usage of simple hardware for gaining information. Although the above mentioned approaches gave a better performance, however, some of them implied complex sensing mechanisms. Others only provided a few number of robot formations. The paper work proposes a simple vision approach based on infra"red transmitters and receivers by which the distance and the angle of neighboring robots can be measured. It enables each robot to see all other nearby robots (although each robot transmits data using unique frequency). The controllers are designed such that each robot starts to follow the first observed teammate. This mechanism enables the robots to construct the robot chain rapidly from any randomly oriented               E.M. Saad, M.H. Awadalla, A. M. Hamdy and H. I. Ali T INTERNATIONAL JOURNAL OF COMPUTERS Issue 3, Volume 2, 2008 330