VECIMS 2005 - IEEE International Conference on Virtual Environments, Human-Computer Interfaces, and Measurement Systems Giardini Naxos, Italy, 18-20 July 2005 Experimenting with a Robotic System for Localizing Magnetic Field Sources F. Amigoni * , S. Cadonici, V. Caglioti, and G. Fontana Dipartimento di Elettronica e Informazione, Politecnico di Milano, Piazza Leonardo da Vinci, 32 – 20133 Milano, Italy * Contact author, Phone: +39 02 23993475, Fax: +39 02 23993411, E–mail: amigoni@elet.polimi.it. Abstract – In this paper we present the implementation and the pre- liminary experimental validation of a distributed robotic system for monitoring magnetic fields. In particular, we concentrate on localiz- ing magnetic field sources starting from measurements performed by mobile robots equipped with magnetic sensors. I. INTRODUCTION The monitoring of electro-magnetic fields (EMFs) in indoor environments for verifying the levels for human exposure is an increasingly important activity. The use of distributed mea- surement systems composed of mobile robots carrying the sen- sors [1] can be indispensable when a large number of measure- ments is required, the measurement points must be determined at run time, and the environment is dangerous [2]. In this paper we present the implementation and the prelim- inary experimental validation of a distributed robotic system for monitoring magnetic fields. The general multirobot archi- tecture of our system has been introduced in [3], [4], where the sensors and the fields have been only simulated. The main original contributions of this paper are the use of real magnetic sensors (described in [2]) and the presentation of the results obtained from a preliminary experimental validation with real magnetic fields. In particular, we are concerned with the local- ization of magnetic field sources starting from the data returned by the sensors mounted on mobile robots. This paper is structured as follows. The next section presents the general architecture of our system. Section III discusses the issues connected to the use of our real magnetic sensors for source localization. Section IV shows the prelim- inary experimental results. Section V relates our work with part of the literature on distributed sensing. Finally, Section VI concludes the paper. II. THE GENERAL ARCHITECTURE Our multirobot architecture, presented in full detail in [3], [4] and called EMF MORO agency, has been developed for general tasks connected to monitoring EMFs. In particular, the robots of the EMF MORO agency are able to localize an EMF source in the environment and to circumnavigate it in order to verify the levels for human exposure. The architecture is hierarchical. A coordinator (a computer) supervises the activ- ities of the system, while a number of explorers (mobile robots equipped with EMF sensors) navigate in the environment and perform measurement tasks. The main cycle of activities of the EMF MORO agency can be summarized as follows: (a) the explorers (independently) measure the EMF in their current locations, (b) the coordinator integrates the measurements and builds an hypothesis about the location of the EMF sources, (c) the explorers move in the environment to reach new in- teresting measurement positions and the cycle starts again from (a). We explicitly note that the explorers are autonomous mobile robots that carry both “navigation” sensors (for example, cam- eras and sonars) and “monitoring” sensors (such as the EMF sensors). Currently, the representation of the EMF phenomena is limited to the representation of the estimated positions of the EMF sources within a grid map of the environment (that is supposed known). For simplicity, we considered a single EMF source. The position of the EMF source is given by the cell in which the source is located. The most significant feature of our system is the task alloca- tion of step (c) of the above algorithm. We use the contract net paradigm for allocating tasks to the robots [5]. The allocated tasks are “interesting” cells to be visited. The coordinator pro- poses a cell for visit if it needs an EMF measurement from that cell, either for triangulation purposes (to estimate the position of the source, see next section) or for determining whether the EMF is within the limits established by the law. The coordi- nator sends the proposed task to free explorers, waits for their proposed paths (or until a timeout expires), and assigns the goal to an explorer. A proposed path is a sequence of movements that brings an explorer from its current position to the cell to be visited. The coordinator considers the proposed path of mini- mum length and assigns the task to the proposing explorer. The other explorers are notified that their proposed paths have been rejected. To give a more precise idea of how this task allocation process works, we present one of the several tests we per- formed with simulated sensors and sources. In this experiment