Self-Location of a Mobile Robot with uncertainty by cooperation of an heading sensor and a CCD TV camera E. Stella, G. Cicirelli, A. Distante Istituto Elaborazione Segnali ed Immagini - C.N.R. Via Amendola, 166/5 - 70126 Bari (Italy) E-mail: stella@iesi.ba.cnr.it Abstract Goal-oriented navigation of a mobile robot by land- mark based techniques is a straightforwardand suitable ap- proach. Most of methods, normally, give an estimation of the position and orientation of the vehicle, but, often, they are not able to provide a good estimate of the uncertainty in the measurement. That information is useful in application where multisensor fusion is requested. Our method permits to determine the vehicle location and relative uncertainty, when its orientation is obtained by an heading sensor, using a visual landmark based method. The approach is straight- forward and suitable for real time performance on general purpose hardware. Experimental results are provided by im- plementation on our autonomous mobile vehicle SAURO. Keywords: Camera Localization, Navigation, Autonomous mobile vehicle, Self-Location. The research is partially supported by Progetto Speciale Robotica (Comitato Scienze Fisiche) of the C.N.R. 1. Introduction Self-location is the capability of an autonomous mobile vehicle to determine its position in an environment. Visual approaches to self-location are suitable for navigating a mo- bile robot in indoor contexts. Two classes of methods can be considered: Absolute technique: are those approaches providing an estimation of the location and orientation of the ve- hicle using artificial or natural landmarks whose posi- tions in the environment are a-priori known [11, 4, 7]. Relative technique: are those methods where only the relative motion of the vehicle is estimated (rota- tion and translation) so for goal oriented applications the absolute location of the start point must be known. Those techniques, however, do not require any a-priori knowledge about the environment [6, 1, 2]. Methods in both classes work with a good performances (referred to execution time) and provide an acceptable es- timation of the vehicle location and orientation. Most of them, however, have a fundamental drawback: they do not provide an estimation of the uncertainty in the vehicle loca- tion and orientation. In the last years, a common tendency of robotic researchers has been to analyze the problems of the mobile vehicles in term of multisensor fusion framework, so the estimation of uncertainty of sensor measurements has played an important rule. In this paper we propose a straightforward approach to self-location of a mobile robot by localizing the Center-of- projections (CP) of the camera when the orientation of its optical axis is given. In fact, using an heading sensor (we have used a compass), the orientation of the vehicle can be recovered, so three landmarks are enough to estimate loca- tion of the camera and relative uncertainty. The experimen- tal setup is made up by three beacons (infrared LEDS) used as landmarks at known positions in the environment. Using the excellent angular resolution of CCD cameras, visual an- gles are evaluated and, as described in sect. 2, the location of the CP and optical axis orientation are recovered. The visual angle of a landmark is the angle formed by the ray from the CP to the landmark and the optical axis (see fig.1). The tech- nique used to estimate the visual angle is similar to [3]. We have used artificial beacons, in order to speed up the search for the landmarks in the image, but natural features could be considered. The technique is described in section 2., the estimation of uncertainty is explained in sect.3. . The experimental setup is described in sect. 4 and, finally, the results on self-location of a mobile robot are shown. 1