1-4244-0342-1/06/$20.00 ©2006 IEEE ICARCV 2006 LOCAL POSITIONING SYSTEM OF A MOBILE ROBOT : A PRACTICAL PERSPECTIVE. Iman H. Kartowisastro Computer Engineering Department Bina Nusantara University Jakarta, Indonesia. imanhk@binus.ac.id Abstract— Localization problem in determining position and orientation of an autonomous robot with respect to a world reference frame is commonly encountered. Personal service robot which has the largest share in the growth of service robot needs a simple and affordable system to overcome this problem. This paper presents an alternative approach to solve localization problem by utilizing Local Positioning System based on ultrasonic sensors. With this system, 3D information in regards to position and information of a mobile robot can be obtained fast enough to cope with a robot moving with a speed of 30 cm/s. The use of LPS to enhance the performance of control system has been already proven for a mobile robot moving on area with high irregularity surface Keywords— Local Positioning System (LPS), personal service robot, absolute measurement, relative measurement, feedback control system. I. INTRODUCTION Penetration of service robots into market grows significantly in which estimated growth of market value in 2007 will be 2.5 times of value in 2003 and the population number in 2007 is predicted to be 6 times of population in 2003 [1]. The biggest portion of this growth falls into a category of personal/ domestic service robots. Hence, research in this category of service robots are conducted intensively world wide. Furthermore, in the ubiquitous computing environment, service robot, such as IDRO [5], can take advantages of it and this in turn will accelerate the opening of the service robot market in the post PC era. Works in the field of robot navigation is important as mobile robot platforms which constitute most personal robots depend heavily on the navigation capabilities and environmental navigation in turn provides a basis to carry out tasks. Localization problem in determining position of a robot can be solved via a hybrid approach, namely by using data base of reference images throughout the regions and the use 802.11b wireless network signals [6]. In this work, directing the camera to obtain a view of the upper portion of the region and polling the wireless signals coming out from several access points will allow computer to process those data in obtaining information about where the robot is. Some researchers proposed a multi agent based architecture for outdoor mobile robot navigation [2]. A mobile robot equipped with CCD cameras, binocular CCD cameras, two LADARs and a GPS device was used in this work, and it illustrated that the architecture improved the reasoning ability about the world by making use of apriori global knowledge. Ultrasonic/ sonar sensors as range finding sensors are commonly used to support localization tasks. Multilobal detection patterns can be modeled and neural network approach is implemented to solve problem in localization [11]. The presence of this sensor in detecting objects or environments can be used for concurrent mapping and localization purposes. Identification and localization of environmental features are taken from sparse and noisy sonar data. With the use of stochastic approach, a map building process can be carried out. A sequence of independent limited size stochastic maps were then joined in a globally consistent way [9]. Others work more on algorithm aspects, such as localization algorithm [3]. Simulation study showed that the algorithm gave results in 5.51 seconds with an error of 15 cm. This result is somewhat longer in real time condition, considering during this period a robot may already be moving to somewhere else or may hit anything surrounding it. Localization problem is further expanded into a team of robots in which each robot is equipped with proprioceptive and exterceptive sensors [7], [8]. An extended Kalman filter for fusing the data coming out from these 2 sensors are used and from simulation result the accuracy on the localization is improved by the use of relative bearing. Another work based on Kalman Filter were also proposed to perform global localization with the use of multiple hypothesis tracking technique [10]. This approach of using multi hypothesis Kalman Filter based pose tracking has an advantage over a more common grid based methods in such away that incomplete world model information can be utilized. Even though there are so many research carried out in the field of localization and navigation, however, few works are dedicated to producing simple systems but robust enough which are developed specifically for personal service robots in which price, affordability, and simplicity aspects are very important. There are so many algorithms developed in localization subjects, but implementations are sometimes not easy due to obtaining data in real time fashion and impractical problem imposed. This paper presents a practical perspective of