RFID-based Indoor Positioning by Using Monte Carlo Algorithm

An indoor positioning scheme for a moving vehicle equipped with UHF radio-frequency identification (RFID) system is proposed by using Monte Carlo-based algorithm. By evaluating and updating the random sampling positions marked by passive RFID tags on ground, Monte Carlo algorithm can be implemented successfully for positioning. Performance of the positioning algorithm has been demonstrated by both numerical and experimental results. Introduction RFID-based tracking systems have been considered for wireless indoor positioning for years. RFID-based positioning system usually consists of RFID reader and specially-deployed passive RFID tags on ground. By interrogating the RFID tags, recording tags responses and combing the positions of RFID tags, position of RFID reader or platform carrying RFID reader can be obtained [1]. Owing to the limited answer range of passive RFID tags, RFID-based indoor positioning scheme may provide more accurate positioning results if compared with other wireless positioning indoor schemes [2]. Since the RFID readers and RFID tags are getting cheaper and cheaper, RFID-based positioning scheme has shown great potentials. The most common RFID-based indoor positioning scheme is constructed with a RFID reader on a moving platform connected to a single RFID reader antenna [3]. For positioning, scheme of [3] tried to range the distances between the reader antenna and RFID tags by utilizing the received signal strength indicator (RSSI). And several statistical algorithms based on RSSI have been proposed by eliminating or decreasing the interference or noise from the indoor environment [4,5]. Unfortunately, the RSSI may be severely distorted by the poor indoor wireless propagation due to interference from different obstacles. Moreover, measuring RSSI usually requires specially designed RFID reader. In this paper, we propose an RFID-based indoor positioning scheme by using ordinary RFID reader. Based on the position information of RFID tags, by using Monte Carlo algorithm, positioning or tracking of a moving vehicle equipped with a RFID reader can be implemented. The scheme doesn't need to measure RSSI and may avoid many problems in practical indoor environment. Monte Carlo-based Positioning Algorithm The model for RFID-based indoor positioning is shwon in Fig. 1. On the ground, UHF passive tags are arranged in a designeted pattern, and the data of tag's coordinates has been stored in the RFID system. The interrogation zone of RFID antenna on the ground, referred to as footprint, is shown in Fig. 2. RFID tags within the footprint can be recognized by the RFID reader. Their coordinates can be acquired by checking teh deployment pattern, which can be used for vehicle positioning. For positioning, Monte Carlo algorithm can be inplemented in two phases: prediction phase and update phase, which are given as follows. I. Prediction Phase. At time k , the antenna can detect some tags whose number is set ' n ' and the system can get the position coordinates of these tags. So we can estimate the position of the footprint center of the antenna preliminarily by Eq. 1. 4th International Conference on Machinery, Materials and Computing Technology (ICMMCT 2016) © 2016. The authors Published by Atlantis Press 1880 Fig. 1. RFID-based indoor positioning system. Fig. 2. Antenna footprint in moving. ' ' 1 2 1 2 ... ... ( , ) ( , ) n n k k x x x y y y ax ay n n        . (1) Where ' ' ( , ) k k ax ay is the estimated footprint center position and ( , ) i i x y is the position of the detected tag ith ( 1... ) i n  . The footprint center k o is defined as an arbitrary probability density function within the footprint near the position ' ' ( , ) k k ax ay . The probability coordinates of the footprint center are formed by N particles. Each particle is composed of the position hypothesis ( , )( 1... ) j j j s sx sy j N   of the antenna and the weight of the jth particle ( ) p j . {( , ( )) | 1,2, , } k j j S s p j j N    is the set of the sampling particles at time k . II. Update Phase. At time k , some particles can't meet the condition that if the footprint center is on the positions of these particles, the antenna can recognize the tags which are detected in fact. So for each particle like this, we find other two particles which meet the condition, take an average of the coordinates of the two particles and use the average to replace the particle position which doesn't meet the condition. At time 1 k  , the antenna has detected some other tags and has traveled a certain distance from time k . The particles at time k would travel with the vehicle in the same direction and the same distance, then reach the new positions at time 1 k  . Then we analyze if each new particle can meet the condition. If the particle can't meet the condition, we give up the particle at time k . Then we use the Gaussian probability model to get the weight of the remained particles described by Eq. 2. 2 1 2 ( ) 2 1 1 ( ) ( | ) 2 j k d l j k p j p d l e     