Comparison of a Fast Probabilistic Propagation Model against an Analytical Computational-EM Model and Measurements for the Evaluation of Passive RFID Systems Antonis G. Dimitriou 1 , Achilles Boursianis 2 , Ioannis Markakis 2 , Stavroula Siachalou 1 , Theodoros Samaras 2 and John N. Sahalos 2 1 School of Electrical & Computer Engineering, Aristotle Univerisity of Thessaloniki, Greece, antodimi@auth.gr 2 Physics Department, Aristotle Univerisity of Thessaloniki, AUTh, Thessaloniki, Greece, bachi@physics.auth.gr Abstract—This paper presents the comparison of a fast prob- abilistic propagation model against an analytical computational electromagnetic (EM) model and measurements performed for the evaluation of passive RFID systems. The results of the probabilistic model compared to the ones of the analytical model, as well as to the measurements, are in fairly good agreement. Index terms—probabilistic model, analytical model, RFID. I. I NTRODUCTION In this paper a comparison between a fast probabilistic model and an analytical Full-Wave model is presented. Both models aim to evaluate the identification performance of passive RFID systems operating at the UHF frequency band (860MHz-930MHz). The probabilstic model, [1], calculates the probabilities of succesfull identification of passive RFID tags at the specified locations. It has been designed so that it can be integrated in automated planning softwares [2], where numerous estimations are needed in small time. Therefore, a rough description of the surrounding environment is only considered, without considering furniture or other objects. Only major indoor propagations mechanisms are considered [4]; that is the direct field and multiple reflections. However, all important characteristics of passive RFID systems are included in the estimations, like antennas’ radiation patterns, polarization of the tags and the reader’s antennas etc [3]. The analytical Full-Wave model (FDTD), [5], considers a detailed representation of the actual environment. 285 solids representing actual objects inside the simulation room were included in the estimations. The model results in an actual ”screenshot” of the field inside the simulated area, where maxima and minima are shown. The simulation time is pro- hibitive for planning applications. However, it can be used as a benchmark for the evaluation of the performance of the ”abstract” probabilistic model. The models are compared against each other and against measurements conducted in a real environment. Comparison of the estimations demonstrated good agreement. Furthermore, both models were compared with measurements conducted in the simulated room. Again, both models demonstrated good agreement. The simulation time was less than 4s in an average laptop for the probabilistic model, while for the same problem, the Full-wave model needed 135h in an advanced workstation. II. PROBABILISTIC MODEL The probabilistic model was analytically presented in [1]. The probability of successful identification of passive RFID tags is calculated. Line of Sight Conditions (LOS) are ex- pected, justified by the power constraints of battery-less RFID systems [6]. A passive RFID tag is typically considered suc- cessfully identified, if the power that reaches the tag is greater than its wake-up threshold, assuming that the sensitivity of the reader is small enough to receive the backscattered signal from an ”awaken” tag. The probability of successful identification equals the probability that the instantaneous power at the tag IC is greater than its wake-up threshold γ . In the presence of a strong LOS path, fading is well described by a Rician prob- ability density function. Hence, the probability of successful identifications is P (X ≥ γ )=1 - F X (γ |ν,σ), (1) F X (x|ν,σ)=1 - Q 1  ν σ , x σ  , (2) and ν 2 is the power of the LOS path, 2σ 2 is the average power of the other contributions x is the signal’s amplitude and Q 1 (a,b) is the Marcum Q-function. Therefore, by defining ν and σ at each reception point, we can calculate the desired probability for a single reader-antenna configuration. For the calculation of the average power of the multiply reflected rays (2σ 2 in (2)), we consider ray-clusters that include all rays that initially bounce on the same wall, e.g. for a typical room/building, six ray clusters are considered for the six surrounding walls. Within each cluster, we consider the phases of the rays as random variables (key assumption of the model), identically and independently distributed, uniformly over [0, 2π]. Furthermore, we approximate the magnitude of