On the Optimal Frame-length Configuration on Real Passive RFID Systems M.V. Bueno-Delgado, J. Vales-Alonso Telematics Engineering Group, Technical University of Cartagena, Plaza del Hospital 1, Cuartel de Antiguones, Cartagena, 30202, Spain mvictoria.bueno, javier.vales@upct.es Abstract The majority of the anti-collision protocols proposed for passive RFID systems are based on Frame Slotted Aloha (FSA). They assume a classical result in FSA-based protocols which states that the theoretical identification throughput is optimized when the number of competing tags in coverage equals the number of slots in the frame. However, this is not exact in real RFID systems, as the so-called capture effect is neglected. The capture effect occurs when a tag identification signal is successfully decoded from a collision slot. This paper analyzes the identification performance of real RFID systems, taking into account not only the capture effect, but also the requirements imposed by the de facto standard EPCglobal Class-1 Gen-2. The analysis is addressed by Discrete Time Markov Chains. From the analysis, a set of relevant results is extracted: the frame-length values that, configured into the readers studied, guarantee the best identification performance (maximum throughput). The analytical results have been confirmed by means of simulations and by a set of measurements performed on a real passive RFID system. Results closely match the analysis predictions, which demonstrate a notable impact of the configuration on the performance. Keywords: FSA, Markov Chain, Capture Effect 1. Introduction In RFID systems, the communication between readers and tags takes place in a shared communication channel. An anti- collision mechanism is required to minimize the collisions caused by simultaneous transmissions. Meanwhile, in passive RFID, the extreme simplicity of the tags is a severe constraint on the design of collision resolution methods, and complexity must rely almost exclusively on the reader. UHF passive RFID readers available in the market imple- ment the anti-collision protocol EPCglobal Class-1 Gen-2 (aka EPC-C1G2) [1], based on a variation of Frame Slotted Aloha (FSA). As Figure 1 shows, the reader divides the time into identification cycles (frames). Each frame is in turn divided into slots. At the beginning of each identification cycle, the reader announces the length of the frame (K slots) with a Query packet. Tags in coverage receive the information and randomly select a slot in that cycle to transmit their identifier [2]. If there are not tag responses (z=0) in a slot, it is considered empty. If only one tag answer in a slot (z=1), it is considered successful, and those slots with two or more simultaneous tag responses (z ≥2) are collision slots. Then, having a frame with K slots and N tags competing, the fill level of z tags in a slot is is given by the binomial distribution function: Pr(z) =  N z  1 K  z  1 − 1 K  (N−z) (1) The expected number of slots filled with exactly z tags is given by E(z)=K · Pr(z), and the theoretical throughput (Ω) of FSA is calculated as follows: Figure 1: FSA procedure Ω= E(z = 1) K = N K  1 − 1 K  (N−1) (2) When the number of tags is much larger than the number of slots or vice versa, the identification delay increases and the throughput is negatively affected. The anti-collision protocols based on FSA perform optimally if N =K giving the theoretical maximum throughput Ω=e −1 ≈0.36. However, current readers on the market cannot achieve this maximum throughput due to the following reasons: • EPC-C1G2 restricts the frame-length to {K = 2 Q : Q = 0,..., 15}. • Commercial readers do not adjust K depending on N. Preprint submitted to Journal of Network and Computer Applications March 25, 2010