Modified Slotted ALOHA Protocol for RFID Tag Collisions 1 Wei Zhou, 2 Gaurav Kapoor, 2 Selwyn Piramuthu 1 Information Systems and Operations Management ESCP Europe, France 2 University of Florida, USA wzhou@escpeurope.eu, {grkapoor,selwyn}@ufl.edu Abstract As RFID tags become popular, their volume in both the number of applications and the sheer number per application is bound to increase. This increase in density of RFID tags at any given location invariably results in collision problems that occur when a reader tries to simultaneously read multiple tags that are present in its field. The Slotted ALOHA protocol and its variants have been used over the years to alleviate similar collision problems in communication systems. We consider the basic Slotted ALOHA protocol for RFID applications, and propose a modified approach with improved performance characteristics. 1. Introduction The framed slotted ALOHA has been proposed as a means to alleviate problems associated with collisions when multiple RFID tags are simultaneously present in a reader's field. Almost all of these approaches assume that the reader first transmits its frame size to the tags. The tags then randomly pick a slot and reply in that slot. From the reader's side, in any given slot, it receives (a) no signal, (b) signal from one tag, or (c) signal from more than one tag. The slot under case(a) is considered a loss since no useful information was transmitted during that slot. The slot under case(b) is a successful case since the signal from this tag is received by the reader. Collision occurs in case(c), and the number of tags that replied in this slot is generally unknown. The minimum number of tags involved in collision is clearly 2 and the maximum probabilistically depends on the number of slots and the number of tags in the field of the reader. It's hard to estimate the initial number of tags present in the field of the reader. However, the frame length depends on the number of tags. In reality, the reader initially guesstimates the frame length (i.e., the number of slots per frame) from past experience and broadcasts it to the tags in its field. The tags then choose a slot to reply to from among the available slots. Existing papers on RFID collision (as well as ALOHA, in general) assume that the tags choose slots to reply to using a uniform distribution (e.g., [1],[2],[4]). However, in practice, the random numbers chosen tend to bunch together such that several tags respond to the same slot while many other slots remain vacant (i.e., with no signal from any tag). Based on the number of collisions, the reader dynamically adjusts the frame length and repeats this process until there are no collisions. In principle, the frame length initially chosen can be made large enough to accommodate all the tags. However, this could drastically slow down the tag reading process. At the other extreme, the frame length can be chosen such that only one tag is read in each read cycle. However, since tags often move in and out of the field of the reader quite frequently, the probability of missing some tags increases as the number of read cycles increases. Therefore, there is a tradeoff between frame length and the time it takes to read all the tags that are present in the field of the reader. Piramuthu ([5]) used a knowledge-based system to dynamically choose the appropriate frame length based on the estimate of the number of tags in the reader's field. The knowledge-base was generated from past experience in the domain of interest. Although this is a step in the direction, there is still a need to improve on the randomness with which the tags choose a slot to reply as well as the frame size chosen during each cycle. There are at least two issues that need to be addressed in this scenario: (1) How to dynamically modify the frame length during each tag read cycle, and (2) instead of letting the tags choose among the slots uniformly (i.e., UNIFORM(0,frame-length)), can this process be improved to more accurately reflect the number of tags in the field of the reader as well as the associated collision probabilities? In the general communication systems literature, both sides (i.e., reader and tag in the RFID scenario) are assumed to have a reasonable amount of processing power to accomplish this process. However, with its extremely constrained resources including limited memory and processing power, the basic RFID tags that are commonly used are incapable of any computations related to collision- Copyright © 2009 by the Institute of Electrical and Electronics Engineers, Inc. All rights reserved.