An Optimal Q-Algorithm for the ISO 18000-6C RFID Protocol
An Optimal Q-Algorithm for the ISO 18000-6C RFID Protocol
复制标题
DOI：
10.1109/tase.2008.2007266
复制
发表时间：
2009-01-01
影响因子：
5.6
通讯作者：
Pappu, Ravikanth
中科院分区：
文献类型：
--
作者：
Maguire, Yael;Pappu, Ravikanth
基于国际标准化组织/国际电工委员会18000-6C(又名EPC Gent)协议的无源射频识别系统的典型读取率高达每秒1200个唯一的96位标签。这种性能部分是通过使用称为Q算法的媒体访问控制算法来实现的，该算法是时隙ALOHA多用户信道访问算法的变体。分析了国际标准化组织/国际电工委员会18000-6C射频识别空中接口协议所采用的介质访问控制算法，并给出了实现最佳读取率的步骤。我们还表明，在许多实际用例中，理论性能可能会超出理论性能，并提供了一个将真实世界数据纳入读取率估计的模型。在频谱的一端，在纯理论方法中，通过获取每个比特的持续时间并计算每秒可以解码的总比特数来估计速率。这种方法没有考虑任何协议开销或实际情况。在纯实验方法中，当标签、读取器、固件、协议等几个因素不同时，使用标准测试用例来比较读取率。这两种方法都不能很好地理解一般情况下的读取率估计问题。在本文中，我们通过建立Gent介质访问控制层的冲突概率的第一性原理模型来解决这个问题。标签响应的冲突是决定Gent系统中读取速率的主要因素。使用该模型，我们证明了协议的最坏情况下的效率不低于36.8%，即每单位时间应该可以看到超过36.8%的给定标签群体。我们开发了一个动态Q-算法，它的性能远远好于最坏的情况，并展示了它相对于静态Q-算法的性能。然后，我们放松了上述算法背后的假设，以便能够结合真实世界的情况，并提供一个框架，其中从业者可以对特定情况进行一些测量，并使用我们的模型来估计预期的阅读率。需要考虑的三个重要因素是：(I)不同类型的时隙占用的不同解码时间；(Ii)捕获效应，其中，由于后向散射功率足够不同，两个占用时隙被解码为有效标签；以及(Iii)后向散射功率的分布。我们开发了一个模型来解释这三个因素。尽管我们的模型做出了几个假设，但我们设计和部署的阅读器几乎证明了所有这些假设的合理性。我们目前正在致力于开发一组标签的后向散射功率分布的更深层次的表征。这将使我们能够利用阅读器的信号处理能力来消除两个占有率的歧义，并提高阅读率-远远高于我们的模型预测的阅读率。这是我们目前研究的重点。
Passive radio-frequency identification (RFID) systems based on the ISO/IEC 18000-6C (aka EPC Gent) protocol have typical read rates of up to 1200 unique 96-bit tags per second. This performance is achieved in part through the use of a medium access control algorithm, christened the Q-algorithm, that is a variant of the Slotted Aloha multiuser channel access algorithm. We analyze the medium access control algorithm employed by the ISO/IEC 18000-6C RFID air interface protocol and provide a procedure to achieve optimal read rates. We also show that theoretical performance can be exceeded in many practical use cases and provide a model to incorporate real-world data in read-rate estimation.Note to Practitioners-Estimating read-rates in RFID has always been something of a black art. At one end of the spectrum, in the pure-theory approach, rates are estimated by taking the duration per bit and calculating the total number of bits that can be decoded per second. This approach does not take any of the protocol overheads or real-world conditions into account. In the pure-experimental approach, a standard test case is used to relatively compare read-rates as several factors-tags, readers, firmware, protocols, etc., are varied. Neither of these approaches really provides any insight into the problem of estimating read rates for the general case.In this paper, we take on this problem by developing a first-principles model of collision probability in the Gent medium access control layer. Collisions of tag responses are a dominant factor in determining read rates in Gent systems. Using this model, we show that the worst case efficiency of the protocol can be no less than 36.8%, i.e., it should be possible to see more than 36.8% of a given population of tags per unit time. We them develop a dynamic Q-algorithm that performs much better than the worst case, and show its performance relative to a static Q-algorithm.We then relax the assumptions underlying the above algorithm so as to be able to incorporate real-world situations and provide a framework wherein practitioners can make some measurements of a particular situation and use our model to estimate expected read rates. Three important factors that need to be considered are: (i) the different decoding times for different types of slot-occupancy; (ii) the capture effect, wherein a two-occupancy slot is decoded as a valid tag because the backscatter powers are sufficiently different; and (iii) the distribution of backscatter powers. We develop a model to account for these three factors.Although our models make several assumptions, we have designed and deployed readers that justify almost all of them. We are currently working on developing a deeper characterization of the backscatter power distribution of a population of tags. This will allow us to use the signal processing capability of our readers to disambiguate two-occupancy slots and boost read rates well-above those predicted by our model. This is the focus of our current research.