RFID for mobile applications Arne Bosien, Volker Turau Institute of Telematics Hamburg University of Technology (TUHH) Hamburg, Germany {arne.bosien|turau}@tuhh.de Abstract—The availability of fast anti-collision algo- rithms is crucial for most RFID applications. This pa- per aims to evaluate these algorithms for applications in which it is not intended to identify the entirety of moving objects but to detect as much tags as needed to allow orientation. The navigation of Automated Guided Vehicles (AGV) by distributed landmarks is an example which clarifies the discriminative requirements com- pared to supply chain tasks. For the former purpose redundant information can be gained from different tags. This requires the detection of an application dependent percentage of all tags. Because AGVs are moving, the detection and read and write operations have to be close together and very fast, since repetitive communication is not always possible. I. I NTRODUCTION In traditional RFID applications it has been re- quired to detect all RFID tags in range. This may be useful if the determination of all products in a shopping cart is wanted, but this is not suitable for all imaginable uses. The availability of fast anti-collision algorithms is crucial for most systems, but the employment in mobile applications makes higher demands on the communication speed because the reader is moving. A. The meaning of anti-collision To allow the communication with a specific RFID tag, all tags in the range of the RFID reader have to be identified at first. For this purpose the reader sends a single command which causes all tags to respond and send back their ID to the reader. Since the tags are not able to communicate with each other, the transmission of their responses leads to collisions, which inhibits the identification of the tags for the reader. To solve this challenge, anti-collision algorithms are required. B. RFID for navigation purposes Several approaches make use of stationary tags for orientation [Zam07], [Pec08] or navigation services of AGVs [BVT08], [NBOF06]. In this context it is not necessary to determine all tags by all means, and the speed of detection may overrule the importance of the completeness of an inventory if a high driving speed is necessary. 1) Example: Detection of one tag: To detect one single tag, which is exactly lying on the path of the reader, with a realistic detection time of t inv = 0.030 s and a reader range of r =0.15 m a maximum speed of v = x t = 2r t inv = 0.30 m 0.030 s = 10 m/s is possible. From the Nyquist-Shannon sampling the- orem follows that at least two scans are necessary to detect the tag for sure. Therefore, the speed reduces to v =5 m/s. 2) Detection of more tags: Furthermore, the max- imum speed is reduced if more tags are within the interrogation area. The number of requests to detect n tags with a simple anti-collision algorithm (described in section III-B) and incremented tag IDs can be calculated as the number of nodes of a binary tree. The height of this tree is given by h = log 2 (n). The time for a complete anti-collision cycle then can be estimated as: t inv (n) =  2 h+1 - 1  t inv =  2 log 2 (n)+1 - 1  t inv The maximum speed for n =4, t inv =0.030 s and r =0.15 m is reduced to: v = 1 2 · 2r ( 2 log 2 (n)+1 - 1 ) t inv ≈ 0.71 m/s 3) Detection of boundary tags: As the reader is moving, the distance x, while a tag is in range of the reader, also depends on the offset y to the path (see Figure 1), which can easily be computed as: x = 2 p r 2 - y 2 Increasing the radius leads to a greater period when the detection is possible but unfortunately also to a higher amount of included tags if the tag density remains unchanged.