Efficiency analysis of rfid tag collision-resistant mechanisms with multiple labels read together
RFID Knowledge

Efficiency analysis of rfid tag collision-resistant mechanisms with multiple labels read together

PublishedShanghai Zejie Smart Technology Co., Ltd.

Three hundred labels on a stack, and five minutes after reading it, it's not finished.

Last year at an auto parts warehouse, we saw a familiar problem. More than 300 reusable bins were stacked neatly on one pallet, each carrying an RFID tag, an electronic tag that communicates by electromagnetic backscatter and can be read at a distance. A worker pushed the pallet through a portal. After almost five minutes, the reader's progress bar was only at 70%. Dozens of tags still would not read and had to be entered manually. The root cause was signal collisions when multiple tags entered the antenna field at once. The anti-collision mechanism directly determines the upper limit on simultaneous read performance.

How did the collision happen?

In the RFID system, the reader is "questioner" and the label is "responder". When multiple labels are activated at the same time, almost immediately, the signal is superimposed in the air, the reader receives a bunch of mixed wave shapes that cannot be removed. This is collision, not equipment failure, but natural challenges posed by the RFID working method.

The common saying is, "Readers say "the tags present" and "identify them." If a hundred labels together, the reader can't hear anything; if they are divided, line up according to the rules, the efficiency is completely different. This is the rule of "queuing," which is what collision-proof algorithms solve.

Two mainstream technology routes: ALohA system and tree type

The collision-resistant algorithms are broadly divided into two families. The ALOHA system (a time-repeated protocol based on "random retreat" and then re-dispatch without a time gap) is simple but has a maximum amount of swallowing; the other type is tree-based algorithm (a defined agreement based on binary query, which removes collision space repeatedly by two parts until the only remaining label), and the process is determined but varies much longer.

ALOHA is: from pure random to time interval

The family has the advantage of having simple agreements and low costs, with early ISO 14443 using similar thinking. The disadvantage is that when it swallows a ceiling and labels are far above time gaps, performance drops off the cliff.

Tree algorithm: definitive decomposition collision

The tree algorithm changes the same approach: Reader sends a prefixed query command, which only returns with a matching tag for an ID prefix. If you have a collision, add another prefix to continue sifting until only one label responds. Repeats, all labels are identified.

The advantage of this algorithm is certainty, and there will be no "labels never read" starvation; the cost is that the number of interactivity with the label ID length index increases, the average reading time for a single label is high, and the amount of vomiting falls between 30% and 45%.

Active Main Power: EPC Gen-2 Self-adaptation Q Algorithm

Physical storage, logistics, retail scenes, which are basically UHF labels that meet the criteria of the EPC Class-1 Gen-2 (also known as ISO 18000-63). The core of this protocol is self-adaptation to the ALOHA system's "Spacing Gap" idea and introduce Q value parameters.

Q values determine the time gap of each frame N = 2Q. Reader first sends a Query command with Q, and the labels select a time gap between 0-2Q-1. Reader and write on the result of this frame - empty space, collision time interval, proportion of successful time gap - automatically adjust Q: Increase Q by increasing the time gap by increasing Q by more than one time. Successfully, make Q save time. This way, regardless of the label density, the system can approach its optimal state.

Gen-2 protocols match the original Q values and adjust steps (typical Q range 0-15) that are reasonable, with 100 to 200 labels per second for single antennas under a laboratory with a flat distribution of labels; it is not difficult to keep this medium density, the scenario where the label is coded, at 150 tags/seconds or more. Shanghai-based Quitu will typically preset the Q threshold in 4-6 years, then use a oscillator to fine-tune back-transmitting, so as to avoid system harvesting over a dozen rounds.

The real factors that influence and capture efficiency

The agreements are only the basis, on the ground, and their efficiency is often being undermined by some of the "non-calculations" factors. These are key points that have been repeatedly tested in the project.

Match the number of labels to the number of time slots

Framed slotted ALOHA is most efficient when the number of slots is close to the number of tags. If 300 tags are present and a frame has only 16 slots (Q = 4), an average of nearly 20 tags will try to use each slot, making collisions extremely likely. At the other extreme, opening 4,096 slots (Q = 12) for 50 tags leaves most slots empty and makes each read round take far longer. That is why Gen2's adaptive Q is so important.

Reader Power and antenna configuration

The high power is a new problem: the remote labels are activated but the return signal is too weak, and the readers do not receive them, so they remain in the gap and fail to respond, creating an "unresolved collision." Experience is that the warehouse portals, such as the one that controls radiation power at 1-2 W (the full-irradiation power of equivalents), and the Circumpolar antenna 6-8 dBi can get a cleaner label activation area at 3-5 metres.

Labels towards medium

The surface of the metal is a serious reflection of the UHF signal, and the liquid absorbs electromagnetic waves. If the label is on the same tray, it is placed on the metal, separated from the plastic case, and the number of "effective visible labels" that the reader sees is often only half the true number. The improvement is to mark the metal with anti-metal labels or to isolate bubbles above 5 mm between the label and the metal.

Environmental interference and speed of label movement

Wi-Fi, Bluetooth, and other RFID readers can cause the same frequency interference, with labels moving at high speed (e. g. transfer belts 2 m/s above) across multiple time slots, and a significant drop in the hit rate. The common response is to increase the number of antennas into overlay arrays or to keep the transmission belt within 1 m/s, so that each label can be within at least 3-5 time slots.

A set of measured data: gaps in optimization and post-optimal

Take a typical dresser-store board view. The store has 1,500 items with a tag on them, and closes the door with a hand-held reader to finish the count by walking around the shelf.

The initial settings, using default Q values, 1W fixed power, single antenna scans, take 1,500 labels for 18 minutes, with a missing read rate (unrecognized tags) of about 6%. After adjustment, the start value of Q is changed to 5, power is moved to 0.5W, handheld is closer, and two slow trips are made, down to 7 minutes, and the drop rate is 0.4% of the same device 's label, all in parameters and operational details.

I'm gonna make a few practical suggestions for efficiency.

  1. First, you can use a oscilloscope or protocol analysis tool to grab a frame back and observe the percentage of idle/crash/success time slot as an initial reference for Q values.
  2. The power is kept to the minimum of "stabilized reading of the remotest label" and would prefer to be dense or not to fail in a distant field.
  3. Metals are marked against metal, and the liquid environment is designed inlay, with labels in the direction of polarization as vertically as possible.
  4. The transmission belt and high-speed scenes are activated with multi- antenna partitions, which will allow the label "Standing" to avoid collage within a frame.
  5. The bulk scene is first tested on a small scale, and the best Q, power, antenna layout is detected, then pushed to the whole production line or warehouse.

The mechanism for collision prevention determines how fast the RFID system can read, and how good it is, but no protocol can get the best solution in all scenarios. Gen-2's self-adaptation Q algorithm is the best compromise of current commonality, grasping it, then targeting labels, power, antennas, environment, most of which are capable of being solved by engineering. The remaining small edge case, often not a algorithm, but a physical dimension or process design itself needs to be adjusted.