Source-linked AI summary
Twins:Device-free Object Tracking using Passive Tags
Jinsong Han, Chen Qian, Dan Ma, Xing Wang, Jizhong Zhao, Pengfeng Zhang, Wei Xi, Zhiping Jiang
TL;DR
Device-free tracking needs a lower-cost way to monitor non-cooperative objects than sensor- and active-tag-based systems. Twins uses interference between passive RFID tags to create a critical state, models and validates that phenomenon, and implements tracking with commercial hardware. The prototype detects moving objects and reports 0.75m average localization error.
Problem
Existing device-free tracking solutions mainly use sensors and active RFID tags, which are more expensive than passive tags.
Method
Twins exploits a critical state caused by interference between adjacent passive tags, develops a structure-aware interference model, and combines KNN with particle filtering for tracking.
Results
0.75m average localization error is reported for the implemented Twins tracking scheme, with experiments demonstrating effective moving-object detection.
Takeaways & Limitations
Twins provides a practical passive-RFID approach for intrusion detection and trajectory tracking using existing RFID infrastructure.
Abstract
from arXiv · showhide
Without requiring objects to carry any transceiver, device-free based object tracking provides a promising solution for many localization and tracking systems to monitor non-cooperative objects such as intruders. However, existing device-free solutions mainly use sensors and active RFID tags, which are much more expensive compared to passive tags. In this paper, we propose a novel motion detection and tracking method using passive RFID tags, named Twins. The method leverages a newly observed phenomenon called critical state caused by interference among passive tags. We contribute to both theory and practice of such phenomenon by presenting a new interference model that perfectly explains this phenomenon and using extensive experiments to validate it. We design a practical Twins based intrusion detection scheme and implement a real prototype with commercial off-the-shelf reader and tags. The results show that Twins is effective in detecting the moving object, with low location error of 0.75m in average.
I. INTRODUCTION
Twins introduces passive-RFID device-free tracking by exploiting a critical state between coupled tags, addressing the cost and cooperation limits of prior approaches. The paper develops a structure-aware interference model, validates the phenomenon experimentally, and implements tracking with average localization error of 0.75m.
- Existing device-free tracking mainly relies on sensors or expensive active RFID tags, while passive-tag solutions lacked accurate and reliable motion detection.
- Twins creates a critical state in which coupling between adjacent passive tags makes one tag unreadable.The effect is attributed to reduced RF energy at the affected tag’s antenna.
- Moving objects can trigger state jumping by reflecting or refracting RF waves toward the Twins, making the previously unreadable tag readable.
- The proposed structure-aware interference model uses antenna-circuit structure and T-match analysis to explain the critical state, unlike structure-oblivious models.
- Twins combines KNN and particle-filter algorithms to approximate intruder trajectories and is implemented with commercial off-the-shelf RFID hardware.
- 0.75m average localization error is reported for the implemented tracking scheme, with extensive experiments validating passive-tag motion detection.
II. BACKGROUND
RFID uses passive tags for cost-efficient identification, but their antenna structure and near-field coupling shape how adjacent tags interact. The section motivates experiments and structure-aware modeling of these interactions.
- Passive RFID tags are battery-free and more cost-efficient than active tags, but have simpler circuitry and less communication capability.
- Passive tags communicate through far-field backscatter, while interference between adjacent tags occurs through near-field coupling.
- The T-match structure improves antenna-to-IC matching by adapting a short antenna to the capacitive IC load.
- Experiments use 20 commercial tags arranged into 10 Twins pairs, with tag spacing fixed at 10mm and reader distance at 2m.
- The placements of Twins and the minimum reader transmission power are varied to characterize how tag geometry affects readability.
- Existing structure-oblivious models predict equal tag currents, contradicting experiments that show shadowing and higher activation power for the Rear-tag.
B. Structure-aware model
The structure-aware model represents each passive tag with electric- and magnetic-dipole components aligned with its T-match structure. It derives unequal currents in adjacent tags and explains why the Rear-tag becomes shadowed.
- Structure-aware model: The model represents each tag antenna with an electric dipole line and a magnetic dipole rectangle, matching the T-match structure.
- Structure-aware model: Twins are modeled as four conductors: two lines and two rectangles, with the IC embedded in the magnetic dipole.
- Currents: The model identifies the tag currents with the currents in their magnetic-dipole components: I1 = I_S1 and I2 = I_S2.
- Theorem: The theorem establishes I_S1 < I_S2 for the two tags in the modeled Twins placement.
- Mutual inductance: Mutual inductance is computed among the two line and two rectangular conductors, including interactions within S1 and S2 and between S1 and S2.
- Critical state: When tag spacing is sufficiently large, the inductive coupling contributions become nearly identical; when l and b share a length scale, their difference can trigger the critical state.
IV. VALIDATION OF THE CRITICAL STATE
Experiments validate the critical state between adjacent passive RFID tags and show that moving objects trigger state jumping within a measurable detection region.
- A critical state occurs when one tag becomes unreadable while the Twins are interrogated.
- 15mm is the distance threshold below which the two tags exhibit a critical state at D = 2m.
- The rear tag B is consistently unreadable in the critical state and requires higher minimum PTX than tag A.
- State jumping events are measured by dividing the experiment area into cells and counting events caused by a moving volunteer.
- The effective detection region is about 2m × 1m between the reader and Twins, shrinking to 1m behind the Twins.
V. TRACKING MOVING OBJECTS
Twins tracks a single moving object by first detecting state jumps and then applying particle-filter tracking over a grid-covered monitored area.
- The tracking procedure first identifies state jumping on Twins and then uses a particle filter to track the object.
- A monitored region is partitioned into a 2D grid represented as a graph whose cells are vertices and adjacent cells are connected.
A. Identifying state jumping
State-jump identification must query many Twin pairs despite a reader power constraint that permits only one critical state at a time. MPLL uses graph-based priority polling to detect jumps quickly without starvation.
- A fixed PTX allows the reader to create a critical state for only one Twin pair at a time, making sequential querying necessary but potentially too slow.
- Naive priority scheduling can leave low-priority Twin pairs unqueried, causing scheduling starvation.
- MPLL addresses this problem with graph-based polling designed for timely movement detection without starvation.
- MPLL queries Twin pairs in a BFS pattern, retaining jumping pairs in the high-priority list and querying neighboring pairs.
- An access-status bit prevents unnecessary repeated queries within one polling round, while the algorithm identifies jumping pairs with high probability.
B. Localizing an object
The localization stage converts detected state jumps into a connected candidate region and estimates the object's position from the centroid of the Twins within that region.
- A smaller region inferred from state jumping provides a more accurate location estimate.
- Disconnected candidate subgraphs caused by multipath or noise are filtered by selecting a connected subgraph Gs.
- The estimated position is the centroid C = (X1+X2+…+Xk)/k of all Twin positions in the selected region.
- Centroid-based estimation might increase detection errors, so tracking accuracy is evaluated separately.
C. Tracking the object based on the particle filter
Twins tracking uses a particle filter to estimate an object's position from observed state-jumping events. Offline fingerprints support online prediction, weighting, resampling, and coordinate approximation.
- The particle filter estimates latent object states by repeatedly filtering and resampling random samples in the state space.
- The observation vector V records state-jumping counts n0 through n8 at the corresponding Twins during Δt.
- Offline training estimates the distribution of state-jumping counts at known positions and discretizes it into histogram fingerprints.
- Online tracking initializes particles near the origin, predicts their states, computes observation-based weights, resamples, and approximates object coordinates iteratively.
- The detailed particle-filter tracking procedure is provided in the authors’ technical report.
A. Proper detection region
The deployment geometry affects state jumping and detection reliability. The system is most suitable for single-object monitoring with Twins placed between the reader and the moving object, while multiple-object tracking remains unresolved.
- State jumping is easier when the object moves between the Twins and reader than behind the Twins, so experiments use the in-between deployment.
- Single-tag critical states are more unstable than Twins because ambient changes cause larger variation in the required reader transmission power.
- The study focuses on tracking a single moving object; multiple objects may create multiple subgraphs, but crossing or overlapping trajectories remain challenging.
- Reader and antenna cost can leave large surveillance areas uncovered when only a few RFID readers are affordable.
- The evaluation includes prototype hardware, warehouse deployment, and experiments assessing detection and tracking performance.
A. Hardware
The prototype uses unmodified commercial RFID hardware and evaluates calibration, warehouse deployment, detection settings, and tracking accuracy. Across these tests, detection depends on geometry and height, while average localization error is 0.75m.
- Hardware: The prototype uses an Impinj Speedway R420 reader, 920~928MHz antennas, 10–32.5 dBm transmission power, and 500 commercial E41-b passive tags.
- Calibration: Calibration varies antenna angle, reader-to-Twins distance, tag spacing, and height to identify settings for motion detection.
- Prototype: The warehouse prototype attaches each Twins to shelves spaced 2m apart, with adjacent Twin pairs separated by 0.6m.
- Experiment environment: Experiments evaluate tracking in a 30m×20m warehouse with a 1.70m intruder moving at 1.5m/s.
- Detection settings: 5.8m is the maximum deployment distance when PTX = 32.5dbm with the Impinj R420 reader.
- Detection settings: 95.17% average detection is achieved when Twins are 75cm above the floor, while 50cm or 100cm produce lower detection rates.
- Tracking accuracy: 0.75m is the average localization error for Twins, which outperforms LANDMARC and TASA in the reported comparison.
VIII. RELATED WORKS
Prior work includes device-based and device-free localization using wireless signals, sensors, active tags, and passive-tag interference. Twins extends this line with passive-tag critical states for device-free tracking and reports experimental effectiveness.
- Wireless activity sensing and tracking research has used GSM, WiFi, GPS, FM, acoustic signals, and other sensors.
- Device-based approaches require the target to carry a tracking device, whereas device-free methods monitor uncooperative objects through wireless-signal disturbances.
- Passive-tag localization has used tag interference and multipath profiles to locate desired items or objects.
- Earlier experiments and models studied interference among nearby tags, but Twins addresses a critical state involving two adjacent passive tags.
- Twins combines the critical-state finding with a practical tracking scheme and validates it through extensive real experiments.