An Online RFID Localization in the Manufacturing Shopfloor

Andri Ashfahani, Mahardhika Pratama, Edwin Lughofer, Qing Cai, Huang ShengarXiv 2018 · DOI · OpenReview

Abstract

Radio Frequency Identification (RFID) technology has gained popularity for cheap and easy deployment. In the realm of manufacturing shopfloor, it can be used to track the location of manufacturing objects to achieve better efficiency. The underlying challenge of localization lies in the non-stationary characteristics of manufacturing shopfloor which calls for an adaptive life-long learning strategy in order to arrive at accurate localization results. This paper presents an evolving model based on a novel evolving intelligent system, namely evolving Type-2 Quantum Fuzzy Neural Network (eT2QFNN), which features an interval type-2 quantum fuzzy set with uncertain jump positions. The quantum fuzzy set possesses a graded membership degree which enables better identification of overlaps between classes. The eT2QFNN works fully in the evolving mode where all parameters including the number of rules are automatically adjusted and generated on the fly. The parameter adjustment scenario relies on decoupled extended Kalman filter method. Our numerical study shows that eT2QFNN is able to deliver comparable accuracy compared to state-of-the-art algorithms.

Introduction: The Challenge of Shopfloor Localization

In the "Industry 4.0" era, real-time tracking of manufacturing objects (MOs)—such as tools, trolleys, and materials—is critical for operational efficiency. RFID technology is a preferred choice due to its low cost and ease of deployment. However, the manufacturing shopfloor presents a hostile environment for radio signals; metal machinery causes multipath reflections, and the movement of personnel and equipment creates constant interference.

System architecture for RFID-based localization on a manufacturing shopfloor.

The primary technical hurdle is that Received Signal Strength (RSS) data is non-stationary. Traditional static machine learning models struggle to adapt to these shifting patterns without frequent, costly retraining. This research addresses the need for a "life-long learning" strategy that can update itself autonomously as data streams in.

Methodology: The eT2QFNN Framework

The authors propose the evolving Type-2 Quantum Fuzzy Neural Network (eT2QFNN). This system is built upon three technical pillars designed for robustness and real-time performance:

  1. Interval Type-2 Quantum Fuzzy Sets (IT2QMF): Standard fuzzy systems often use Gaussian curves. In contrast, eT2QFNN utilizes quantum fuzzy sets with graded membership degrees. This allows the system to more accurately distinguish between overlapping location classes in noisy signal environments.
  2. Evolving Architecture: The model begins with an empty rule base. It employs a modified Generalized Type-2 Datum Significance (mGT2DS) mechanism to evaluate incoming data. If a data point is sufficiently unique, the network automatically generates a new rule, allowing the "brain" of the system to grow as needed.
  3. Decoupled Extended Kalman Filter (DEKF): To ensure the system remains fast enough for edge deployment, the authors use DEKF for parameter tuning. By only adjusting the "winning" rule (the most relevant one for the current data), the system maintains high accuracy with minimal computational overhead.
Architecture of the evolving Type-2 Quantum Fuzzy Neural Network (eT2QFNN).

Experimental Setup and Results

The model was validated using a real-world setup at the SIMTech laboratory in Singapore, involving a "Smart Rack" equipped with reference RFID tags. Over a 20-hour period, the system processed 283,100 observations.

The eT2QFNN was benchmarked against four state-of-the-art evolving algorithms (gClass, pClass, eT2Class, and eT2ELM). Key results include:

  • Near-Perfect Accuracy: The system achieved a 99% classification rate in 10-fold cross-validation, outperforming all competitors.
  • High Efficiency: It was the second-fastest algorithm tested, proving its suitability for real-time applications.
  • Compact Logic: The network maintained a very small rule base (averaging 6 rules), demonstrating that efficient logic can replace architectural complexity.

Conclusion

The eT2QFNN offers a powerful solution for the non-stationary challenges of shopfloor localization. By combining the nuanced classification of quantum fuzzy sets with a self-evolving structure, it provides a scalable, low-maintenance tracking system. Future research will explore rule pruning to handle drastic layout changes and optimize the architecture for larger-scale multi-class environments.

Summary updated

Abstract

Radio Frequency Identification technology has gained popularity for cheap and easy deployment. In the realm of manufacturing shopfloor, it can be used to track the location of manufacturing objects to achieve better efficiency. The underlying challenge of localization lies in the non-stationary characteristics of manufacturing shopfloor which calls for an adaptive life-long learning strategy in order to arrive at accurate localization results. This paper presents an evolving model based on a novel evolving intelligent system, namely evolving Type-2 Quantum Fuzzy Neural Network (eT2QFNN), which features an interval type-2 quantum fuzzy set with uncertain jump positions. The quantum fuzzy set possesses a graded membership degree which enables better identification of overlaps between classes. The eT2QFNN works fully in the evolving mode where all parameters including the number of rules are automatically adjusted and generated on the fly. The parameter adjustment scenario relies on decoupled extended Kalman filter method. Our numerical study shows that eT2QFNN is able to deliver comparable accuracy compared to state-of-the-art algorithms.
BibTeX
@misc{ashfahani2018online,
  author = {Andri Ashfahani and Mahardhika Pratama and Edwin Lughofer and Qing Cai and Huang Sheng},
  title = {{An Online RFID Localization in the Manufacturing Shopfloor}},
  year = {2018},
  eprint = {1805.07715},
  archivePrefix = {arXiv},
  url = {https://arxiv.org/abs/1805.07715},
}
Figure