DISTRIBUTED ANOMALY DETECTION SYSTEM IN MANUFACTURING USING EDGE AI AND COMPUTER VISION

Olha Shymchyshyn, Kostiantyn Pidhornyi, Maryan Shymchyshyn

Abstract


Background. Modern automated production facilities are characterized by a high level of robotization and the drive to minimize human involvement in routine operations. At the same time, the complete elimination of the operator is unacceptable due to safety require­ments and the need for decision-making in the event of anomalies. The use of centralized computer vision systems results in significant network load, limiting their scalability. One promising direction for solving this problem is the application of Edge AI technology.

Materials and Methods. A three-tier architecture for an anomaly detection system is proposed. The first tier is implemented based on the ESP32-P4 edge nodes with digital cameras, which perform local object detection using quantized computer vision models. Only structured metadata is transmitted to the central cluster via the CAN bus. The second tier performs motion trajectory analysis based on root mean square (RMS) deviation. The third tier implements an operator notification subsystem with automatic generation of an event video replay.

Results and Discussion. Offloading object detection to edge nodes significantly reduces network load and ensures system scalability regardless of the number of cameras. The use of the ESP32-P4 enables real-time video processing at up to 30 frames/s with a power consumption of less than 500 mW per node. RMS analysis ensures the detection of deviations from the reference trajectory in less than 1 s. After an anomaly is recorded, the operator receives a push notification and a 5-second video clip for rapid situation assessment.

Conclusion. The proposed architecture demonstrates the feasibility of practical Edge AI utilization for building scalable production process monitoring systems. The combination of local computer vision, transmission of structured data via the CAN bus, and centralized statistical analysis ensures continuous technological process control, reduces network load, and shortens response time to anomalous events. Further research will focus on the adaptive tuning of RMS thresholds and the implementation of federated learning for the models.

Keywords: edge computing, computer vision, ESP32-P4, anomaly detection, CAN bus, industrial automation


Full Text:

PDF

References


[1] Romero, D., & Stahre, J. (2021). Towards the resilient operator 5.0: The future of work in smart resilient manufacturing systems. Procedia CIRP, 104, 1089–1094. https://doi.org/10.1016/j.procir.2021.11.183

[2] Hasan, M. M., Kasedullah, M., Ripon, M. B. B., & Khan, M. M. H. (2025). AI-Driven Quality Control in Manufacturing and Construction: Enhancing Precision and Reducing Human Error. Applied IT & Engineering, 3(1), 1–10.
https://doi.org/10.25163/engineering.3110270

[3] Seeber, I., Bittner, E., Briggs, R. O., de Vreede, T., de Vreede, G. J., Elkins, A., Maier, R., Merz, A. B., Oeste-Reiß, S., Randrup, N., Schwabe, G., & Söllner, M. (2020). Machines as teammates: A research agenda on AI in team collaboration. Information & Management, 57(2), 103174.
https://doi.org/10.1016/j.im.2019.103174

[4] Deng, S., Zhao, H., Fang, W., Yin, J., Dustdar, S., & Zomaya, A. Y. (2020). Edge intelligence: The confluence of edge computing and artificial intelligence. IEEE Internet of Things Journal, 7(8), 7457–7469. https://doi.org/10.1109/JIOT.2020.2984887

[5] Espressif Systems. (2025). ESP32-P4 technical reference manual (Rev. v0.4, Pre-release). Espressif Systems (Shanghai) Co., Ltd. https://documentation.espressif.com/esp32-p4-chip-revision-v1.3 technical_reference_manual_en.pdf

[6] Espressif Systems. (2025). ESP-NN: Optimised neural network functions for Espressif chipsets [Software library]. GitHub.
https://github.com/espressif/esp-nn

[7] Espressif Systems. (2025). ESP-DL: Espressif deep-learning library for AIoT applications (v3.3.8) [Software framework]. ESP Component Registry. https://components.espressif.com/components/espressif/esp-dl

[8] Lazarevich, I., Kozlov, A., Kaglinskiy, M., & Dukhan, M. (2023). YOLOBench: Benchmarking efficient object detectors on embedded systems. arXiv. https://arxiv.org/abs/2307.13901

[9] International Organization for Standardization. (2024). Road vehicles — Controller area network (CAN) — Part 1: Data link layer and physical coding sublayer (ISO Standard No. 11898-1:2024).
https://www.iso.org/standard/86384.html

[10] Montgomery, D. C. (2019). Introduction to statistical quality control (8th ed.). John Wiley & Sons.

[11] Masykuroh, K., Hendrawan, Mulyana, E., & Krishna, F. (2025). A Survey on Video Compression Optimization Techniques for Accuracy Enhancement in Video Analytics Applications (VAPs). IEEE Access, V.13, 75822–75846. https://doi.org/10.1109/ACCESS.2025.3561063

[12] Peluso, V., Cipolletta, A., Calimera, A., Poggi, M., Tosi, F., & Mattoccia, S. (2025). Edge AI in practice: A survey and deployment framework for neural networks on embedded systems. Electronics, 14(24), 4877. https://doi.org/10.3390/electronics14244877

[13] Gauttam, H., Nain, G., Pattanaik, K. K., & Mendes, P. (2025). Edge-AI: A systematic review on architectures, applications, and challenges. Journal of Network and Computer Applications, 104375.
https://doi.org/10.1016/j.jnca.2025.104375




DOI: http://dx.doi.org/10.30970/eli.35.9

Refbacks

  • There are currently no refbacks.