Special Session:
Safety integration in neural networks for sensor systems in critical applications (NNiS)

The increasing use of sensors in many products and applications is intended to improve safety. Nowadays, sensors in cars monitoring driver attention and meteorological sensors preventing weather-related safety risks have become almost standard features. In industrial plants, predictive maintenance and medical technology, sensors are essential for monitoring and controlling the safety of processes. Due to advances in GPU efficiency and an increasing amount of data, neural networks are becoming increasingly utilized in sensor applications in critical environments.

This session introduces a novel perspective by addressing the regulatory and safety integration of neural networks within sensor network environments. While most research focuses on model accuracy or efficiency, few approaches explicitly consider how neural methods can be designed, validated, and deployed in compliance with emerging regulations and safety standards in critical real-time systems. This session aims to close this gap by highlighting compliant neural networks design principles, certification-aware learning strategies, and trustworthy deployment frameworks for intelligent sensor applications.

This session is intended for scientists, engineers and practitioners, who present their latest theoretical and practical results, as well as the challenges they have encountered in the field of sensor technology using neural networks and hybrid methods.

The following is a non-exclusive list of topics:

1. Utilizing sensors for industrial monitoring and control

2. Sensor-based safety and reliability systems in critical environments

3. Smart city and smart home applications combining sensor systems with neural networks

4. Hybrid modeling approaches combining physics-based and data-driven methods

5. Multimodal sensor fusion (e.g., cameras, acoustic, vibration, and environmental data) for automated driving or process monitoring

6. Neural networks for real-time anomaly detection and predictive maintenance in sensor systems with safety constraints

7. Certification-aware model design and validation under the EU AI Act and other regulatory frameworks

8. Trustworthy and explainable AI for sensor applications

9. Methods to enhance data quality in sensor systems (e.g., drift compensation, denoising, data harmonization) and their meaning for regulations

10.  Cyber-physical and medical systems integrating compliant neural networks for monitoring and decision support

11.  Standards and benchmarking for the safe deployment of neural networks in sensor networks

12.  Edge AI and energy-efficient neural inference for distributed sensor systems

Another novelty of the session lies in its interactive concept: an Open Discussion Slot (ODS) designed to openly discuss unpublished or sensitive real-world challenges in sensor applications will be integrated. Several issues, such as trustworthiness of models, data safety, data quality, data migration, or explainability with large amounts of data, cannot or may not be published. These topics will be openly addressed in the roundtable discussion to facilitate an exchange of possible solutions. The organizers and invited experts will provide examples to stimulate discussion; further impetus is welcome.  The session will feature five 15-minute paper presentations and a 45-minute open discussion.

The session directly addresses core research areas of IJCNN, FUZZ-IEEE, and IEEE CEC by focusing on neural and hybrid computational intelligence methods for sensor-based systems. It promotes the combination of (deep) neural networks with sensor systems to improve sensing, control, and reliability. Furthermore, the integration of trustworthy AI and compliance aspects links theoretical advances in intelligent systems to their responsible real-world deployment—an increasingly central topic across all three conferences.

The organizers aim to bridge the gap between academic research and industrial application in intelligent sensing. By bringing together experts from machine learning, control engineering, and regulatory sciences, this session fosters interdisciplinary collaboration toward safer and more reliable AI-driven sensor technologies. The session is expected to generate shared insights and potential best practices for the design, certification, and maintenance of compliant deep learning systems, contributing to both scientific progress and sustainable innovation in industry and society.

We expect high-quality contributions. For further details, see

https://attend.ieee.org/wcci-2026/information-for-authors/

Paper submission deadline: 31 January: (23:59, anywhere on Earth, i.e. UTC-12) 

for details see https://attend.ieee.org/wcci-2026/important-dates-deadlines/