SPECIAL SESSION #08

Martin Atzmueller

Osnabrück University & German Research Center for Artificial Intelligence, Germany

• martin.atzmueller@uos.de

Juan Carlos Saborío

Osnabrück University, Germany

• jcsaborio@uos.de

Trustworthiness, as a quality of AI systems, is difficult to assess but often associated with indicators such as transparency and reliability. Guaranteeing these properties is especially challenging in domains such as agriculture, characterized by incomplete information, imperfect predictability and partial control. The complex underlying dynamics give rise to data with no a priori structure and non-trivial relationships, which demands a tight interplay between method robustness and data representation techniques.

This session captures the latest methodological developments, advances and real‑world applications for building trustworthy AI systems that operate on complex agricultural data. Contributions are invited that address techniques for ensuring or improving system reliability, transparency, explainability, interpretability and robustness in the context of complex data such as images, hyperspectral images, point clouds, graphs, networks, and especially multimodal data from autonomous systems in agriculture, remote sensing imagery and IoT sensor streams. Particular emphasis is placed on methods for robust data interpretation, semantic modeling, explainable AI approaches, uncertainty representation and quantification, and model interpretability tailored to the spatiotemporal dynamics of modern agriculture. In the context of agriculture, the session aims to provide an interdisciplinary forum to investigate fundamental issues that hinder trustworthiness in AI and the efficient handling of complex data, as well as to discuss recent advances, trends, and challenges in various related topics.

We welcome contributions that cover the following topics:

• Explainable AI and interpretable methods on complex spatio‑temporal agricultural data
• Methods for robust data processing and analysis in agriculture
• Reliable machine learning approaches and methods in agriculture
• Knowledge-grounded approaches for trustworthy AI in agriculture
• Semantic modeling methods for trustworthy AI in agriculture
• Uncertainty quantification in the context of trustworthy AI in agriculture
• Resilient and robust decision-making under uncertainty and with noisy sensor data
• Transparent integration of perception, learning and planning approaches for agricultural tasks
• Edge‑ and embedded AI enabling robustness and real‑time reliability
• Validation, benchmarking, and reproducibility for trustworthy AI in agriculture
• Trustworthy AI system design in complex agricultural data domains
• Applications of all of the above

Martin Atzmueller is Full Professor at the Institute of Computer Science at Osnabrück University (Germany), as well as Scientific Director - Research Department Cooperative and Autonomous Systems in Osnabrück - at the German Research Center for Artificial Intelligence (DFKI). He is founding spokesperson of the Joint Lab on Artificial Intelligence and Data Science and founding member of the Research Center Data Science at Osnabrück University. His research areas are artificial intelligence (AI), data science and integrative AI systems, where his particular research interests include knowledge-grounded semantic machine perception, explainable AI, interpretable learning, as well as integrative and trusted AI system design approaches.

Dr. Juan Carlos Saborío is a Postdoctoral researcher at the Joint Lab for Artificial Intelligence & Data Science of the University of Osnabrück, Germany, where he completed his doctorate in Computer Science on the topic of stochastic models for planning under uncertainty. Previously, he was a researcher at the Cooperative Autonomous Systems (formerly Plan-Based Robot Control) group of the DFKI Niedersachsen, where he worked on the integration of probabilistic planning and intention recognition methods onboard robotic assistants. His current interests include approximation and optimization methods for POMDPs, integrated and hybrid AI, as well as safe and explainable AI applications in industry and agriculture.