Document Type
Lecture
Abstract
In the age of embodied AI and smart automation, autonomous agents are increasingly deployed in high-stakes, real-world environments. Ensuring the robustness and resilience of these systems in the face of rare but critical failure events is essential for their safe and reliable operation. Accurate forecasting of such rare events is particularly crucial, as a single overlooked anomaly can lead to catastrophic consequences. In manufacturing, for instance, unplanned downtime due to rare failures costs industries over \$50 billion annually, with sectors like automotive losing more than \$2 million per hour—even with preventive maintenance systems in place.
However, the extreme rarity and complexity of these events pose significant challenges for AI methods. The scarcity of high-quality data, methodological gaps in the literature, and limited practical experience with multimodal signals further complicate rare event prediction and detection.
In this talk, I explore two real-world domains critical for autonomous systems. First, I focus on smart manufacturing/ Industry 4.0, presenting the complete lifecycle of rare event prediction -- ranging from analog and multimodal dataset development to addressing data scarcity, improving data quality, and building and evaluating robust predictive models.
In the second part, I introduce the problem of unobserved entity prediction in autonomous driving, which involves identifying potentially missing entities due to perceptual failures in edge-case scenarios. I discuss how contextual and structured knowledge can be leveraged to develop a knowledge-enhanced predictive framework, validated across multiple real-world driving datasets.
I will share the datasets and resources we have developed and made publicly available for both cases to support and encourage further research in rare event prediction for autonomous systems.
Publication Info
Published in 14th Conference on Extreme Value Analysis, Probabilistic and Statistical Models and their Applications (EVA 2025), Summer 2025.
APA Citation
Wickramarachchi, R. (2025, June 25). From assembly lines to the open road: Predicting rare events in autonomous systems. 14th Conference on Extreme Value Analysis, Probabilistic and Statistical Models and Their Applications (EVA 2025), University of North Carolina at Chapel Hill, Chapel Hill, NC, USA.