Toward Foundation Models for Online Complex Event Detection in CPS-IoT: A Case Study

2025-03-15

AI Summary

Complex event (CE) detection in Cyber-Physical Systems–Internet of Things (CPS-IoT) faces challenges including modeling long-term temporal dependencies, handling high-noise sensor data, and mitigating interference from irrelevant atomic events. Method: This paper introduces Mamba—a state-space model—as the first foundational architecture for online CE detection in CPS-IoT. We propose a novel neuro-symbolic fusion paradigm that synergistically integrates Mamba’s long-range temporal modeling capability, large language model (LLM)-adaptation mechanisms, data-driven neural rule learning, and joint symbolic reasoning. Contribution/Results: Experiments demonstrate that our approach significantly outperforms mainstream LLMs and purely neural models in both accuracy and cross-scenario generalization. Notably, it achieves substantial gains in detecting CEs from previously unseen, long-duration sensor trajectories—validating Mamba’s effectiveness and superiority as a foundational backbone for CPS-IoT CE detection.

Technology Category

• Cognitive Modeling & Cognitive Systems: Conceptual Inference and Reasoning
• Machine Learning: Neuro-Symbolic Learning
• Intelligent Robots: Multimodal Perception & Sensor Fusion

Application Category

• Systems and Infrastructure for Web, Mobile and WoT: Applied ML and AI for Web-based mobile applications
• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs

Abstract

Complex events (CEs) play a crucial role in CPS-IoT applications, enabling high-level decision-making in domains such as smart monitoring and autonomous systems. However, most existing models focus on short-span perception tasks, lacking the long-term reasoning required for CE detection. CEs consist of sequences of short-time atomic events (AEs) governed by spatiotemporal dependencies. Detecting them is difficult due to long, noisy sensor data and the challenge of filtering out irrelevant AEs while capturing meaningful patterns. This work explores CE detection as a case study for CPS-IoT foundation models capable of long-term reasoning. We evaluate three approaches: (1) leveraging large language models (LLMs), (2) employing various neural architectures that learn CE rules from data, and (3) adopting a neurosymbolic approach that integrates neural models with symbolic engines embedding human knowledge. Our results show that the state-space model, Mamba, which belongs to the second category, outperforms all methods in accuracy and generalization to longer, unseen sensor traces. These findings suggest that state-space models could be a strong backbone for CPS-IoT foundation models for long-span reasoning tasks.

Problem

• Detecting complex events in CPS-IoT systems
• Overcoming challenges of long-term reasoning in noisy data
• Evaluating models for accurate and generalizable event detection

Innovation

• Leveraging large language models for CE detection
• Employing neural architectures to learn CE rules
• Adopting neurosymbolic approach integrating neural and symbolic engines