CroTad: A Contrastive Reinforcement Learning Framework for Online Trajectory Anomaly Detection

2025-11-20

AI Summary

Sub-trajectory-level anomaly detection in intelligent transportation systems remains challenging due to limited fine-grained localization capability, strong dependence on manually tuned thresholds, and poor robustness to irregular sampling and trajectory noise. Method: This paper proposes a threshold-free, online detection framework that innovatively integrates contrastive learning with deep reinforcement learning: contrastive learning adaptively models diverse normal mobility patterns and extracts trip-specific features, while deep reinforcement learning enables end-to-end real-time point-level and sub-trajectory-level anomaly scoring. Contribution/Results: The framework natively handles irregularly sampled and noisy trajectories without requiring predefined thresholds. Evaluated on two real-world datasets, it significantly outperforms state-of-the-art methods in detection accuracy, robustness to noise and sampling irregularities, and computational efficiency—demonstrating both theoretical effectiveness and practical deployability for real-time traffic monitoring.

Technology Category

• Machine Learning: Life-Long and Continual Learning
• Data Mining & Knowledge Management: Anomaly/Outlier Detection
• Planning, Routing, and Scheduling: Replanning and Plan Repair

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Systems and Infrastructure for Web, Mobile and WoT: Applied ML and AI for Web-based mobile applications
• Graph Algorithms and Modeling for the Web: Representation, reconstruction, and subgraph or motif discovery in Web-related graphs

Abstract

Detecting trajectory anomalies is a vital task in modern Intelligent Transportation Systems (ITS), enabling the identification of unsafe, inefficient, or irregular travel behaviours. While deep learning has emerged as the dominant approach, several key challenges remain unresolved. First, sub-trajectory anomaly detection, capable of pinpointing the precise segments where anomalies occur, remains underexplored compared to whole-trajectory analysis. Second, many existing methods depend on carefully tuned thresholds, limiting their adaptability in real-world applications. Moreover, the irregular sampling of trajectory data and the presence of noise in training sets further degrade model performance, making it difficult to learn reliable representations of normal routes. To address these challenges, we propose a contrastive reinforcement learning framework for online trajectory anomaly detection, CroTad. Our method is threshold-free and robust to noisy, irregularly sampled data. By incorporating contrastive learning, CroTad learns to extract diverse normal travel patterns for different itineraries and effectively distinguish anomalous behaviours at both sub-trajectory and point levels. The detection module leverages deep reinforcement learning to perform online, real-time anomaly scoring, enabling timely and fine-grained identification of abnormal segments. Extensive experiments on two real-world datasets demonstrate the effectiveness and robustness of our framework across various evaluation scenarios.

Problem

• Detecting sub-trajectory anomalies in intelligent transportation systems
• Eliminating dependency on manually tuned detection thresholds
• Handling irregular sampling and noise in trajectory data

Innovation

• Contrastive reinforcement learning for trajectory anomaly detection
• Threshold-free framework robust to noisy irregular data
• Online real-time anomaly scoring using deep reinforcement learning