trajectory pattern analysis

Existing trajectory clustering methods are highly sensitive to transient anomalies, often fragmenting trajectories due to local perturbations—thereby compromising cluster integrity, stability, and interpretability. To address this, we propose a robust trajectory clustering framework built upon segment-based DBSCAN, supporting both full-trajectory and sliding-window subtrajectory modeling. We introduce the Mean Absolute Deviation (MAD) to quantitatively measure local perturbation intensity, enabling selective suppression of transient deviations. Furthermore, we design a motion-history-driven, event-triggered segment splitting and merging mechanism to mitigate over-segmentation induced by anomalies. Experiments on real-world trajectory datasets demonstrate that our method significantly improves clustering stability, parameter robustness, and pattern consistency, outperforming state-of-the-art trajectory clustering algorithms.

Existing trajectory anomaly detection and cleaning tools lack systematic, standardized evaluation, hindering fair comparison and practical deployment. Method: We propose a unified taxonomy categorizing methods into five classes—statistical, sliding-window, clustering, graph-based, and heuristic—and introduce a reproducible ground-truth generation mechanism tailored to real-world trajectory scenarios. This forms the first standardized evaluation framework for trajectory anomaly detection. Contribution/Results: We conduct comprehensive efficiency and accuracy benchmarking across ten mainstream open-source tools on diverse real-world trajectory datasets, analyzing performance degradation under distinct anomaly types (e.g., positional drift, sampling noise, semantic inconsistency). Our empirical study yields an evidence-based tool selection guideline, significantly enhancing comparability and practicality of trajectory preprocessing methods. The framework establishes a foundational benchmark for both academic research and industrial applications in trajectory data quality management.

To address the lack of standardized datasets, inconsistent preprocessing protocols, and non-uniform evaluation metrics in UAV trajectory prediction research, this paper proposes the first comprehensive standardization framework for the field. We introduce an integrated pipeline encompassing data cleaning, coordinate normalization, multi-granularity evaluation (ADE, FDE, and collision detection), and interactive visualization. We publicly release Dronalize—a Python-based end-to-end toolbox built on NumPy, Pandas, Matplotlib, and Plotly—that supports seamless adaptation to major benchmarks including UAV123 and DroneVehicle, and incorporates customizable modules such as physics-aware collision detection. Experiments demonstrate a 70% average reduction in preprocessing time; consistent and comparable evaluation results across six state-of-the-art models; and broad adoption, evidenced by over 320 GitHub stars and widespread use in academia.

High-dimensional features in trajectory analysis impair computational efficiency and model interpretability, limiting predictive performance. To address this, we propose a structured feature selection method grounded in a geometric–kinematic dual-classification framework: trajectory features—including curvature, convexity, velocity, and acceleration—are semantically partitioned into geometric and kinematic categories, and group-wise constraints are imposed to drastically reduce the combinatorial search space. Unlike black-box optimization approaches, our method ensures both computational efficiency and decision transparency. Experiments across multiple public trajectory datasets demonstrate that the proposed approach maintains or improves prediction accuracy while reducing feature selection time by 42% on average. Moreover, it reveals dataset-specific sensitivities to geometric versus kinematic features, offering an interpretable and reusable feature engineering paradigm for trajectory modeling.

This paper addresses the problem of compactly representing large-scale trajectory data (e.g., GPS traces): given *n* input trajectories, select the minimum number of representative polygonal curves—each of complexity at most *l*—such that every point on any input trajectory lies within Fréchet distance Δ of some subtrajectory of a representative curve. We propose a novel geometric set cover framework that, for the first time, supports multi-segment polyline representatives (not merely line segments), reducing the required number of representatives from *O(kl log(kl))* to *O(k log n)*, where *k* is the optimal cover size. Our algorithm guarantees coverage radius 11Δ and runs in Õ(*l²n⁴ + kln⁴*) time. Extensive evaluation on ocean current and human motion datasets demonstrates significant improvements in modeling complex real-world movement patterns and practical applicability.

This study addresses the challenge of clustering individual mobility trajectories derived from mobile phone signaling data by proposing a novel framework that integrates compositional data analysis with state-space modeling. The approach represents trajectories as temporal sequences of compositional vectors in the simplex space, explicitly incorporating both localization uncertainty and road network structure. By leveraging a mixture state-space model, the method enables interpretable trajectory clustering. As the first work to combine compositional data analysis with temporal state-space models for human mobility analysis, this research successfully identifies semantically meaningful group travel patterns in an empirical study of Padua, offering urban planners highly interpretable, data-driven insights for transportation decision-making.

This study addresses the structural distortions introduced when linearizing two-dimensional geospatial data into one-dimensional orderings, which often produce misleading visual artifacts that obscure genuine spatiotemporal patterns. To mitigate this issue, the authors propose a metric-driven visual analytics approach that uniquely integrates neighborhood-preserving metrics with visual enhancement techniques—specifically glyphs, halos, and stippling—to interactively and interpretable identify and annotate linearization artifacts through a dedicated interface. By coupling quantitative fidelity measures with perceptually effective visual encodings, the method significantly enhances analysts’ ability to distinguish authentic spatial structures from distortion-induced artifacts. The efficacy of the proposed framework is demonstrated through a case study on COVID-19 incidence rates in Germany, where it successfully supports accurate pattern recognition amidst complex spatial data.

This work addresses the high cost and limited scalability of traditional turning movement count methods, which rely on manual annotation and camera calibration. The authors propose an unsupervised pipeline that automatically generates persistent polygonal entry and exit zones solely from vehicle trajectory start and end points, leveraging spatial clustering without requiring human intervention, camera calibration, or prior intersection knowledge. This approach introduces reusable spatial regions for the first time, enabling cross-camera deployment and substantially reducing computational overhead. Integrated with object detection, multi-object tracking, and systematic parameter optimization, the method achieves efficient trajectory classification. Evaluated on 25 real-world cameras—including 16 unseen scenarios—it attains a median classification error of approximately 3% and meets engineering standards under the GEH metric, demonstrating superior stability and efficiency compared to existing baselines.

This study addresses the challenges of low matching accuracy and poor computational efficiency in map-matching low-frequency GPS trajectories within dense road networks. To this end, the authors propose an enhanced spatiotemporal trajectory matching method that integrates a dynamic buffer mechanism, an adaptive observation probability model, an improved temporal scoring function, and a path inference strategy based on historical behavioral patterns. Notably, the approach achieves high-quality matching without requiring ground-truth annotations. Experimental evaluation on real-world trajectory data from Milan demonstrates the superiority of the proposed method, with significant improvements across multiple metrics compared to existing algorithms, while simultaneously maintaining high computational efficiency and accurate path reconstruction.

This study addresses the challenge of systematically evaluating stay-point detection algorithms under noisy trajectory conditions, which has been hindered by the absence of publicly available benchmark datasets with ground-truth annotations. To bridge this gap, we construct the first large-scale synthetic trajectory dataset annotated with real-world stay points and propose both a novel unsupervised and a supervised detection algorithm. Leveraging this dataset alongside a noise-robustness evaluation framework, we conduct a comprehensive assessment of nine representative algorithms. Our results reveal that state-of-the-art methods suffer significant performance degradation under realistic noise, whereas the proposed approaches substantially improve detection accuracy, with the supervised variant notably outperforming existing baselines.