aiqbench analysis
This research introduces a Temporal Convolutional Network capable of reconstructing missing segments in trajectory data by utilizing both past and future context, effectively relaxing standard causality constraints. The approach combines weighted mean squared error with boundary continuity penalties to handle sensor failures or occlusions common in real-world datasets. While the study uses synthetic 2D trajectories for validation, it represents a significant methodological advance in handling incomplete time-series inputs within machine learning frameworks.
Inferring Missing Trajectory Data with Temporal Convolutional Networks
arXiv:2607.25147v1 Announce Type: new Abstract: Trajectory data collected in real-world settings is frequently incomplete due to sensor failure, communication loss, or occlusion. We address the task of \emph{trajectory inpainting}: reconstructing contiguous missing segments from observed context. We propose a Temporal Convolutional Network (TCN) with symmetric dilation that relaxes the standard causality constraint, allowing each time step to draw on both past and future observations, a property that is essential for inpainting, but absent from forecasting-oriented architectures. The model is trained with a composite loss that combines weighted mean squared error, boundary--continuity penalties, and a smoothness regularizer. Trained on a synthetic dataset of $1,000$ (train), $200$ (validation), and $300$ (test) two-dimensional trajectories with randomly placed 20% masked segments, the model achieves good R$^{2}$, MSE and MAE metrics.
Abstract:Trajectory data collected in real-world settings is frequently incomplete due to sensor failure, communication loss, or occlusion. We address the task of \emph{trajectory inpainting}: reconstructing contiguous missing segments from observed context. We propose a Temporal Convolutional Network (TCN) with symmetric dilation that relaxes the standard causality constraint, allowing each time step to draw on both past and future observations, a property that is essential for inpainting, but absent from forecasting-oriented architectures. The model is trained with a composite loss that combines weighted mean squared error, boundary--continuity penalties, and a smoothness regularizer. Trained on a synthetic dataset of $1,000$ (train), $200$ (validation), and $300$ (test) two-dimensional trajectories with randomly placed 20% masked segments, the model achieves good R$^{2}$, MSE and MAE metrics.
Artificial Intelligence (cs.AI)
arXiv:2607.25147 [cs.AI]
(or arXiv:2607.25147v1 [cs.AI] for this version)
https://doi.org/10.48550/arXiv.2607.25147 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.
Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.
Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.
Full text under the publisher's open licence · 215 words
Related coverage
Impute On-Demand: Adaptive Correlated Time Series Imputation for Changing Environments
arXiv · Jul 28, 2026
User Trajectory Prediction Unifying Global and Local Temporal Information
arXiv · Oct 12, 2026
3-D Emissions Mapping and Social Cost Estimation for US Domestic Aviation at West Coast Hubs
arXiv · Sep 30, 2026
[AINews] The Future of Latent Space
Latent Space · Sep 25, 2026