Time Series Domain Adaptation: A Review

ABSTRACT Time‐series domain adaptation (TSDA) has emerged as a critical paradigm for addressing distribution shifts across sensing environments, operating conditions, and acquisition devices in sequential data. Building upon the theoretical foundations of vision domain adaptation, TSDA introduces additional challenges stemming from temporal dependency, spectral variability, and privacy constraints inherent to real‐world systems. This survey provides a comprehensive and systematic review of TSDA methods from the perspectives of access‐privacy constraints, category‐space semantics, and source‐target topology, three orthogonal axes that unify existing approaches within a common taxonomy. Specifically, we analyze representative algorithms across four access regimes (UDA, SFUDA, BBDA, FDA), four category‐space configurations (closed, partial, open, universal), and three topological structures (single‐, multi‐, and multi‐target), highlighting their theoretical motivations, architectural trends, and cross‐domain generalization strategies. We further examine real‐world applications across diverse temporal domains, linking methodological evolution to practical deployment constraints. Benchmarking results from recent studies are summarized to establish empirical baselines across representative time‐series datasets and applied domains, encompassing both core classification benchmarks and real‐world TSDA applications. Finally, we discuss open challenges and future directions in this field.