SIF-TF: A Scene-Interaction fusion Transformer for trajectory prediction

Gao, Fei; Huang, Wanjun; Weng, Libo; Zhang, Yuanming

doi:10.1016/j.knosys.2024.111744

Cited by 1 publication

(1 citation statement)

References 11 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…By leveraging the weight allocation of GRU and the attention mechanism, the model can utilize the latitude, longitude, tidal value, water velocity, and wind speed as the inputs of the model feature, thereby accurately predicting the short-term drift of the target. In recent years, the Transformer has made significant achievements in various fields [24][25][26][27][28][29][30], largely because the attention mechanism plays a key role in the Transformer. The extraction of local and dimensional features is paramount for sequence prediction, wherein features of varying dimensions can complement one another.…”

Section: Related Workmentioning

confidence: 99%

Prediction of Drift Trajectory in the Ocean Using Double-Branch Adaptive Span Attention

Zhang,

Zhao

et al. 2024

JMSE

View full text Add to dashboard Cite

The accurate prediction of drift trajectories holds paramount significance for disaster response and navigational safety. The future positions of underwater drifters in the ocean are closely related to their historical drift patterns. Additionally, leveraging the complex dependencies between drift trajectories and ocean currents can enhance the accuracy of predictions. Building upon this foundation, we propose a Transformer model based on double-branch adaptive span attention (DBASformer), aimed at capturing the multivariate time-series relationships within drift history data and predicting drift trajectories in future periods. DBASformer can predict drift trajectories more accurately. The proposed adaptive span attention mechanism exhibits enhanced flexibility in the computation of attention weights, and the double-branch attention structure can capture the cross-time and cross-dimension dependencies in the sequences. Finally, our method was evaluated using datasets containing buoy data with ocean current velocities and Autonomous Underwater Vehicle (AUV) data. The raw data underwent cleaning and alignment processes. Comparative results with five alternative methods demonstrate that DBASformer improves prediction accuracy.

show abstract