EUVIMEDEUVIMED entdecken
Open Access · CC BY-NC-SA 4.0

Conflict-Weighted Meta-Gradient Correction for Continual Learning in Multi-Sensor UAV Positioning

Truong Nguyen Xuan, Van Tran Thi, Thang Nguyen Van, Quang-Anh N.D, Khoa Nguyen Dang

EAI Endorsed Transactions on AI and Robotics · 06.11.2026 · 5

DOI: 10.4108/airo.13236 ↗

Kurz gefasst

Fachartikel in der Zeitschrift EAI Endorsed Transactions on AI and Robotics, online erschienen am 06.11.2026. Verfasst von Truong Nguyen Xuan und 4 weiteren Autor:innen. Der Volltext ist frei zugänglich (Open Access, CC BY-NC-SA 4.0).

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Multi-sensor UAV positioning systems are increasingly deployed in sequences where new sensor modalities are integrated mid-deployment, requiring a model to expand its input space while retaining knowledge of previously observed environments. Existing continual learning methods assume a fixed-dimensional input and fail in this modality-incremental settings with earlier single-modality task memories produce stale gradients that actively oppose learning of newly introduced sensor features. We propose RAMGCL, a method that addresses this dual challenge of catastrophic forgetting and input-space expansion through three components operating within a fixed-size architecture. An asymmetric modality filter selectively suppresses stale gradients from zero-padded memories while retaining cross-modality signal that remains informative. A conflict-weighted per-task memory gradient continuously corrects the update direction by upweighting tasks whose gradients most conflict with the current update. A bi-level diagonal gradient preconditioner meta-learns a per-dimension scaling vector for the latent gradient slice, calibrating stability and plasticity without solving a quadratic programme. Episodic memories are populated using farthest-point sampling in position space to ensure spatially diverse coverage of each environment. Experiments on the Q-Drone UWB benchmark across five environments and two modality-incremental orderings show that RAMGCL achieves state-of-the-art performance in both average MAE and forgetting, and is the only method to achieve positive backward transfer in the IMU-to-UWB ordering.

Zitieren

APA

Xuan, T. N., Thi, V. T., Van, T. N., N.D, Q., & Dang, K. N. (2026). Conflict-Weighted Meta-Gradient Correction for Continual Learning in Multi-Sensor UAV Positioning. EAI Endorsed Transactions on AI and Robotics, 5. https://doi.org/10.4108/airo.13236

Vancouver

Xuan TN, Thi VT, Van TN, N.D Q, Dang KN. Conflict-Weighted Meta-Gradient Correction for Continual Learning in Multi-Sensor UAV Positioning. EAI Endorsed Transactions on AI and Robotics. 2026;5. doi:10.4108/airo.13236

BibTeX
@article{Xuan2026,
  author = {Xuan, Truong Nguyen and Thi, Van Tran and Van, Thang Nguyen and N.D, Quang-Anh and Dang, Khoa Nguyen},
  title = {Conflict-Weighted Meta-Gradient Correction for Continual Learning in Multi-Sensor UAV Positioning},
  journal = {EAI Endorsed Transactions on AI and Robotics},
  year = {2026},
  volume = {5},
  doi = {10.4108/airo.13236}
}
Kostenlos starten →