Researchers have developed NAViLoss, a novel objective function designed to improve the accuracy and robustness of underwater vehicle velocity estimation. This method addresses limitations in current learning-based approaches by incorporating physics-consistent and uncertainty-aware principles. NAViLoss jointly penalizes estimation errors in both the navigation-state and DVL measurement domains, while an adaptive mechanism regulates measurement uncertainty. When integrated into a DeepONet architecture, the resulting NAVi-DeepONet model achieved a 44% improvement in velocity estimation accuracy compared to existing methods, as demonstrated on extensive real-world AUV experimental data. AI
IMPACT This research could lead to more accurate and reliable navigation for autonomous underwater vehicles.
RANK_REASON The cluster describes a new objective function and model presented in an arXiv paper. [lever_c_demoted from research: ic=1 ai=1.0]
AI-generated summary · Google Gemini · from 2 sources. How we write summaries →
