# Mitigating Error Accumulation in Continuous Navigation via Memory-Augmented Kalman Filtering Source: [Mitigating Error Accumulation in Continuous Navigation via Memory-Augmented Kalman Filtering]() Collection index: [llms.txt]() Data updated: 2026-05-09T05:00:04.795Z Authors: Yin Tang, Jiawei Ma, Jinrui Zhang, Alex Jinpeng Wang, Deyu Zhang Published: 2026-01-30T05:03:08.000Z ## Abstract Continuous navigation in complex environments is critical for Unmanned Aerial Vehicle (UAV). However, the existing Vision-Language Navigation (VLN) models follow the dead-reckoning, which iteratively updates its position for the next waypoint prediction, and subsequently construct the complete trajectory. Then, such stepwise manner will inevitably lead to accumulated errors of position over time, resulting in misalignment between internal belief and objective coordinates, which is known as "state drift" and ultimately compromises the full trajectory prediction. Drawing inspiration from classical control theory, we propose to correct for errors by formulating such sequential prediction as a recursive Bayesian state estimation problem. In this paper, we design NeuroKalman, a novel framework that decouples navigation into two complementary processes: a Prior Prediction, based on motion dynamics and a Likelihood Correction, from historical observation. We first mathematically associate Kernel Density Estimation of the measurement likelihood with the attention-based retrieval mechanism, which then allows the system to rectify the latent representation using retrieved historical anchors without gradient updates. Comprehensive experiments on TravelUAV benchmark demonstrate that, with only 10% of the training data fine-tuning, our method clearly outperforms strong baselines and regulates drift accumulation. ## Recorded benchmark results A research result applies to its stated dataset, split, metric and evaluation configuration. Compare like-for-like settings, preserve the metric's unit and direction, and distinguish paper-reported results from independent reproduction. The best result recorded here is not a claim of exhaustive, current global coverage. - [Continuous Navigation on UAV-Need-Help Full (test-seen)](): Continuous Navigation; UAV-Need-Help Full (test-seen); Navigation Error (NE): 71.56. - [Continuous Navigation on UAV-Need-Help Easy (test-seen)](): Continuous Navigation; UAV-Need-Help Easy (test-seen); Navigation Error (NE): 42.7. - [Continuous Navigation on UAV-Need-Help Hard (test-seen)](): Continuous Navigation; UAV-Need-Help Hard (test-seen); Navigation Error (NE): 105.07. - [Continuous Navigation on UAV-Need-Help UO, Easy L1 assistant (test-unseen)](): Continuous Navigation; UAV-Need-Help UO, Easy L1 assistant (test-unseen); Navigation Error (NE): 44.5. - [Continuous Navigation on UAV-Need-Help UM, Full L1 assistant (test-unseen)](): Continuous Navigation; UAV-Need-Help UM, Full L1 assistant (test-unseen); Navigation Error: 100.32. - [Continuous Navigation on UAV-Need-Help UM, Easy L1 assistant (test-unseen)](): Continuous Navigation; UAV-Need-Help UM, Easy L1 assistant (test-unseen); Navigation Error: 69.5. - [Continuous Navigation on UAV-Need-Help UM Hard L1 assistant (test-unseen)](): Continuous Navigation; UAV-Need-Help UM Hard L1 assistant (test-unseen); Navigation Efficiency (NE): 140. - [Continuous Navigation on UAV-Need-Help UO, Full L1 assistant (test-unseen)](): Continuous Navigation; UAV-Need-Help UO, Full L1 assistant (test-unseen); NE: 71.01. - [Continuous Navigation on UAV-Need-Help UO Hard L1 assistant (test-unseen)](): Continuous Navigation; UAV-Need-Help UO Hard L1 assistant (test-unseen); NE: 84.5. ## Original sources and implementation [Original paper]() [PDF]()