arXiv cs.AI· Mohamed Bachar (LINEACT), Ilyass Abouelaziz (LINEACT), Yuehua Ding (LINEACT)·· 10 小时前AI 评分18
连续时间 Set-Transformer 实现异步 WiFi-IMU 室内定位
Continuous-Time Set-Transformers for Asynchronous WiFi-IMU Indoor Localization
AI 导读
研究者提出一种连续时间 Transformer，将 WiFi RSSI 与 IMU 观测表示为带时间戳的无序 token 集合，用单个注意力模块联合融合跨模态与跨时间信息，无需重采样和模态专用分支。训练时随机遮蔽整个模态和单条观测，使模型可在输入缺失或过期时工作。在跨会话漂移场景下优于开放 WiFi 指纹基线，与学习融合及航位推算基线相当，并随观测缺失或过期平缓退化。
正文 · 原文
Abstract:Indoor localization commonly combines WiFi received signal strength indicator (RSSI) fingerprints with inertial measurements. However, these sensor streams are sampled asynchronously at different rates; therefore, observations rarely align in time and may be missing or stale. Most existing methods resample the data onto a shared clock and process each modality using a separate branch. We instead present a continuous-time transformer that represents observations as an unordered set of timestamped tokens. Each token encodes its measurement, modality, and elapsed time since acquisition. A single attention block jointly fuses information across modalities and time, avoiding both resampling and modality-specific branches while remaining invariant to token order. During training, we randomly mask entire modalities and individual observations, enabling the model to operate under missing or outdated inputs. We validate the modality-specific encoders on their respective benchmarks and evaluate the fused model on both simulated and real datasets, including a crosssession dataset collected on unseen days. The proposed approach improves upon an open WiFi fingerprinting baseline under cross-session drift, remains competitive with learned fusion and dead-reckoning baselines, and degrades gracefully as observations become missing or stale. These results suggest that a single continuoustime transformer provides a practical framework for indoor localization from asynchronous sensors.
| Subjects: | Machine Learning (cs.LG); Artificial Intelligence (cs.AI) |
| Cite as: | arXiv:2611.13057 [cs.LG] |
| (or arXiv:2611.13057v1 [cs.LG] for this version) | |
| https://doi.org/10.48550/arXiv.2611.13057 arXiv-issued DOI via DataCite (pending registration) |
|
| Journal reference: | 23rd International Conference on Informatics in Control, Automation and Robotics, Oct 2026, Angers, France. pp.331-342 |
Submission history
From: Mohamed BACHAR [view email] [via CCSD proxy]
[v1]
Mon, 9 Nov 2026 15:54:11 UTC (1,606 KB)
来源：arXiv cs.AI · arxiv.org