ReNiL: Event-Driven Pedestrian Bayesian Localization Using IMU for Real-World Applications

TL;DR

ReNiL reframes pedestrian localization from dense trajectory tracking to on-demand IPDP-based estimation, producing displacement and uncertainty at contextually meaningful points. It combines a motion-aware orientation filter with ASLE, a multi-scale, Laplace-parameterized regression module, and a Bayesian inference chain to propagate uncertainty across IPDPs and optionally fuse with external sensors. The approach yields state-of-the-art displacement accuracy and homogeneous Euclidean uncertainty on RoNIN-ds and a new WUDataset, while reducing computation. This event-driven, uncertainty-aware framework has strong practical potential for mobile and IoT localization, enabling robust, infrastructure-free positioning across diverse motion patterns and devices.

Abstract

Pedestrian inertial localization is key for mobile and IoT services because it provides infrastructure-free positioning. Yet most learning-based methods depend on fixed sliding-window integration, struggle to adapt to diverse motion scales and cadences, and yield inconsistent uncertainty, limiting real-world use. We present ReNiL, a Bayesian deep-learning framework for accurate, efficient, and uncertainty-aware pedestrian localization. ReNiL introduces Inertial Positioning Demand Points (IPDPs) to estimate motion at contextually meaningful waypoints instead of dense tracking, and supports inference on IMU sequences at any scale so cadence can match application needs. It couples a motion-aware orientation filter with an Any-Scale Laplace Estimator (ASLE), a dual-task network that blends patch-based self-supervision with Bayesian regression. By modeling displacements with a Laplace distribution, ReNiL provides homogeneous Euclidean uncertainty that integrates cleanly with other sensors. A Bayesian inference chain links successive IPDPs into consistent trajectories. On RoNIN-ds and a new WUDataset covering indoor and outdoor motion from 28 participants, ReNiL achieves state-of-the-art displacement accuracy and uncertainty consistency, outperforming TLIO, CTIN, iMoT, and RoNIN variants while reducing computation. Application studies further show robustness and practicality for mobile and IoT localization, making ReNiL a scalable, uncertainty-aware foundation for next-generation positioning.

Figures (14)

• Figure 1: Overview of the ReNiL framework. The figure illustrates the ReNiL workflow in three stages.
• Figure 2: Motion-aware orientation filter based on adaptively weighted complementary. Gyroscope, accelerometer and magnetometer are weighted and updated orientation estimation with different windows.
• Figure 3: The structure and training tasks of the ASLE. The solid line represents the Bayesian regression task, and the loss is calculated by the label and the network output. The dashed line represents the self-supervised task, and the loss is calculated by the output of the contextual builder module for the input with data augmentation and without it.
• Figure 4: Bayesian process for pure multi-IPDPs localization or involved with external sources. ReNiL uses Bayesian inference to combine ASLE outputs of IPDP $t_n$ with prior position $p_{t_{n-1}}$ of IPDP $t_{n-1}$. For pure multi-IPDPs localization, ReNiL produces final posterior estimates $p_{t_n}$ directly; For multi-IPDPs localization involved with external sources, ReNiL updates Bayesian inference result with the external observations $z_{t_{n}}$ to produce final posterior estimates $p_{t_{n}}$. Then move to the next loop.
• Figure 5: Four typical cases from RoNIN-ds and WUDataset. The top figure of each column is a direct comparison of position estimation at different time scales, and the bottom figure shows the change of cumulative error over time.