PALMS: Plane-based Accessible Indoor Localization Using Mobile Smartphones

TL;DR

PALMS is an innovative indoor global localization and relocalization system for mobile smartphones that utilizes publicly available floor plans that leverages the Certainly Empty Space (CES) constraint along with principal orientation matching to create a spatial probability distribution of the device’s location.

Abstract

In this paper, we present PALMS, an innovative indoor global localization and relocalization system for mobile smartphones that utilizes publicly available floor plans. Unlike most vision-based methods that require constant visual input, our system adopts a dynamic form of localization that considers a single instantaneous observation and odometry data. The core contribution of this work is the introduction of a particle filter initialization method that leverages the Certainly Empty Space (CES) constraint along with principal orientation matching. This approach creates a spatial probability distribution of the device's location, significantly improving localization accuracy and reducing particle filter convergence time. Our experimental evaluations demonstrate that PALMS outperforms traditional methods with uniformly initialized particle filters, providing a more efficient and accessible approach to indoor wayfinding. By eliminating the need for prior environmental fingerprinting, PALMS provides a scalable and practical approach to indoor navigation.

Figures (3)

• Figure 1: Architecture of PALMS' particle filter initialization pipeline. The ARKit-based application (top left) detects vertical planar patches (walls) in a $360^{\circ}$ LiDAR scan. We used their 3-D poses to create 4 convolution kernels through principal orientation matching, shrinking, or smoothing. Each kernel is rotated to match a possible orientation. We then convolve the rasterized floor plan with each of the kernels to create 4 heatmaps $H_{0\sim3}$. We extract the top 1% locations from each heatmap and randomly sample 4 groups of particles, each with $p$ particles sharing the same initial drift from the corresponding orientation. These particles will then be placed in one particle filter. Colored edges and dots indicate different orientation groups. Orange represents the correct orientation, and the green cross indicates the ground truth observation point.
• Figure 2: Certainly Empty Space (CES). CES enforces visibility constraints to eliminate physically impossible matches. The figure shows two scenarios: in match 1 (top right), the constraint is met as no floor plan segments fall within the CES (orange). In match 2 (bottom right), several segments intersect with the CES (bright red), making it a less viable match.
• Figure 3: Visualization of the particle filters under different experiment settings at the time of initialization, at different stages of convergence, and when the path ends. Red dot: mean location of all particles (before $t_1$), mean position of the dominant group (after $t_1$), mean position of the dominant cluster (after $t_2$). Green: ground truth path and location. All three settings share the same starting point, tracking data, ground truth path, and particle count. Particles in different orientation groups are represented by different colors. $t = x$ shows the time in seconds. We can see that PALMS successfully localized the user at $t_2$ and tracks the user until $t_{end}$ while the other methods failed to do so. Note that for some cases the 1st stage and 2nd stage convergence happen simultaneously.