(Journal) Inverse k-visibility for RSSI-based Indoor Geometric Mapping
Junseo Kim^13, Matthew Lisondra^23, Yeganeh Bahoo^3, and Sajad Saeedi^3
(1) Tu Delft, (2) University of Toronto, (3) Toronto Metropolitan University
*** Journal Extension Submitted to IEEE Sensors Journal (ISJ), 2024 ***
Special Issue on Machine Learning for Radio Frequency Sensing
Paper (Preprint)
Graphical Abstract (Concept)
Inverse k-visibility for RSSI-based Indoor Geometric Mapping
In recent years, the increased availability of WiFi in indoor environments has gained an interest in the robotics community to leverage WiFi signals for enhancing indoor SLAM (Simultaneous Localization and Mapping) systems. SLAM technology is widely used, especially for the navigation and control of autonomous robots. This paper discusses various works in developing WiFi-based localization and challenges in achieving high-accuracy geometric maps. This paper introduces the concept of inverse k-visibility developed from the kvisibility algorithm to identify the free space in an unknown environment for planning, navigation, and obstacle avoidance. Comprehensive experiments, including those utilizing single and multiple RSSI signals, were conducted in both simulated and real-world environments to demonstrate the robustness of the proposed algorithm. Additionally, a detailed analysis comparing the resulting maps with ground-truth Lidar-based maps is provided to highlight the algorithm’s accuracy and reliability.
What is the Project about
Knowing accurately the mapped free space is imperative for the control of many autonomous systems, which need to know how to plan paths, making sure not to collide with occupied and obstacle spaces. The contributions of our work are:
(I) A novel algorithm that is capable of generating geometric maps using WiFi signals received from multiple routers,
(II) benchmarking the WiFi-generated maps with Lidar-generated maps by comparing the area, number of data points, k-value prediction True/False setting, k-value accuracy percentage, IOU and MSE scores,
(III) evaluation on real-world collected from indoor spaces.
FIGURE 1: The concept of k-visibility is demonstrated where different k-values are shown: k = 0 (red), k = 1 (green), k = 2 (blue) and k = 3 (yellow). Based on a straight-line measure from a reference point such as a router (shown in dark blue) to a desired location, k-visibility is a metric on how many times traverses through a wall/obstacle.
FIGURE 2: Diagram illustrating the ray-drawing principle fundamental to the inverse k-visibility algorithm. The wall is located where two consecutive k-value regions intersect along a ray projected from the router point. Only the region within the k = 1 area is displayed.
FIGURE 3: Ray-drawing for an arbitrary trajectory coordinates Ti with an associated k-value ki. According to the definition of k-visibility, the ray RTi must intersect exactly ki walls along its path.
FIGURE 4: Initial wall estimate along a ray (left). Enhanced wall estimation along a ray after adjustments to the lower and upper endpoints (right).
FIGURE 5: Assigning probability distributions based on the variations in kvalues across ray subsegments.
FIGURE 6: Visual demonstration of the geometric rules for areas with k = 0 and k = 1. Rule 1, depicting the robot’s trajectory, was omitted from the legend for clarity. Figure adapted from SfW Conference Paper.
ALGORITHM I: Multi-router case as the robot traverses a space.
FIGURE 7: Real-world experimental results for single-router and multi-router SfW mapping. (a) and (b) are for the same environment, but with one and two routers respectively. The same applies to (c) and (d). Our proposed method with RSSI-based mapping detects most of the free space and potential walls with k ≥ 1. The multi-router case has improved accuracy in estimating its map.
TABLE I: Data corresponds to Fig. 7. Please read caption in Fig. 7 to reference which set of data corresponding.
Contact
If you have any questions, feel free to reach out to us at the following email us at:
{junseo.kim, matthew.lisondra, bahoo, s.saeedi}@torontomu.ca
Reference and Bibtex Entry
Junseo Kim, Matthew Lisondra, Yeganeh Bahoo and Sajad Saeedi, "Inverse k-visibility for RSSI-based Indoor Geometric Mapping", 2024. Preprint.
@article{kim2024inverse,
title={Inverse k-visibility for RSSI-based Indoor Geometric Mapping},
author={Kim, Junseo and Lisondra, Matthew and Bahoo, Yeganeh and Saeedi, Sajad},
journal={arXiv preprint arXiv:2408.07757},
year={2024}
}