Tracking Wildfire Assets with Commodity RFID and Gaussian Process Modeling

TL;DR

The paper tackles scalable wildfire asset tracking with low-cost passive RFID by introducing a Gaussian Process-based environmental fingerprinting framework that localizes tags without requiring known ground-truth tag locations. It builds an environment model dictionary and uses a weighted likelihood to match observed RF signatures to environmental models, enabling GPS-like accuracy with RFID. Extensive experiments across six environment types and multi-tag scenarios demonstrate sub-meter to a few-meter localization errors, outperforming KNN baselines and showing robustness to unseen terrains. The approach is scalable to mobile readers and dozens of tags, though it relies on adequate dictionary coverage and coordinated deployments for large fires.

Abstract

This paper presents a novel, cost-effective, and scalable approach to track numerous assets distributed in forested environments using commodity Radio Frequency Identification (RFID) targeting wildfire response applications. Commodity RFID systems suffer from poor tag localization when dispersed in forested environments due to signal attenuation, multi-path effects and environmental variability. Current methods to address this issue via fingerprinting rely on dispersing tags at known locations {\em a priori}. In this paper, we address the case when it is not possible to tag known locations and show that it is possible to localize tags to accuracies comparable to global positioning systems (GPS) without such a constraint. For this, we propose Gaussian Process to model various environments solely based on RF signal response signatures and without the aid of additional sensors such as global positioning GPS or cameras, and match an unknown RF to the closest match in a model dictionary. We utilize a new weighted log-likelihood method to associate an unknown environment with the closest environment in a dictionary of previously modeled environments, which is a crucial step in being able to use our approach. Our results show that it is possible to achieve localization accuracies of the order of GPS, but with passive commodity RFID, which will allow the tracking of dozens of wildfire assets within the vicinity of mobile readers at-a-time simultaneously, does not require known positions to be tagged {\em a priori}, and can achieve localization at a fraction of the cost compared to GPS.

Figures (7)

• Figure 1: Proposed concept of commodity RFID to track wildfire assets where RFID tags are distributed across assets and are tracked using a system of readers mounted on mobile vehicles.
• Figure 2: The six environments used for data collection: Environments 1, 2, and 5 are forested environments that were 100m - 200m away from each other, with Environment 6 nearly half a kilometer away from the other environments. Environments 3 and 4 have minimal obstructions and are used as baselines for grassy and rocky terrain.
• Figure 3: Data was collected by placing the tag at varying distances and angles from the antenna centerline, using a camera and measuring tape to record positions. Over 15 days, approximately 27,000 data points were gathered per environment, totaling 162,000 across six environments. While mapping signal attributes for a single tag, multiple tags were also placed to assess system performance under multi-tag conditions. Although RFID systems activate one tag at a time due to anti-collision protocols, the physical presence of other tags can still interfere with the signal.
• Figure 4: This figure shows the signal attributes observed in Environment 2. Although each environment had its own unique signal response, they all shared a general trend in signal behavior. Each environment exhibited distinct construction and destructive interference regions, including dead zones, which manifested as outliers in the $\Delta\phi$ data.
• Figure 5: This figure illustrates the mean distance errors for each environment and the corresponding 90th percentile error ranges generated by the KNN and both GP methods. The GP outperforms KNN, and by incorporating information about the tag's trajectory, the accuracy can be further improved, achieving a range error of less than 2.5 m in most cases, a full meter below the errors obtained through the application of KNN.