Improving BLE Beacon Proximity Estimation Accuracy Through Bayesian Filtering

Andrew Mackey, Petros Spachos, Liang Song, Konstantinos N. Plataniotis

TL;DR

This paper tackles the challenge of accurately estimating proximity using Bluetooth Low Energy (BLE) beacons in indoor environments, where signal strength (RSSI) is notoriously unreliable. The authors developed a mobile application and evaluated three Bayesian filtering techniques (Kalman, Particle, and Non-parametric Information filters) against a simple moving average, demonstrating significant improvements in proximity estimation accuracy, especially at close distances.

What Problem Does This Paper Address?

The Internet of Things (IoT) is expanding rapidly, enabling context-aware applications like Proximity-Based Services (PBS) and Location-Based Services (LBS) within buildings. A critical component for these services is accurate indoor positioning, which is problematic for GPS due to signal obstruction. While various wireless solutions exist, Received Signal Strength Indicators (RSSI) from BLE beacons are a popular choice due to their low cost, low power consumption, and ease of deployment. However, RSSI is highly susceptible to environmental factors like signal interference and multipath fading, leading to inaccurate proximity estimations. This paper aims to improve the accuracy of BLE beacon-based proximity estimation by applying advanced filtering techniques to mitigate these RSSI fluctuations.

Methods

The core of this research lies in the application and comparison of three Bayesian filtering techniques implemented on a mobile platform:

  • Simple Moving Average (SMA): Used as a baseline for comparison, this method calculates the average of the last N RSSI measurements. The authors found a window size of 20 to be optimal.
  • Kalman Filter (KF): A recursive filter well-suited for linear and Gaussian systems. The paper explores both a static (KF-ST) and a dynamic (KF-DN) version, where the latter accounts for changing process noise. The KF is implemented directly on the mobile application.
  • Particle Filter (PF): A nonlinear Bayesian filter that can handle noisy measurements and non-Gaussian distributions. It estimates the current state by recursively utilizing all available measurement observations. The authors used 100 particles, determined through iterative experimentation.
  • Non-parametric Information (NI) Filter: A closed-form Bayesian filter designed to handle nonlinear and non-Gaussian conditions, making it particularly suitable for the non-parametric path loss model of RSSI. It combines a kernel density estimator (KDE) with a dynamic model prediction.

The authors developed a custom Android application to measure RSSI values from BLE beacons and convert them into distance estimations, which were then fed into these filters.

Datasets

The paper does not explicitly mention pre-existing datasets. Instead, the authors conducted their own experiments to collect data. They gathered approximately 1000 RSSI data points at each of 14 defined distances for each beacon in each environment. This data was used for both calibrating the path loss model and evaluating the performance of the filtering techniques.

Key Insights and Ideas

  • RSSI is inherently noisy: The experiments clearly demonstrate that raw RSSI values are highly variable and unreliable for accurate proximity estimation without filtering.
  • Bayesian filters offer significant improvements: All tested Bayesian filters (Kalman, Particle, NI) consistently outperformed the simple moving average, highlighting the effectiveness of these advanced techniques.
  • Environment matters: The small room exhibited higher interference, leading to lower overall accuracy, underscoring the need for robust filtering in challenging environments.
  • Close proximity is key: The most significant improvements in accuracy were observed when the beacon and receiver were within 3 meters, where the filters could better resolve fine-grained proximity.
  • Beacon hardware differences exist: While all beacons benefited from filtering, there were variations in performance, with the Gimbal beacon showing consistently good results across experiments.
  • Mobile implementation is feasible: The successful implementation of these filters on a smartphone demonstrates the practicality of on-device processing for real-time proximity estimation.

Key Results

The Bayesian filtering techniques significantly improved proximity estimation accuracy. Specifically:

  • Mean Absolute Error (MAE):
    • In Environment 1 (large room), the Particle filter achieved the best average MAE of 0.27 m within 3 meters.
  • On average, the Particle filter improved accuracy by 39.17% compared to SMA.
  • In Environment 2 (small room), the NI and Particle filters showed the best performance, with NI having the lowest RMSE of 0.63 m within 3 meters.
  • Root Mean Squared Error (RMSE):
    • On average, the Particle filter improved RMSE by 49.48% compared to SMA in Environment 1.
  • Overall, the paper reports improvements of up to 30% in proximity estimation accuracy compared to traditional filtering when the beacon and receiver are within 3 meters.

Limitations and Future Work

  • Environmental dependency: The path loss model and filter parameters were tuned for specific environments. The authors acknowledge that these parameters may need re-calibration for new environments.
  • Beacon placement: The paper suggests that strategic placement of numerous beacons is crucial for high accuracy LBS, especially beyond 1.5 meters.
  • Interference: The small room's higher interference highlighted the challenges and the necessity of filtering. Further research could explore optimal beacon layouts to minimize interference.
  • Beacon hardware: While three different beacons were tested, the authors suggest that further tests with other BLE beacon devices would be beneficial to confirm the generalizability of their findings.
  • Future work: The authors suggest that alternative deployment scenarios with greater system noise and physical obstructions should be investigated, as well as exploring optimal filter tuning parameters for various environments.

Summary updated

Abstract

The interconnectedness of all things is continuously expanding which has allowed every individual to increase their level of interaction with their surroundings. Internet of Things (IoT) devices are used in a plethora of context-aware application, such as proximity-based services (PBSs), and location-based services (LBSs). For these systems to perform, it is essential to have reliable hardware and predict a user’s position in the area with high accuracy in order to differentiate between individuals in a small area. A variety of wireless solutions that utilize received signal strength indicators (RSSIs) have been proposed to provide PBS and LBS for indoor environments, though each solution presents its own drawbacks. In this article, Bluetooth low energy (BLE) beacons are examined in terms of their accuracy in proximity estimation. Specifically, a mobile application is developed along with three Bayesian filtering techniques to improve the BLE beacon proximity estimation accuracy. This includes a Kalman filter, a particle filter, and a nonparametric information (NI) filter. Since the RSSI is heavily influenced by the environment, experiments were conducted to examine the performance of beacons from three popular vendors in two different environments. The error is compared in terms of mean absolute error (MAE) and root mean squared error (RMSE). According to the experimental results, Bayesian filters can improve proximity estimation accuracy up to 30% in comparison with traditional filtering, when the beacon and the receiver are within 3 m.
BibTeX
@article{mackey2020improving,
  author = {Andrew Mackey and Petros Spachos and Liang Song and Konstantinos N. Plataniotis},
  title = {{Improving BLE Beacon Proximity Estimation Accuracy Through Bayesian Filtering}},
  journal = {IEEE Internet of Things Journal},
  year = {2020},
  url = {https://openreview.net/forum?id=tvC7HkBKTO},
}
Figure