TL;DR
This paper introduces a novel indoor positioning system using visible light communication (VLC) that leverages the geometric features of circular luminaires, specifically their arcs, to achieve centimeter-level accuracy without relying on inertial measurement units (IMUs) or being limited by camera tilt angles. The proposed V-PA system adaptively uses two algorithms, V-PCA and OA-V-PA, to handle scenarios with complete or incomplete luminaire captures, demonstrating robust performance in both simulations and real-world experiments.
What Problem Does This Paper Address?
Accurate indoor positioning is crucial for a wide range of location-based services, but existing technologies like Wi-Fi, RFID, and Bluetooth often struggle to balance accuracy with implementation cost. Visible Light Positioning (VLP) using VLC has emerged as a promising alternative due to its potential for high accuracy and low cost. However, many VLP systems treat luminaires as simple point sources, neglecting their geometric characteristics. This simplification often requires multiple luminaires and can be sensitive to the user's field of view. Furthermore, existing methods that utilize the circular shape of luminaires often rely on IMUs, which are prone to measurement errors, or have limitations regarding the user's camera tilt angle. This paper addresses the need for a more practical and accurate VLP system that can effectively utilize the geometric information of circular luminaires, even when they are partially occluded or viewed at various angles, without requiring additional sensors like IMUs.
Methods
The core of the proposed system is the V-PA (Visible Light Positioning with Perspective Arcs) approach, which adaptively employs two algorithms:
- V-PCA (Visible Light Positioning with Perspective Circle and Arc): This algorithm is designed for scenarios where the camera captures at least one complete circular luminaire and potentially an incomplete one. It leverages plane and solid geometry to estimate the luminaire's orientation and position in the camera's coordinate system. Key steps include:
- Estimating the normal vector of the luminaire in the camera coordinate system (CCS) by fitting an ellipse to the luminaire's projection and using auxiliary coordinate systems.
- Estimating the 3D coordinates of the luminaire's center and a designated mark point in CCS.
- Utilizing single-view geometry theory to determine the user's pose and location in the world coordinate system (WCS).
- OA-V-PA (Anti-Occlusion Visible Light Positioning with Perspective Arcs): This algorithm is introduced to handle situations where only incomplete circular luminaires (arcs) are captured, which can occur due to limited field of view or occlusions. It addresses the challenge of determining the luminaire's center when a full circle or mark point is not visible.
- It fits ellipses to the captured arcs and approximates the luminaire's center with the center of the fitted ellipse.
- It then proceeds with pose and location estimation similar to V-PCA, but using the approximated center.
The system employs a fused image processing scheme to simultaneously extract VLC signals and geometric information. This involves capturing images at different exposure times: a long exposure to capture the luminaire's outline (circle or ellipse) for geometric feature extraction, and a short exposure to capture fringe patterns for VLC signal demodulation.
Datasets
The paper does not explicitly mention training on specific datasets. Instead, it relies on simulations and experimental data collected from a custom-built prototype. The experimental setup involves a controlled indoor environment with 12 circular luminaires.
Key Insights and Ideas
- Leveraging Luminaire Geometry: The paper demonstrates that the geometric features of circular luminaires (arcs and circles) provide rich information for accurate positioning, going beyond treating them as simple points.
- Eliminating IMU Dependency: By using geometric cues from camera images, the system achieves accurate pose estimation without the need for IMUs, thus avoiding their inherent measurement errors.
- Handling Partial Occlusions: The OA-V-PA algorithm is a significant contribution, enabling positioning even when luminaires are partially visible, which is a common challenge in real-world VLP applications.
- Adaptive Algorithm Selection: The V-PA framework, which dynamically chooses between V-PCA and OA-V-PA based on the captured luminaire information, enhances the system's robustness and practicality.
- Centimeter-Level Accuracy: The experimental results show that the proposed V-PA system can achieve centimeter-level positioning accuracy, which is highly competitive for indoor positioning.
Key Results
- Simulation Results:
- V-PA achieves a 90th percentile location accuracy of approximately 10 cm.
- V-PA demonstrates superior robustness to image noise compared to PnP and V-IMU algorithms.
- Positioning accuracy improves with increasing luminaire radius.
- The length of the captured arc significantly impacts accuracy, with longer arcs leading to better performance.
- Experimental Results:
- The V-PA algorithm achieves an average location error of less than 5 cm, even with the user tilted at various angles (0°, 15°, 30°).
- The minimum error observed was 0.87 cm, and the maximum error was 6.82 cm.
- The experimental prototype confirms the feasibility and centimeter-level accuracy of the proposed V-PA system.
Limitations and Future Work
- Limitations:
- The OA-V-PA algorithm uses an approximation for the luminaire's center when only arcs are available, which can introduce slight estimation errors.
- The performance is still affected by image noise, although V-PA shows better robustness than comparative methods.
- The current system relies on specific luminaire designs (circular) and may not be directly applicable to other shapes without modifications.
- Future Work:
- The paper suggests that the V-PA approach can be extended to handle more complex scenarios and luminaire configurations.
- Further research could explore more sophisticated methods for approximating luminaire centers in OA-V-PA to reduce estimation errors.
- Investigating the impact of different luminaire shapes and arrangements on the V-PA system's performance would be a valuable direction.
Summary updated