Applied Visible Light Positioning

Positioning is not a new problem. Humanity has attempted to determine its location for centuries with tools such as sextants, clocks, maps, and more. One of the largest revolutions in this field is arguably the advent of the global positioning system (GPS), which can provide position information almost anywhere on earth. Unfortunately, GPS and similar satellite navigation systems cannot be used indoors. The signal strength is reduced significantly by building walls, rendering the indoor accuracy of such systems unusable. Many researchers have attempted to find a solution. However, a single standard indoor positioning technology has not been achieved. Each solution has its advantages and drawbacks. The most apparent trade-off is between accuracy and cost. However, the novel technology of visible light positioning may be able to change that. In Visible Light Positioning (VLP), the visible light spectrum is used as a wireless communication medium. The transmitted signals are then used to determine the location of a receiver. A significant advantage of VLP is that the existing lighting infrastructure can often be re-used. Additionally, the local character of light reduces the impact of reflections, which leads to high accuracy. VLP received much attention in recent years. However, that research is often focused on improving the accuracy. For example, by introducing new modulation techniques, novel receiver designs, or some other technical feature. While this work certainly has its merits, this dissertation will instead focus on the bigger picture. By zooming out rather than in, this thesis aims to remove some entry barriers for VLP and increase the application potential. To that end, four concrete objectives are put forth. Before a VLP system can be used, it must first be calibrated. More specifically, the location and identity of each light source must be known accurately. The calibration step is often performed manually, which is a tedious and error-prone process. The topic of calibration is rarely discussed in literature yet is an integral step in deploying a VLP system. Therefore, the first objective of this work is to develop a simpler calibration method that can be used for high-performance VLP systems. A novel method is proposed that uses a mobile robot. Extensive experiments show that this approach is at least as accurate as manual calibration. Additionally, robot calibration is reliable and performs well under a range of circumstances. Indoor positioning research is increasingly using hybrid systems to cover the weaknesses of individual technologies. However, these hybrid systems have not often taken advantage of VLP. The second goal of this thesis is, therefore, to fuse VLP data with complementary sensors. Sensor fusion will likely improve the accuracy of VLP systems. More importantly, it will also increase the robustness of a system. Conventional VLP is only possible if multiple light sources are observed. The occlusion of one or more lights may interrupt tracking. By using additional sensors, these temporary outages can be compensated. The methods developed in this thesis will be evaluated with a mobile robot. Tracking of autonomous vehicles is arguably the most challenging application of indoor positioning systems. Therefore, the third objective is to enable the accurate tracking of mobile robots with visible light at a relatively low cost. If these demanding requirements can be met, it stands to reason that VLP could also be expanded to other areas with more modest requirements. Objectives 2 and 3 are combined by tracking the location of a mobile robot through the combination of VLP and encoder data. Three algorithms for sensor fusion are tested: an extended Kalman filter, a particle filter, and a hybrid particle/Kalman filter. Experiments are performed in a sparse lighting installation that results in frequent outages. The evaluation showed that accuracy remains high in this environment. Additionally, performance is not impacted significantly by external factors such as ambient light. Real-time operation on embedded computers is also shown to be possible. A drawback of VLP is the need for some modifications to the infrastructure. Additional hardware must be placed between each light and its power source to enable high-frequency modulation. Therefore, the final objective of this PhD is to explore the possibilities of unmodulated light as a signal of opportunity. This dissertation proposes a novel method for tracking a mobile robot by fusing unmodulated light with encoder data. Compared to modulated light, the accuracy and robustness of this approach are significantly lower. Unmodulated light is, therefore, best reserved for applications with modest requirements. Overall, this work lowers some of the entry barriers to VLP. Future work can expand on the ideas put forth in this thesis to break down these walls even more. VLP may then someday become a ubiquitous indoor positioning technology.

Jaar van publicatie:2021

Toegankelijkheid:Open