| Indoor positioning has become an attractive research topic within the past two decades. However, no satisfying solution has been found with consideration of both accuracy and system complexity. Recently, research on visible light communications offers new opportunities on realizing accurate indoor positioning with relatively simple system configuration. This dissertation investigates several fundamental research topics of indoor positioning systems based on visible light communications (VLC) technology.;First, we introduce a simulation method to generate optical wireless channel impulse response more efficiently and accurately, based on which a novel 2-D positioning algorithm is developed. Using basic framed slotted ALOHA (BFSA) protocol, it successfully addresses channel multi-access issue, removing synchronization requirement on the transmitters. The algorithm makes use of received signal strength (RSS) to calculate the ranges from receiver to light-emitting diode (LED) bulbs, after which trilateration is performed to obtain estimation of the receiver's location. Simulation results show that, with over 95% confidence, the target receiver can be located with an accuracy of 5.9 cm, assuming indirect sunlight exposure and proper installation of LED bulbs.;The dissertation also explores one of the main digital signal processing techniques available: digital filters. Three filtering techniques are employed to improve the positioning accuracy: Kalman filter, sequential importance-resampling particle filter (SIR-PF) and Gaussian mixture sigma-point particle filter (GM-SPPF). Filtering techniques are used to form the trajectory of the receiver while rejecting wild values which are large deviations from correct measurements. Results show that filters help increase the positioning accuracy and Gaussian mixture sigma-point particle filter (GM-SPPF) outperforms other filters such as basic Kalman filter and particle filter (SIR-PF), at a reasonable computational cost.;3-D positioning algorithm is then proposed to keep track of the receiver's height. Simulation results show that the height of receiver can be estimated as accurately as within ten centimeters. Results also show that the estimation is more accurate when light distribution from LED bulbs involved are uniform.;Finally, sensor fusion technology is adopted as a solution to blockage problem and an option to further improve the positioning accuracy. |
