A Probabilistic Framework for Multi-Sensor Fusion Based Indoor Localization on Mobile Platfor

Nowadays, smart mobile devices integrate more and more sensors on board, such as motion sensors (accelerometer, gyroscope), wireless signal strength indicators (WiFi, Bluetooth), and visual sensors (LiDAR, camera). People have developed various indoor localization techniques based on these sensors. In this dissertation, a probabilistic framework for multi-sensor fusion based indoor localization system is developed and partially implemented on a mobile platform.;The probabilistic fusion of multiple sensors is investigated in a hidden Markov model (HMM) framework for mobile device user localization. We propose a graph structure to store the model constructed by multiple sensors during offline training phase, and a multimodal particle filter to seamlessly fuse the information during online tracking phase.;The multi-sensor information for our data fusion and analysis includes WiFi received signal strength (RSS) collected from mobile device's received signal strength indicator (RSSI), motion signals gathered by built in motion sensors including accelerometer and gyroscope, and images captured by camera.;Based on our algorithms, we performed simulations in MATLAB and analyzed the results. We further implemented the indoor localization system on the iOS platform. The experiments carried out in typical indoor environment have shown promising results of the proposed algorithm and system design.