Intelligent fusion of information derived from received signal strength and inertial measurements for indoor wireless localization

In this paper, we focus on improving the accuracy of wireless localization in wireless sensor networks using information derived from the inertial measurement unit (IMU) in a smartphone and the received signal strength (RSS). We propose an algorithm to relate the RSSs and measurements obtained from the IMU to the coordinates of an indoor robot. To deal with the dynamic nature of fingerprint information in an indoor radio environment, we first use the hierarchical Bayesian hidden Markov model (HB-HMM) to process a time series of RSSs. Unlike other HMM-based methods, the HB-HMM depends only on a single initial hyper-parameter for global optimization. Next, we evaluate the measurements obtained from the IMU to identify the robot's state, which includes the rotating, moving, and bumping states. We used the IMU accelerometers to estimate the velocity. Lastly, a method based on the particle filter (PF) was used to fuse the results obtained from RSS and IMU. Experiments show that our algorithm can achieve better accuracy than related algorithms.