A Particle Filter-Based Reinforcement Learning Approach for Reliable Wireless Indoor Positioning

摘要: Positioning is envisioned as an essential enabler of future fifth generation (5G) mobile networks due to the massive number use cases that would benefit from knowing users’ positions. In this work, we propose a particle filter-based reinforcement learning (PFRL) approach for robust wireless indoor positioning system. Our algorithm integrates information zone prediction, inertial measurement units, radio-based ranging, and floor plan into filter. The prediction method designed with ensemble by integrating individual discriminative methods Hidden Markov Models. Further, integrate filter learning-based resampling provide robustness against localization failure problems such kidnapping robot problem. PFRL validated on two-tier architecture, in which distributed machine tasks are hosted at client edge layer. Experiment results show our system outperforms traditional terminal-based approaches both stability accuracy.