Indoor carpark often suffers from unavailable or weak GNSS (Global Navigation Satellite System) and cellular signals, Under such condition, we consider the challenging problem of navigating a driver with an offline smartphone docked at the car dashboard. There is some basic RF (radio-frequency) infrastructure in the premise, but due to signal attentuation by the car body, the location is noisy and intermittent. Previous works on carpark navigation often require special measurement equipment such as on-car additional infrastructure (OCAI), or perform integration of IMU (inertial measurement unit) signals over time. These are either not cost-effective to deploy or prone to high propagation error.

We propose RICH, a novel, real-time, simple and cost-effective docked-phone approach to fuse...[ Read more ]

Indoor carpark often suffers from unavailable or weak GNSS (Global Navigation Satellite System) and cellular signals, Under such condition, we consider the challenging problem of navigating a driver with an offline smartphone docked at the car dashboard. There is some basic RF (radio-frequency) infrastructure in the premise, but due to signal attentuation by the car body, the location is noisy and intermittent. Previous works on carpark navigation often require special measurement equipment such as on-car additional infrastructure (OCAI), or perform integration of IMU (inertial measurement unit) signals over time. These are either not cost-effective to deploy or prone to high propagation error.

We propose RICH, a novel, real-time, simple and cost-effective docked-phone approach to fuse RF and IMU signals for indoor carpark navigation using HMM (Hidden Markov Model). RICH is the first deployment-ready learning-based offline fusion approach executed completely in local phone without any OCAI or error-prone IMU integration. RICH uses IMU signals to classify accurately the car speed pattern and detect its heading and turning. This information and the crude RF localization are then fused in an HMM framework to compute the location distribution of the car. We present an analysis of RICH complexity and present its trade-off between computation and accuracy. We implement RICH in smartphones, and conduct extensive experiments in real carparks. As compared with the state of the art, RICH achieves substantially lower localization error (lower by 40%), and is computationally light-weight and fast suitablefor docked phone navigation(less than 10ms per location).