Robust indoor localization with smartphones through statistical filtering

Location-based service is becoming more and more significant for nowadays mobile applications. In this paper, we develop an indoor localization scheme by only using a smartphone and existing WiFi infrastructure. To fuse both map information and data collected from motion sensors, we exploit particle filters to estimate the probability distribution of location state. However, it is challenging that smartphone sensor data is usually inaccurate due to complicated indoor environment. Therefore we develop a novel filtering technique to reduce the sensor measurement error based on Kalman filter. We evaluate our proposed method over real world traces collected in a large indoor environment of 3750m², the experimental results have shown that we can achieve a low localization median error of 2.69m based on heading measurements filtered with our Kalman filter, 0.81m lower than the case when low-pass filters are applied.