Simultaneous Localization and Rough Indoor Floorplan Mapping Combining PDR and FM/Wi-Fi Radio Signals

Abstract

Among various indoor positioning methods, fingerprinting-based localization technology which utilizes ubiquitous radio signals has been widely studied due to advantages, such as low cost and high accuracy. To tackle the issues of time-consuming fingerprint database construction and enhance its applicability in unknown environments, radio signal simultaneous localization and mapping (Radio-SLAM) technology is utilized to automate the construction process of fingerprint database (also known as radio map). Nevertheless, existing Radio-SLAM methods mainly employ regression models for radio map construction, neglecting the influence of building structures on signal propagation, which leads to limited applicability in nonrepetitive path scenarios. Moreover, existing methods usually fail to generate indoor floorplans. To address these problems, this article integrates pedestrian dead reckoning (PDR) with frequency modulation (FM) radio signal from outdoor stations and Wi-Fi signal from indoor access points (APs) to achieve simultaneous localization and rough indoor floorplan mapping. First, windows, corners, and APs are selected as landmarks. Utilizing the impact of building structures on signal propagation, landmark detection is achieved through extracting the unique attenuation patterns of received signal strength indicator (RSSI). Second, the parameters of the propagation model containing landmark positions are inverted, based on which rough floorplan and large-area radio map are automatically constructed. Finally, PDR and fingerprinting localization results are fused. Experimental results demonstrate that the average positioning errors for landmarks are less than 2 m. Compared with the commonly used Gaussian process regression model, the proposed method significantly improves the accuracy of the radio map, thereby enhancing the localization performance.