Performance Analysis of a New Collaborative and Multimodal Framework for Indoors Navigation using Smartphones

In the last decade, there was many advances in indoors navigation research, driven mainly by the proliferation of miniaturized and low-cost sensors, embedded inside our smartphones, enabling countless commercial opportunities, in entertainment, safety, and location based services (LBS). Many algorithms and techniques have been proposed to mitigate the errors and the drift of these low cost sensors. Some of the most recent techniques used for indoors navigation include: pedestrian dead reckoning (PDR) techniques, WiFi trilateration, WiFi fingerprinting, magnetic matching, map matching techniques, proximity sensing of peers, or a combination of some of these methods. Some of these methods require an accurate and updated maps of the area of interest. Acquiring these maps could be an expensive and time consuming process. The problem is exacerbated by the fact that the mapped features could be time dependent, requiring regular updates to the maps. As a result, some of the recent proposed methods incorporated automatic maps generation and update. The main objective of this paper is to propose and evaluate a new collaborative and multimodal framework for indoors navigation using smartphones to mitigate the current limitations of the indoors navigation algorithms. This framework unifies the positioning and mapping tasks, and uses only smartphones, without requiring previous knowledge of the environment, additionally it does not assume the existence of any specific infrastructure. In the context of the proposed framework, 'multimodal' refers to the different types of mapped features; such as, metric and topological floorplans, WiFi access point location map, WiFi access point power distribution map, and magnetic field density map. Additionally, 'collaborative' refers to the collaboration between different node, and it is defined on two dimensions: spatial and temporal. The spatial collaboration takes place when two or more nodes are in physical proximity. In this case the range information between nodes provides an additional input to the position estimation algorithm, which helps in enhancing the precision of the position estimate for each of the collaborating nodes. The temporal collaboration occurs through building and updating multimodal maps, by applying the simultaneous localization and mapping (SLAM) technique. In this paper the focus will be mainly on WiFi access point received signal strength indicator maps. However, the selected map representation is easily extended to other features, such as magnetic field intensity maps.