Multisource Fusion Framework for Environment Learning–Free Indoor Localization

Abstract

At the core of context-aware jobsite management is location information. For outdoor environments, global positioning systems (GPSs) are widely used. For indoor environments, however, an effective localization system has yet to be fully developed. Existing indoor localization systems usually rely on prior or real-time environment learning, and with just a slight change in the jobsite environment their performances degrade. Furthermore, localization systems that rely on a single sensor can hardly be everything a project manager would want—inexpensive, accurate, and easy to develop. Therefore, to simplify the deployment and enhance the robustness of localization for a dynamic environment, this work proposes a multisensor fusion framework. To simulate a typical residential jobsite’s indoor environment (with moving workers and ongoing activities), this work relies on two testbeds—an office area of 274 m2 with dynamic traffic flow and a lab of 92 m2 with ongoing lab tests. During working hours, researchers conducted performance evaluation tests in the testbeds. The results indicate that with simpler deployment the multisensor fusion algorithm was able to achieve the same accuracy level as existing systems without needing prior environment learning.