Research on Radio Frequency Finerprint Licalization based on Machine Learning

With the proliferation of the internet and mobile devices, location-based service (LBS) has become an indispensable part of everyday activities. Moreover, as a lot of indoor activities are conducted within concrete buildings with dense obstacles, localization methods that are able to provide accurate information with efficiency in such complex environments, is key to successful application of LBS. A variety of positioning technologies have been developed over the years. This paper has investigated and compared various machine learning methods for the prediction of locations based on RSS data. It introduces the recent development of RSS technology in indoor localization and further investigates the application of machine learning methods for location prediction. WIFI-based RSS methods address the challenges in indoor localization where GPS and sensor networks failed to solve. Machine learning models which predict location or coordinates generally achieve high accuracy. However, the choice of specific models depends on the environment setup. The application of RSS methods addresses the difficulties in localization in environments with obstacles. While significant improvement in accuracy can be achieved by machine learning techniques, the computational cost is still controllable with customization in environmental and device setup. The cost-benefit suggests further research in this area will be beneficial and potentially profitable for industrialization.