A Scalable Deep Neural Network Architecture for Multi-Building and Multi-Floor Indoor Localization Based on Wi-Fi Fingerprinting (1712.01990v1)

Abstract: One of the key technologies for future large-scale location-aware services covering a complex of multi-story buildings --- e.g., a big shopping mall and a university campus --- is a scalable indoor localization technique. In this paper, we report the current status of our investigation on the use of deep neural networks (DNNs) for scalable building/floor classification and floor-level position estimation based on Wi-Fi fingerprinting. Exploiting the hierarchical nature of the building/floor estimation and floor-level coordinates estimation of a location, we propose a new DNN architecture consisting of a stacked autoencoder for the reduction of feature space dimension and a feed-forward classifier for multi-label classification of building/floor/location, on which the multi-building and multi-floor indoor localization system based on Wi-Fi fingerprinting is built. Experimental results for the performance of building/floor estimation and floor-level coordinates estimation of a given location demonstrate the feasibility of the proposed DNN-based indoor localization system, which can provide near state-of-the-art performance using a single DNN, for the implementation with lower complexity and energy consumption at mobile devices.

Overview of Scalable DNN Architecture for Indoor Localization Using Wi-Fi Fingerprinting

The paper by Kim, Lee, and Huang explores a novel deep neural network (DNN) architecture that addresses the challenges of scalable indoor localization across multi-building and multi-floor structures using Wi-Fi fingerprinting. The research focuses on overcoming limitations in existing hierarchical localization approaches by leveraging a scalable DNN that integrates building and floor classification with location estimation.

Core Contributions and Methodology

The proposed DNN architecture consists of stacked autoencoders (SAE) for feature space dimension reduction and a feed-forward multi-label classifier designed to process Wi-Fi signal fingerprints for the identification of specific building and floor locations. The architecture aims to provide a more efficient, less complex solution that can operate effectively on mobile devices with reduced energy consumption compared to traditional methods.

The methodology involves reformulating the building/floor/location classification problem into a multi-label classification task, allowing for scalability advantages by significantly reducing the number of required output nodes in the DNN. This approach capitalizes on the hierarchical nature of indoor localization tasks, offering systematic label formation that ensures alignment between DNN outputs and real-world building, floor, and location identifiers.

Experimental Validation

Experiments were conducted using the UJIIndoorLoc dataset, featuring extensive data from real-world environments. The system's performance was evaluated primarily based on the success rates of building and floor hit, as well as the accuracy of position estimation. Key parameters, such as the number of elements selected from the output location vector and a scaling factor for threshold value during estimation, were tuned to optimize performance.

Results showed potential near-state-of-the-art performance, comparable to traditional techniques, with metrics revealing high building hit rates and reasonable floor hit rates. Given the inherent reduction in system complexity, these positive results suggest practical feasibility for large-scale deployment in complex indoor environments.

Implications and Future Directions

The implications of this research are significant both theoretically and practically. The scalable architecture provides a pathway towards localized processing, wherein DNNs enable efficient on-device computations without extensive server communication, facilitating energy-efficient and secure indoor localization systems.

Future prospects for this research include exploring optimizations of DNN parameters to enhance accuracy further, as well as potentially integrating evolutionary algorithms or other heuristic approaches to refine DNN training processes and performance metrics specifically related to building/floor detection and coordinate estimation.

Overall, the research contributes to the advancement of deep learning applications in indoor localization, offering insights and practical approaches that can influence future developments in smart city infrastructure and location-aware services.