Indoor Navigation Using Convolutional Neural Networks and Floor Plans

The Global Positioning System (GPS) is the primary solution for outdoor navigation. However, the signals from these satellites are blocked in indoor environments leading to the need for alternative indoor solutions. The goal of this paper is to evaluate a new indoor navigation technique by incorporating floor plans along with monocular camera images into a Convolutional Neural Network (CNN) as a potential means for identifying camera position. Building floor plans are widely available and provide potential information for localizing within the building. This work sets out to determine if a CNN can learn the architectural features of a floor plan and use that information to determine a location. In this work, a simulated indoor data set is created and used to train two CNNs. A classification CNN, which breaks up the floor plan into 100 discrete bins and achieved 76.1% top 5 accuracy on test data. Also, a regression CNN which achieved a distance error of 25.4 meters or less between the truth and predicted position on 80% of the test data. The models are further improved by combining them with a filter solution.The best performing classification CNN is evaluated on real world data captured via TurtleBot 3, demonstrating the potential for this solution to be useful to real world Air Force indoor localization problems.