Indoor Navigation with Deep Reinforcement Learning

Abstract

Indoor navigation solutions are essential for navigating within the building where GPS based solutions cannot be used. The applications of indoor navigations are robotics, drones, and few gaming apps. Deep reinforcement learning algorithms are used for building indoor navigation solutions, but such algorithms are computationally intensive and practically infeasible. But it is feasible to use reinforcement learning by reducing the scope and indoor scene representation information to propose an indoor navigation solution, which is the focus of the proposed work. The proposed work uses deep learning to detect objects in the given indoor scene and find a path using deep reinforcement learning. The proposed solution gives the navigation path in 2 minutes 10 seconds approximately, which is comparable with previous works. The authors also indicate the research work to be carried out in the future.