Enhanced V-SLAM combining SVO and ORB-SLAM2, with reduced computational complexity, to improve autonomous indoor mini-drone navigation under varying conditions

Mini-drones have a wide range of applications, including weather monitoring, parcel delivery, search and rescue, and entertainment. As their functionality, safety, and performance heavily relies on ubiquitously reliable positioning and navigation, their applications are relatively limited to outdoor environments where Global Positioning System (GPS) and/or similar are available. Indoor localization is improving, e.g, using Visual Simultaneous Localization and Mapping (V-SLAM). However, for the applications of mini-drone navigation with a higher safety requirements, further improvements are still needed. This paper proposes a novel approach to improve the localization performance for mini-drone indoor navigation. The proposed approach enhances V-SLAM techniques by combining Oriented Rotated Brief SLAM (ORB-SLAM2) and Semi-direct monocular Visual Odometry (SVO) algorithms along with an Adaptive Complementary Filter (ACF). The results show that the proposed approach improvement in position estimation in the different conditions (low light, low texture and dynamic environments) in comparison with other commonly used indoor positioning approaches.

I. INTRODUCTION

Mini-drones are used in a wide range of applications, from aerial mapping to smart farming, search and rescue [1], [2]. They primarily use the outdoor positioning technologies, i.e, Global Navigation Satellite Systems (GNSS) [3]–[5] such as GLONASS, Galileo and GPS [6], [7]. However, such technologies don’t fit the indoor use. Also, autonomous minidrone navigation inside buildings is subject to , e.g., constraints on payload size, mandatory weight cap limiting the number and types of sensors to be embedded, among the others, which make the navigation even more challenging.