Geocaching-Inspired Navigation for Micro Aerial Vehicles with Fallible Place Recognition

Abstract

This paper extends an existing decisional framework for the navigation of Micro Aerial Vehicle (MAV) swarms. The work finds inspiration in the geocaching outdoor game. It leverages place recognition methods, information sharing and collaborative work between MAVs. It is unique in that a priori none of the MAVs knows the trajectory, waypoints and destination. The MAVs collectively solve a series of problems that involve the recognition of physical places and determination of their GPS coordinates. Our algorithm builds upon various methods that had been created for place recognition. The need for a decisional framework comes from the fact that all methods are fallible and make place recognition errors. In this paper, we augment the navigation algorithm with a decisional framework resolving conflicts resulting from errors made by place recognition methods. The errors divide the members of a swarm with respect to the location of waypoints (i.e., some members continue the trip following the proper itinary; others follow a wrong one). We propose four decisional algorithms to resolve conflicts among members of a swarm due to place recognition errors. The performance of the decisional algorithms is modeled and analyzed.