Vision Based Distributed Cooperative Navigation for MAVs in GPS denied areas

Abstract

Current Miniature Air Vehicle (MAV) systems rely on the availability of GPS to enable navigation (state estimation) during ∞ight. However, many envisioned MAV usage scenarios require MAVs that can ∞y in tightly constrained spaces (e.g., urban terrain, indoors, dense jungle, etc.) meaning that GPS may not be available. While inertial sensors can be used to estimate navigation state, their estimates will diverge over time without GPS. In this paper, we propose a method for overcoming this drift in a multiple-MAV scenario. We develop a distributed cooperative navigation system where each MAV is equipped with a bearing-only sensor (an electro-optical camera) and an inertial measurement unit. The camera is used to measure the bearing from other MAVs and/or landmarks which are in its fleld of view. By communicating local bearing information with other MAVs in their communication range, each MAV can estimate, without drift, its navigation state assuming the entire system of MAVs observes two or more landmarks at known locations. The distributed cooperative navigation system explicitly minimizes communication both in terms of message size and number of communication links. We describe su‐cient conditions for eliminating drift in the connectivity of the MAVs and how many landmarks are observed. We present simulation results to support the developed theory.