Vision aided INS/GNSS integration for improving the robustness of a navigation system for Mini Unmanned Aerial Vehicles

Mini Unmanned Aerial Vehicles (MUAV) such as mini-drones are becoming of high interest for surveillance, search and rescue. A challenging issue for drone navigation is the capability of navigating with a system equipped with embedded sensors. On the one hand, stringent constraints related to size, power, weight and cost have to be considered. On the other hand, the navigation system has to provide reliable information to the autonomous guidance system. When the navigation system is based on a low cost Inertial Measurement Unit (IMU) coupled to a GNSS receiver, this system cannot fulfill the integrity requirements in very constrained environments such as city centers, where multipath and satellite outages degrade positioning performance. To make the navigation system suitable for this kind of environment, we propose in the frame of this study to design a vision aided Inertial Navigation System (INS) tightly coupled with a GNSS receiver. The algorithm which is described in this paper exploits monocular vision sensor data, barometer output and GNSS measurements for performing INS corrections using an algorithm based on an unscented Kalman filter. The additional monocular vision sensor provides an observation of the MUAV motion by using optical flow techniques. It is exploited for reducing the inertial solution drift and allows the fusion algorithm to manage efficiently GNSS measurements to ensure the required navigation accuracy and reliability. The proposed approach aims at providing reliable GNSS measurements in urban environments for INS calibration. It consists of online multipath detection and processing. To achieve this objective, the algorithm exploits a vision sensor to improve the a priori estimate of the vehicle state before performing chi-square tests for identifying pseudo range (PR) and delta range (DR) errors due to multipath. Moreover it exploits a multiple-measurement model which allows the system to merge efficiently PR and DR measurements. This approach is a good way of tackling multipath problems in a tightly-coupled INS/GNSS integration scheme in case of slow channel fading. It provides a bounded-error state estimation which is required by the autonomous guidance system to carry out the MUAV mission.