POSITION, VELOCITY AND HEADING ESTIMATION FOR UNMANNED AERIAL VEHICLES USING CAMERA AND INERTIAL SENSORS

Raphael Ballet Ballet¹; Davi Antônio dos Santos Santos¹; Zoran Sjanic Sjanic²

¹ Instituto Tecnológico de Aeronáutica; ² Linköping University

doi:10.20906/CPS/CON-2016-0515

Resumo

Precise and robust navigation of Unmanned Aerial Vehicles (UAV) in GNSS-denied environments is a relevant research topic, especially for autonomous systems. Most of the proposed solutions to this problem rely on expensive and high-precision equipment, thus limiting the accessibility of this technology for research. This work proposes a position, velocity and heading estimation algorithm for multirotor UAV using low-cost components, such as a gyro-stabilized platform containing a downward-facing onboard camera, inertial sensors and an ultrasonic range sensor, using a two-axis gimbal. The estimation algorithm is composed of three steps. At first, the vehicle must detect and identify visible landmarks by processing the images. Second, the algorithm computes the vectors from the camera center to each of the visible landmarks. Finally, measurements from the inertial sensors, ultrasonic range sensor and the landmark's vector measurements are fused with an Extended Kalman Filter (EKF) to estimate the vehicle's position, velocity and heading as well as accelerometer and rate-gyro biases. A map of landmarks is assumed available and the vehicle is supposed to operate indoors. The method is evaluated via extensive Monte Carlo simulations, which show its effectiveness and some of its properties.

Palavras-chave: Extended Kalman filter; Visual-inertial navigation; Sensor fusion; Unmanned Aerial Vehicle