Skip to Main Content
Many applications of unmanned aerial vehicles (UAVs) require the capability to navigate to some goal and to perform precise and safe landing. In this paper, we present a visual navigation system as an alternative pose estimation method for environments and situations in which GPS is unavailable. The developed visual odometer is an incremental procedure that estimates the vehicle's ego-motion by extracting and tracking visual features, using an onboard camera. For more robustness and accuracy, the visual estimates are fused with measurements from an Inertial Measurement Unit (IMU) and a Pressure Sensor Altimeter (PSA) in order to provide accurate estimates of the vehicle's height, velocity and position relative to a given location. These estimates are then exploited by a nonlinear hierarchical controller for achieving various navigation tasks such as take-off, landing, hovering, target tracking, etc. In addition to the odometer description, the paper presents validation results from autonomous flights using a small quadrotor UAV.