Redundant System for Precision Landing of VTOL UAVs

Abstract

This thesis investigates how Vertical Take Off and Landing (VTOL) Unmanned Aerial Vehicles (UAVs) can be automatically landed with high precision in a robust way. Automatic precision landing can allow pilots to land in tight spaces without resorting to manual flight. It can also allow for landing onto small docking stations without any pilot involved.

A study on relevant techonolgies for precision positioning and landing for VTOL UAVs is presented. This includes Ultra-Wideband (UWB) localization, RealTime Kinematic (RTK) Global Navigation Satellite System (GNSS), motion capture and visual target tracking.

A precision landing system is built and implemented to two different hexacopters. The presented precision landing system is a product of an improved landing algorithm in PX4, an open source autopilot software, and redundant precision positioning sensors. The two positioning systems selected are RTK GNSS, and IR-LOCK which is a system for visual target tracking of an Infrared Radiation (IR)-beacon.

From multiple flight tests the presented landing system has proved to outperform regular GNSS aided landings using either IR-LOCK or RTK GNSS. The latest results show an estimated worst case landing of 70 cm from target and a Root Mean Square (RMS) error of 30 cm. However, it is expected to be improved by better tuning of the Remotely Piloted Aircraft (RPA) s position controller.

It became obvious that the landing precision is greatly influenced by the position controller and agility of the RPA in use, but the presented landing system should be able to accommodate less agile RPAs by detecting horizontal biases and refusing final descent until being close enough to target. When being close enough the RPA is allowed to drop. The technique may resemble shooting at a target while holding a rifle unsteady and waiting for the crosshairs to enter the target before firing. The refusal of descent showed to only be activated for half of the landing attempts where needed and is therefore not fully working yet.

The system should not be considered fully working before every landing is within the accepted radius and precision loss can be detected reliably in order to retry landing or go to a fallback landing field. The system still needs further testing and improvements, but is already an improvement over the standard PX4 auto landing using regular GNSS.