CBL - Campus del Baix Llobregat

Projecte llegit

Títol: Development, simulation and validation of a follow-me algorithm for UAS based on PX4

Estudiants que han llegit aquest projecte:

Director/a: REMONDO BUENO, DAVID

Departament: ENTEL

Títol: Development, simulation and validation of a follow-me algorithm for UAS based on PX4

Data inici oferta: 31-01-2020     Data finalització oferta: 01-02-2020


Estudis d'assignació del projecte:
    MU DRONS
Tipus: Individual
 
Lloc de realització: Fora UPC    
 
        Supervisor/a extern: Pere Molina
        Institució/Empresa: Geonumerics
        Titulació del Director/a: M.Sc. Professional and Advanced Mathemat
 
Paraules clau:
UAS, Drone, PX4, MAVlink, MAVSDK, ROS, Gazebo, Follow-me, open-source, control, mapKITE, QGroundControl
 
Descripció del contingut i pla d'activitats:
The company plans to use mapKITE as a solution for corridor
mapping that combines a drone and a terrestrial vehicle to do
simultaneous 3D mapping and georefencing. Corridor mapping is the
collection and processing of environmental data along straight
paths like roads, rivers and powerlines. This technique uses
traditional Ground Control Points or GCPs in order to get accurate
global positioning of the point cloud. One of the innovations of
mapKITE is the use of novel Kinematic Ground Control Points or
KGCPs, this optical target is located in the top of the car and it
needs to be seen by the drone flying above it at all times.

Current autopilot technology based on open-source PX4 flight stack
provides a basic follow me flight mode that doesn’t meet the
performance criteria of mapKITE. The aim of this work is to
improve the current “follow me” algorithm, perform simulations and
then test the code on an actual commercial drone for validation.
MAVSDK is a MAVlink library with API that enables programmatic
control of the vehicle. The code will be written in Python and
will be deployed on the GCS device (laptop or tablet). Different
control and estimation techniques will be evaluated, as well as AI
methods for optical tracking.
 
Overview (resum en anglès):

The commercial drone industry has seen a huge growth in the recent years. UAS regulations are being drafted all over the world, and many industries are looking to improve the efficiency and safety of their operations with the use of autonomous drones and robotics. Inspection and Maintenance (I&M) represents a large economic activity with a global market estimated at 450 billion Euros. Drone mapping is a new and growing area within the I&M market that has seen a lot of innovation in the past years thanks to advancements in both autonomous vehicles technology and photogrammetry/mapping software.

This thesis focuses on the research and development (R&D) of the technology needed for mapKITE, the flagship drone-based mapping project of Geonumerics S.L. MapKITE is a new type of mapping concept, which combines two new paradigms: a new tandem terrestrial-aerial geodata acquisition technique and the use of a terrestrial vehicle to provide continuous Ground Control Point (GCP) information. In order to achieve the operational requirements of mapKITE, the company is developing a technology named vTether.

The main challenge is that the current open-source follow-me flight mode does not match the technical requirements needed for a mapKITE mission. This work provides a solution to this problem and answers the question: what is the best way to support future mapKITE¿s operations in the long term? After many weeks of research and consultation with the PX4 community, the solution found was to: 1) modify a PX4 firmware v1.10 in order to enhance the basic follow-me flight mode available and 2) code a command line program that uses MAVSDK API to provide off-board control of the drone.

The final part of the thesis includes the validation of the autopilot modifications with the help of Software in The Loop (SITL) simulations. The robotics simulator Gazebo was used for this purpose. We demonstrated the capacity of the custom PX4 version to provide a basic follow-me in 3D space at a speed of 19km/h. Future work involving the addition of a visual tracking system and artificial intelligence is discussed.

© CBLTIC Campus del Baix Llobregat - UPC