1/1
4 files

The open database TRAM-FPV. Indoor flight sequences for the study and training of racing drones.

dataset
posted on 28.10.2022, 14:58 authored by Jose M Castiblanco QuinteroJose M Castiblanco Quintero, Dmitrii IgnatevDmitrii Ignatev, Sergio García-Nieto Rodriguez

  

  

BRIEF INTRODUCTION: 

  

The flight sequences were performed with five kinds of racing drones. They have two Symmetrical (SY) airframes, two non-Symmetrical (NSY) and a Hybrid (HS). It stores data from the optical (OMS) and inertial (IMU) sensors. In addition to the mass distribution of each model.


They were conducted in the Cranfield University Flight Arena. The test area sizes are around 20X20X10 metres. Also, it is equipped with 30 Vicon cameras. Those are for capturing the motion of an object in real time. The tracker software handles it.


The layout of the flight arena allowed high speeds on short lengths. It is solved into flight sequences in a controlled background. Thus, to study dynamic behaviour mainly. In addition, it shows the option to design new kinds of controllers. Also, to study trajectory planning or training autonomous drones. Fusing data between sensors is another everyday use. 


The Tracker software can be combined with third-party software. It is via UDP transmission and a self-developed SDK. Software like Matlab/Simulink or LabView is an excellent example of this.

 

FILES FORMATS:

 
 

This dataset contains three file formats: CSV, XLSX and WebM files. The CSV files store the data of the Vicon-tracker software. Also, the XLSX files store the data collected by the inertial measurement unit (IMU). Moreover, the WebM files store the flight series in video format.

 

STORED INFORMATION:

 
 

The CSV files show relative position data (x, y, z). Also, it has an angular position (Rx, Ry, Rz). The XLSX files have the data on acceleration. Also, it includes the data on angular velocity. At the same time, the video files show the data synchronised. In addition, camera calibration files have been provided in XCP format along with the drone's heading angles.

 

ROBUSTNESS OF THE DATASET:

 
 

Each model will undergo 30 flight tests. A test flight lasted between 2.5 and 3 minutes, approximately 40 laps, for a total of 1200 laps for each model tested. The vision system was set up to capture motion at 240 frames per second, while the IMU data is stored at 500 frames per second.

 

BRIEF DESCRIPTION OF THE TESTED MODELS:

 
 

The electronic components are the same for all models. Also, the controller settings were by default. Three types of airframes were tested. They are named symmetrical (SY), non-symmetrical (NSY) and hybrid (HS) airframes. 

The kind of airframes has different weights. They vary between 250 and 320 grams without a battery. In addition, the wheelbase (lambda distance) is around 220 and 250 millimetres. The angular distance (alpha and beta) is about 65 and 90 degrees. (For further information, see bibliographical references.)


The mass distributions of the models have been added. These excel files include the model's density, area, volume and specific mass values. The moments of inertia for each component embedded in the airframe and the fully assembled system are also presented. (To read the Spanish version please go to the bibliographic reference)


 

History

Authoriser (e.g. PI/supervisor)

D.Ignatyev@cranfield.ac.uk