Pelican Dataset Overview

Basic Information

• Maintainer: Nima Mohajerin (mohajerin.nima@gmail.com)
• Format: MATLAB .mat
• Size: 238.1 MB
• Download link: AscTec_Pelican_Flight_Dataset

Dataset Content

Vehicle Information

• Model: AscTec Pelican lightweight quadrotor
• Configuration: Real‑time autopilot board with an onboard computer running Ubuntu, communicating via UART
• Software: ROS Indigo for collecting motor speed and IMU data
• Operating environment: Indoor, piloted by an expert using a Futaba T7C remote controller

Dataset Description

• Number of flights: 54
• Data structure: MATLAB cell containing fields:
  • len: number of samples
  • Pos: position measurements (x, y, z) from Vicon, 3‑by‑len matrix
  • Euler: orientation measurements (roll, pitch, yaw) from Vicon, 3‑by‑len matrix
  • Motors: actual motor speeds, 4‑by‑len matrix
  • Motors_CMD: commanded motor speeds, 4‑by‑len matrix
  • Vel: computed velocities (v_x, v_y, v_z), 3‑by‑(len‑1) matrix
  • pqr: computed body rates (p, q, r), 3‑by‑(len‑1) matrix

Citation

If you use this dataset, please cite the following papers:

• [1] Deep Learning a Quadrotor Dynamic Model for Multi‑Step Prediction, N. Mohajerin, M. Mozifian, S. L. Waslander, IEEE International Conference on Robotics and Automation (ICRA), 2018.
• [2] Multistep Prediction of Dynamic Systems With Recurrent Neural Networks, N. Mohajerin, S. L. Waslander, IEEE Transactions on Neural Networks and Learning Systems, 2019.