Miguel Castillón

Miguel Castillón

Autonomous Navigation Lead

Keybotic

Hi!

I’m the autonomous navigation lead at Keybotic, a start-up developing four-legged robots for industrial inspections. Before this, I did my PhD on 3D perception and navigation of autonomous underwater robots: theory, software development and field tests. My main research interests intertwine numerical optimization and sensor perception for autonomous navigation in challenging scenarios.

Interests
  • SLAM and autonomous navigation
  • Optimization
  • Sensor Perception
Education
  • PhD in Autonomous Robotics, 2023

    University of Girona (Spain)

  • MSc in Robotics, 2018

    KU Leuven (Belgium)

  • BSc in Industrial Engineering, 2015

    University of Zaragoza (Spain)

Research experience

 
 
 
 
 
Autonomous Navigation Lead
Feb 2023 – Present Barcelona (Spain)
Responsible for the mapping and the autonomous navigation of our four-legged robot Keyper to enable autonomous inspections of real industrial plants.
 
 
 
 
 
PhD Researcher
Jan 2019 – Feb 2023 Girona (Spain)
Enabling underwater robots to use range data from laser-based sensors for mapping and manipulation tasks. To this end we built an underwater laser scanner, which allowed me to face challenges ranging from sensor calibration to integration in autonomous platforms.
 
 
 
 
 
Visiting Researcher
Sep 2021 – Apr 2022 Zurich (Switzerland)
Presented a new non-rigid point cloud registration algorithm that corrects motion distortion in dynamic scans, improving accuracy of localization and mapping tasks.
 
 
 
 
 
MSc Thesis
Oct 2017 – Jul 2018 Leuven (Belgium)
Developed a vision-based algorithm capable of reconstructing full-field vibrations of mechanical structures based on observations of natural-frequency deformations.
 
 
 
 
 
Research intern
Jun 2017 – Aug 2017 Leuven (Belgium)
Trained a ML classifier to assign ripeness category to berries detected by the autonomous strawberry picking robot.
 
 
 
 
 
BSc thesis
Jan 2015 – Jul 2015 Vienna (Austria)
Compared different feature-based strategies to match 2D floor plan LiDAR scans for autonomous navigation.