Ioannis Mandralis

Ioannis Mandralis

Ioannis Mandralis

PhD Candidate | Center for Autonomous Systems and Technologies | Caltech

About Me

I am a PhD Candidate at the Center for Autonomous Systems and Technologies (CAST) at Caltech, where I primarily work with Professor Morteza Gharib, and Professor Richard M. Murray.

I work on a broad range of topics all related to autonomy of robotic systems. My approach relies on using new tools from learning & control with an emphasis on applying them to hardware and bringing ideas to reality. I am currently working on aerial robots with non-trivial morphologies and a particular emphasis on showcasing multi-modal behaviours that enhance ground aerial locomotion. I am a recipient of the Onassis Foundation Scholarship.

Before Caltech, I earned my Master Degree in Mechanical Engineering and Robotics from ETH Zurich in 2020 and a Bachelor Degree in Mechanical Engineering from EPFL in 2018. During my time at ETH Zurich, I was fortunate to work with Professor Petros Koumoutsakos on deep reinforcement learning for control of soft swimming bodies.

I have also spent 6 months working as a Robotics Software Engineer at Anybotics (ETH Zürich spinoff) in Zurich. Here I developed novel parameter estimation algorithms to improve walking performance of the quadruped robot Anymal.

Research

Multi-Modal Robotics

Design and testing multi-modal robots, such as the Aerially Transforming Morphobot (ATMO), capable of dynamically switching between aerial and terrestrial modes.

Dynamic Ground-Aerial Morpho-Transition

Mid-air robotic transformation leads to unexpected aerodynamic effects which I analyze using aerodynamic experimental testing and rigorous theory. The findings are used to inform the control algorithms for smooth transition behaviours. Thrust Recapture

Relevant Publications

Soft Robotics

For soft robots operating in fluid media the time variation of the body shape required to produce favorable behaviours are very complex to uncover using traditional control methods. I am working on using AI based methods for learning behaviours of such systems. An example is producing the maximum possible acceleration with a limited energy expenditure in swimming fish robots.

Demo 1

Soft aerial robotics for energetically efficient multi-modal flight. Currently developing a new concept for a flexible surface controlled by thrusters. Such a surface can improve agility and crash tolerance, while reducing energetic cost of locomotion.

Relevant Publications

AI for Control

I am actively investigating how AI can be used for control in situations when the system dynamics are too complex to model and control with standard approaches, and also for robotic explorers with partial information.

Learning Navigation

Relevant Publications

Talks

  • Thrust recapture for morphing aerial vehicles with out of plane thrusters APS Division of Fluid Dynamics, 2023. [PDF]
  • Investigation of surface pressure fluctuations in airfoils subject to transverse gust: towards gust mitigation APS Division of Fluid Dynamics, 2021. [PDF]

Teaching

Media Coverage

We built a second version of ATMO (Aerially Transforming Morphobot) for our collaborators at the Technology Innovation Institute (Abu Dhabi). We demonstrated a package delivery mission in an outdoor environment. The result was covered by the TII media team. Learn more

We also presented ATMO at IROS 2024 (Abu Dhabi). The robot was covered by the national television channel Sky News Arabia. See media coverage.

© Ioannis Mandralis | Hosted on GitHub Pages

Maintained by Ioannis Mandralis. Theme customized from Minimal by Steve Smith.