DIKU Bits: Fast Physics for Robotics and Games

This DIKU Bits lecture will be held online via Zoom. To participate, find the link at https://di.ku.dk/begivenhedsmappe/begivenheder-2020/diku-bits-mihail-francu/.

Speaker

Mihail Francu, Postdoc in the Image Analysis, Computational Modelling and Geometry Section at DIKU.

Abstract

I have worked on physics simulation for more than a decade both as a hobby and professionally at companies like Havok, Crytek and Ubisoft. Game physics engines deliver lightning-fast simulations of rigid bodies and cloth for some of the best-selling AAA titles. The challenge now is to take that technology and apply it to the emerging fields of soft robotics and machine learning. Deformable squashy objects are a challenge to simulate efficiently because of the curse of dimensionality, but multi-core CPUs and GPUs are coming to the rescue. Real-time simulation and faster than that are required by reinforcement learning in robotics. Will we be seeing more realistic physical simulations in games and VR driven by this robotics revolution? Come to my presentation and see what the current state of the art is and what I and others are currently researching.

Zooming in on Mihail Francu

Which courses do you teach (BSc and MSc)?
I  do not teach any courses as I am exempt, being a Marie Curie Postdoc.

Which technology/research/project/startups are you excited to follow the development of?
I’m particularly interested in simulation technologies like Nvidia Flex or open-source frameworks like SOFA or Chrono as they are particularly close to my field of expertise.

What is your favorite sketch from the DIKUrevy?
I’m sorry, I have not been aware of DIKUrevy for the past year and a bit of being here at DIKU.