News

Online talk by Dr. Claude Lacoursière: What the non-expert should know about physics simulation

Community, VERANSTALTUNGEN |

Dr. Lacoursière is a Chief Scientist at Algoryx Simulation, a company that has provided fast, modular, and high-fidelity physics engines such as AGX Dynamics. He is currently visiting MIRMI. In his talk, Dr. Lacoursière will speak about physics simulation. Algoryx and TUM are the partners of the European Union's Horizon 2020 I.AM. project (Impact Aware Manipulation by Dexterous Robot Control and Learning in Dynamic Semi-Structured Logistic Environments).

Zoom Link: tum-conf.zoom.us/j/62028706285


Meeting ID: 620 2870 6285
Passcode: 488776

 

Dr. Claude Lacoursière

Abstract: Simulation toolkits are often taken as nearly equivalent and interchangeable black boxes in some way. Numerical errors come in all shapes and forms from model limitation, time discretization errors, errors due to the specific time integration method, errors due to numerical solvers because of approximations which must be made to maintain reasonable speed.  In addition, error can be biased or not, and this affects the interpretation of the results and the choice of parameters and time step.  Various forms of errors will be explained, along with the models and numerical methods that can best suppress them or at least, make them predictable and faithful.  I will do a brief presentation of variational time integration methods to illustrate they suppress many sources of error or make them unbiased and predictable. All this plays a great role when using simulations to develop controllers and one should be judicious in the choice of simulation library.

Biography: After detours in music school and biochemistry Dr. Claude Lacoursière enrolled in the physics programme at McGill University where he obtained a Masters' degree in condensed mater phenomena, his first encounter with simulations. Eventually he joined some of his class colleagues in a startup in 1996: they wanted to inject real physical motion in 3D graphics virtual worlds.  These were the early days of consumer 3D graphics, the days of the Nintendo 64.  He ended up developing the main simulation library for multibody dynamics with frictional contacts, and applied this to heavy ground machine training simulators in 1998.

He eventually found his way back to university to complete a PhD in applied mathematics, mostly physics modeling and numerical methods for nonsmooth Lagrangian system dynamics, and indeed during that time, he finally found good methods.  He finished that in 2007, founded another software company, and continued developing nonsmooth models and methods for a variety of mechanical systems since. It's been tremendous fun and greatly rewarding for him to solve customers' problems over the years.


Receive our featured stories, news, events, and videos in your inbox every month, subscribe here!
Are you a journalist? For press inquiries, contact office@mirmi.tum.de