Scientific Machine Learning Using Functional Mock-Up Units - JuliaCon 2024

Explore the integration of physical system knowledge into machine learning processes through a hands-on workshop focused on modeling a robot capable of writing messages with a pen. Learn about Functional Mock-Up Units (FMUs) and their incorporation into machine learning topologies to create NeuralFMUs. Dive into a challenging robotics use case involving a Selective Compliance Assembly Robot Arm (SCARA). Examine the physical simulation model, understand model deviations, and address complex phenomena like slip-stick-friction using hybrid modeling techniques. Participate in coding sessions, explore notebook designs, and gain practical experience in designing and training hybrid models. Discover the advantages and challenges of hybrid modeling compared to traditional physical modeling and pure machine learning approaches. Understand the connection between physical modeling and machine learning worlds, learn about NeuralODEs and NeuralFMUs, and explore the impact of signal choice on training success and computational performance. Acquire the knowledge and skills necessary to tackle hybrid modeling applications in your own field.