Few days ago, thanks to my supervisor, I had a great opportunity to participate in top quality spring school on topic Theory and Numerics for Nonlinear Model Predictive Control. The school was organised by Department of Microsystems Engineering – IMTEK University of Freiburg, Germany. The course was a part of international TEMPO program for PhD students, hosting 80 participants from 23 countries.
Here I would like to share with you, my experience from this 9 day intensive course.
The course was divided into two parts, each led by two extraordinary lecturers. First part was about Optimization, Optimal Control, and Numerical Methods led by prof. Moritz Diehl.
Second part was mainly about Nonlinear Model Predictive Control, and Moving Horizon Estimation led by famous prof. James B. Rawlings.
Very first day of the course was optional and devoted to short introduction of programming in Python, basic knowledge of which was necessary for using symbolic framework CasADi for algorithmic differentiation and numeric optimization, which was used during following excercises. I have to say that I really enjoyed it even if my previous experience with Python was very poor. (You know, initial day enthusiasm…)
The structure of the course was divided into lectures and excercises, wich were subsequently and periodically changing every few hours, to keep us focused on the topic.
And as I said this one was realy intense course, with huge ammount of information presented in fast mode, many of which I heard very first time. Yet the lectures itself did not lack consistency and their meaning was not to make us experts in 9 days, rather to provide comprehensive overview of the topic.
On the other hand, regarding the exercises part I have to be honets, that I like them less and less each day. My critics is based on the rapid accelerating pattern of difficulty for excercises, which were hard to follow in short time ranges given for the solution. Especially in second week the difficulty of excercise was so high, that even the PhD students leading the excercise were uable to finish and explain the solution in given time, leading to increased level of frustration among participants.
This issues were caused In my opinion mainly due to
less precise preparation of organizers for exercises due to lack of time resulting from extremely dense content of the course.
So as in this case, sometimes is less more.
However even this critics has aslo a positive point of view, which are provided solutions of mentioned difficult excercises as complex examples of how to use CasADi as a powerful tool for nonlinear MPC.
All materials, i.e. slides and recored lectures as well as excercises, their solutions and also final projects of the participants can be found online here.
For summary I was comming home bit tired, saturated with lot of information for post processing in following days, weeks, and months… Yet very satisfied with lectures and meeting new interesting people from the field of my proffesional interest.
0 Comments