Join ASPE

Spring Topical 2020 – Tutorials

Spring Topical 2020 – Tutorials

AM Tutorial
Wednesday May 6
8:00 am – 12:00 pm

Introduction to Machine Learning and Reinforcement Learning for Precision Engineers
Dr. Amir Barati Farimani (Carnegie Mellon University)

In this entry-level tutorial, we will explore machine learning techniques with the goal of identifying suitable opportunities to use them in problems related to precision engineering and control engineering. The course assumes no past experience with these techniques and begins with a review of the terminology and differences and similarities between machine learning, reinforcement learning, deep learning, and artificial intelligence tools. With the nomenclature demystified, we will next focus on identifying problems for which these tools can provide a benefit. Some hands-on tutorials with Jupyter Notebooks will be overviewed and practiced.  In the second part of the tutorial, we will cover reinforcement learning basics with some classical control examples such as Cart pole, Inverted Pendulum, etc. An introduction to Open AI Gym environment will be given. Finally, we will review case studies of interest to the precision engineers and control with more focus on RL.

Bio:

Professor Farimani joined the Department of Mechanical Engineering at Carnegie Mellon University in the fall of 2018. He was previously a postdoctoral fellow at Stanford University. He received his PhD in Mechanical Engineering in 2015. His lab at CMU focuses on the problems at the interface of Mechanical Engineering, data science and machine learning. His lab uses the state of the art deep learning and machine learning algorithms and tools to learn, infer and predict the physical phenomena pertinent to mechanical engineering. Currently, he is teaching AI and ML to a large class of graduate students at CMU. He received the Stanley I. Weiss best thesis award from the University of Illinois in 2016 and was recognized as an Outstanding Graduate Student in 2015. During his post-doctoral fellowship at Stanford, Dr. Barati Farimani has developed data-driven, deep learning techniques for inferring, modeling, and simulating the physics of transport phenomena and for materials discovery for energy harvesting applications. He taught multiple workshops on machine learning and artificial intelligence for both industry and academia such as  Autodesk (ai4engineering.com) and precision engineering society.

 

PM Tutorial
Wednesday May 6
1:30 pm – 5:30 pm

Advanced Feedforward and Iterative Learning Control for Precision Mechatronics
Dr. Tom Oomen (Eindhoven University of Technology)

Do you also have a system that has the same error in each task? In this tutorial, we investigate advanced feedforward and iterative learning control (ILC) algorithms to improve the performance of your system. The tutorial consists of several topics.

First, traditional ILC algorithms can learn from measured error signals of previous tasks. After just a few iterations, such algorithms can compensate any repeating disturbance perfectly. We will first explore which applications are suitable for ILC, and a very simple test allows you to immediately determine the achievable ILC performance of your system. A complete design framework that resembles standard loop-shaping control techniques is presented, which guarantees fast and safe learning. Essential technical aspects such as robust and monotonic convergence as well as non-causal filtering are explained in an intuitive and directly usable manner. An overview of optimization-based ILC techniques is outlined, as well as their advantages and disadvantages.

Second, advanced feedforward is investigated. Indeed, ILC algorithms achieve exceptional performance for repeating tasks, however, they are far from standard in industrial practice. One of the key reasons is that iterative learning control cannot deal with varying tasks. Even a small variation can lead to disastrous performance deterioration. Many industrial systems perform such very similar yet slightly different tasks, necessitating new concepts for advanced feedforward control, including high-order feedforward (snap, jerk, etc.), input shaping, automated tuning, etc. These new concepts have been developed in recent years, whereas most of these ‘iterative learning control’ (ILC) techniques have been developed in the past two decades and many successful industrial applications have been reported, in particular in precision mechatronics such as printing systems, additive manufacturing, wafer scanners, pick-and-place machines, etc.

Throughout, extensions of the design framework to multivariable systems are outlined, as well as several recent connections to machine learning techniques. The tutorial is of interest to anyone interested in improving the performance of their control system. It is aimed to attract a broad audience: no prior knowledge of learning control is required and only a basic knowledge of feedforward/feedback control (including block diagrams, frequency domain, etc.) is assumed. The tutorial will also be of interest to experts in the field, since it covers recently developed techniques as well.

Bio:

Tom Oomen received the M.Sc. degree (cum laude) and Ph.D. degree from the Eindhoven University of Technology, Eindhoven, The Netherlands. He held visiting positions at KTH, Stockholm, Sweden, and at The University of Newcastle, Australia. Presently, he is associate professor with the Department of Mechanical Engineering at the Eindhoven University of Technology. He is a recipient of the Corus Young Talent Graduation Award, the IFAC 2019 TC 4.2 Mechatronics Young Research Award, the 2015 IEEE Transactions on Control Systems Technology Outstanding Paper Award, the 2017 IFAC Mechatronics Best Paper Award, the 2019 IEEJ Journal of Industry Applications Best Paper Award, and recipient of a Veni and Vidi personal grant. He is Associate Editor of the IEEE Control Systems Letters (L-CSS), IFAC Mechatronics, and IEEE Transactions on Control Systems Technology. He is a member of the Eindhoven Young Academy of Engineering. His research interests are in the field of data-driven modeling, learning, and control, with applications in precision mechatronics.