Machine Learning and System Identification for Estimation in Physical Systems
(2018) In PhD Thesis TFRT-1122- Abstract
- In this thesis, we draw inspiration from both classical system identification and modern machine learning in order to solve estimation problems for real-world, physical systems. The main approach to estimation and learning adopted is optimization based. Concepts such as regularization will be utilized for encoding of prior knowledge and basis-function expansions will be used to add nonlinear modeling power while keeping data requirements practical.
The thesis covers a wide range of applications, many inspired by applications within robotics, but also extending outside this already wide field.
Usage of the proposed methods and algorithms are in many cases illustrated in the real-world applications that motivated the... (More) - In this thesis, we draw inspiration from both classical system identification and modern machine learning in order to solve estimation problems for real-world, physical systems. The main approach to estimation and learning adopted is optimization based. Concepts such as regularization will be utilized for encoding of prior knowledge and basis-function expansions will be used to add nonlinear modeling power while keeping data requirements practical.
The thesis covers a wide range of applications, many inspired by applications within robotics, but also extending outside this already wide field.
Usage of the proposed methods and algorithms are in many cases illustrated in the real-world applications that motivated the research.
Topics covered include dynamics modeling and estimation, model-based reinforcement learning, spectral estimation, friction modeling and state estimation and calibration in robotic machining.
In the work on modeling and identification of dynamics, we develop regularization strategies that allow us to incorporate prior domain knowledge into flexible, overparameterized models. We make use of classical control theory to gain insight into training and regularization while using tools from modern deep learning. A particular focus of the work is to allow use of modern methods in scenarios where gathering data is associated with a high cost.
In the robotics-inspired parts of the thesis, we develop methods that are practically motivated and make sure that they are implementable also outside the research setting. We demonstrate this by performing experiments in realistic settings and providing open-source implementations of all proposed methods and algorithms.
(Less)
Please use this url to cite or link to this publication:
https://lup.lub.lu.se/record/ffb8dc85-ce12-4f75-8f2b-0881e492f6c0
- author
- Bagge Carlson, Fredrik LU
- supervisor
- opponent
-
- Professor Kjellström, Hedvig, Royal Institute of Technology, Stockholm, Sweden
- organization
- publishing date
- 2018-12-20
- type
- Thesis
- publication status
- published
- subject
- in
- PhD Thesis TFRT-1122
- pages
- 184 pages
- publisher
- Department of Automatic Control, Faculty of Engineering LTH, Lund University
- defense location
- Lecture hall B, building M, Ole Römers väg 1, Lund University, Faculty of Engineering LTH, Lund
- defense date
- 2019-01-18 10:15:00
- ISSN
- 0280-5316
- ISBN
- 978-91-7753-921-6
- 978-91-7753-920-9
- project
- SARAFun—Smart Assembly Robot with Advanced FUNctionalities
- Flexifab
- language
- English
- LU publication?
- yes
- id
- ffb8dc85-ce12-4f75-8f2b-0881e492f6c0
- date added to LUP
- 2018-12-11 16:11:51
- date last changed
- 2019-02-21 08:43:36
@phdthesis{ffb8dc85-ce12-4f75-8f2b-0881e492f6c0, abstract = {{In this thesis, we draw inspiration from both classical system identification and modern machine learning in order to solve estimation problems for real-world, physical systems. The main approach to estimation and learning adopted is optimization based. Concepts such as regularization will be utilized for encoding of prior knowledge and basis-function expansions will be used to add nonlinear modeling power while keeping data requirements practical.<br/><br/>The thesis covers a wide range of applications, many inspired by applications within robotics, but also extending outside this already wide field.<br/>Usage of the proposed methods and algorithms are in many cases illustrated in the real-world applications that motivated the research.<br/>Topics covered include dynamics modeling and estimation, model-based reinforcement learning, spectral estimation, friction modeling and state estimation and calibration in robotic machining.<br/><br/>In the work on modeling and identification of dynamics, we develop regularization strategies that allow us to incorporate prior domain knowledge into flexible, overparameterized models. We make use of classical control theory to gain insight into training and regularization while using tools from modern deep learning. A particular focus of the work is to allow use of modern methods in scenarios where gathering data is associated with a high cost.<br/><br/>In the robotics-inspired parts of the thesis, we develop methods that are practically motivated and make sure that they are implementable also outside the research setting. We demonstrate this by performing experiments in realistic settings and providing open-source implementations of all proposed methods and algorithms.<br/>}}, author = {{Bagge Carlson, Fredrik}}, isbn = {{978-91-7753-921-6}}, issn = {{0280-5316}}, language = {{eng}}, month = {{12}}, publisher = {{Department of Automatic Control, Faculty of Engineering LTH, Lund University}}, school = {{Lund University}}, series = {{PhD Thesis TFRT-1122}}, title = {{Machine Learning and System Identification for Estimation in Physical Systems}}, url = {{https://lup.lub.lu.se/search/files/55399444/phdthesis.pdf}}, year = {{2018}}, }