Introduction to reinforcement learning and control theory

Introduction to reinforcement learning and control theory#

Hint

Use ctrl+k to search.

This page contains material and information related to the spring 2024 version of the course Introduction to reinforcement learning and control, offered at DTU.

If you are curious about the course, you can read more about the course here. If you are enrolled and just starting out, you should begin with the Installation. You can find the exercises, project descriptions in the menu to the left.

Practicalities#

Note

This page is automatically updated with typos, etc. I therefore recommend bookmarking it and using the newest version of the exercises.

Time and place:: Building B341, auditorium 21, 08:00–12:00
DTU Learn:: 02465
Exercise code:: https://lab.compute.dtu.dk/02465material/02465students.git
Course descriptions:: kurser.dtu.dk
Lecture recordings:: panopto.dtu.dk
Discord:: Discord channel (invitation link)
Campus-wide python support:: pythonsupport.dtu.dk
Contact:: Tue Herlau, tuhe@dtu.dk.

Course schedule#

The schedule and reading can be found below. Click on the titles to read the exercise and project descriptions.

#	Date	Title	Reading	Homework	Exercise	Slides
	Jan 26th, 2024	Installation and self-test	Chapter 1-3 , [Her24]		[PDF]
1	Feb 2nd, 2024	The finite-horizon decision problem	Chapter 4, [Her24]	1, 2	[PDF]	[1x] [6x]
2	Feb 9th, 2024	Dynamical Programming	Chapter 5-6.2, [Her24]	1, 2	[PDF]	[1x] [6x]
3	Feb 16th, 2024	DP reformulations and introduction to Control	Section 6.3; Chapter 10-11, [Her24]	1, 2	[PDF]	[1x] [6x]
4	Feb 23th, 2024	Discretization and PID control	Chapter 12-14, [Her24]	1, 2	[PDF]	[1x] [6x]
	Feb 29th, 2024	Project 1: Dynamical Programming
5	Mar 1st, 2024	Direct methods and control by optimization	Chapter 15, [Her24]	1	[PDF]	[1x] [6x]
6	Mar 8th, 2024	Linear-quadratic problems in control	Chapter 16, [Her24]	1	[PDF]	[1x] [6x]
7	Mar 15th, 2024	Linearization and iterative LQR	Chapter 17, [Her24]	1	[PDF]	[1x] [6x]
8	Mar 22th, 2024	Exploration and Bandits	Chapter 1; Chapter 2-2.7; 2.9-2.10, [SB18]	1	[PDF]	[1x] [6x]
	Apr 4th, 2024	Project 2: Control theory
9	Apr 5th, 2024	Policy and value iteration	Chapter 3; 4, [SB18]	1, 2	[PDF]	[1x] [6x]
10	Apr 12th, 2024	Monte-carlo methods and TD learning	Chapter 5-5.4+5.10; 6-6.3, [SB18]	1	[PDF]	[1x] [6x]
11	Apr 19th, 2024	Model-Free Control with tabular and linear methods	Chapter 6.4-6.5; 7-7.2; 9-9.3; 10.1, [SB18]	1	[PDF]	[1x] [6x]
12	Apr 26th, 2024	Eligibility traces and value-function approximations	Chapter 10.2; 12-12.7, [SB18]	1	[PDF]	[1x] [6x]
	May 2nd, 2024	Project 3: Reinforcement Learning
13	May 3rd, 2024	Q-learning and deep-Q learning	Chapter 6.7-6.9; 8-8.4; 16-16.2; 16.5; 16.6, [SB18]	1	[PDF]	[1x] [6x]

The reading material is available here:

[Her24]:: 02465_Notes.pdf
[SB18]:: Introduction to Reinforcement Learning (2020) (Authors homepage)

You can find the exam QA slides here. Details about the exam QA session will be announced on DTU Learn.

Note

Chapters 1–3 is background information about python and are therefore not part of the main course content (pensum). Knowledge of python is required for the exams.
The Homework column list those problems that will be discussed during class. They are also indicated by a in the margin of the PDF file. I encourage you to prepare them at home and present your solution during the exercise session.

Exercise sessions#

Hint

I will upload solutions to some of the python problems on gitlab.

The teaching assistants will be available Fridays 10:00–12:00 after the lecture.

Location	Instructor	Email
Building B341, IT-019	Calle Ryge Carlsen	s204114@student.dtu.dk
Building B341, IT-015	Karl Meisner-Jensen	s204106@student.dtu.dk
Building B341, IT-015	Victor Tolsager Olesen	s204141@student.dtu.dk
Building B341, auditorium 21	Tue Herlau	tuhe@dtu.dk

For the exercises, you are encouraged to prepare the homework problems at home (see syllabus above), and present your solution during the exercise session.

Additional reading material#

The material below is referenced in the course but it is not part of the course syllabus.

Reference name	Download
(Tassa et al. [TET12])	tassa2012.pdf
(Kelly [Kel17])	kelly2017.pdf

Contents#

Indices and tables#

Bibliography#

[Her24] (1,2,3,4,5,6,7,8)

Tue Herlau. Sequential decision making. (Freely available online), 2024.

[Kel17]

Matthew Kelly. An introduction to trajectory optimization: how to do your own direct collocation. SIAM Review, 59(4):849–904, 2017. (See kelly2017.pdf). URL: https://epubs.siam.org/doi/pdf/10.1137/16M1062569.

[SB18] (1,2,3,4,5,6)

Richard S. Sutton and Andrew G. Barto. Reinforcement Learning: An Introduction. The MIT Press, second edition, 2018. (Freely available online). URL: http://incompleteideas.net/book/the-book-2nd.html.

[TET12]

Yuval Tassa, Tom Erez, and Emanuel Todorov. Synthesis and stabilization of complex behaviors through online trajectory optimization. In 2012 IEEE/RSJ International Conference on Intelligent Robots and Systems, 4906–4913. IEEE, 2012. (See tassa2012.pdf). URL: https://ieeexplore.ieee.org/abstract/document/6386025.

This page was last updated at:

>>> from datetime import datetime
>>> print("Document updated at:", datetime.now().strftime("%d/%m/%Y %H:%M:%S"))
Document updated at: 22/08/2024 12:09:12