A number of lectures will be given connected with homework problems. Finally, each student will carry out a small project in a simulation environment or real robot system.
FDD3025 Introduction to Behavior Trees in Robotics and AI 3.0 credits
A behavior tree (BT) is a way of creating an overall robot controller from a set oflow level controllers doing separate things, such as: Goto position X, Grasp object, Place object, Open door, Say X. BTs have been shown to be optimally modular, and well suited for creating designs that are both reactive and goal driven. In this course we will learn how BTs work, and in what sense they are modular, reactive and goal directed. We will also see how BTs can be combined with classical control design as well as learning based methods. Finally we see how guarantees on performance, in terms of safety and goal convergence can be made for BTs.
Information for research students about course offerings
Contact the examiner for information on course rounds.
About course offering
For course offering
Spring 2024 Start 18 Mar 2024 programme students
Target group
No information insertedPart of programme
No information insertedPeriods
P4 (3.0 hp)Duration
Pace of study
17%
Form of study
Normal Daytime
Language of instruction
English
Course location
KTH Campus
Number of places
Places are not limited
Planned modular schedule
Course memo
Course memo is not publishedSchedule
Schedule is not publishedApplication
For course offering
Spring 2024 Start 18 Mar 2024 programme students
Application code
61082
Contact
For course offering
Spring 2024 Start 18 Mar 2024 programme students
Contact
Petter Ögren
Examiner
No information insertedCourse coordinator
No information insertedTeachers
No information insertedContent and learning outcomes
Course disposition
Course contents
BT design principles. Reactivity, modularity and goal directedness of BTs. BTs and classical control methods, BTs and reinforcement learning. How BTs can be used to guarantee properties such as safety and goal convergence.
Intended learning outcomes
Upon completion the students will:
- Know how to use a BT to design the controller of a robot or artificial agent
- Know the advantages of BTs in terms of reactivity, modularity and goal directedness
- Know several design principles for BTs
- Know how to connect BTs with classical control such as Control Barrier Functions
- Know how to connect BTs with Reinforcement Learning
- Know how BTs can be used to prove performance in terms of safety and reaching a set of given goal states
Literature and preparations
Specific prerequisites
None
Recommended prerequisites
What is needed to start a PhD program in computer science, electrical engineering or similarly.
Equipment
No equipment needed beyond a standard computer.
Literature
The book: Behavior Trees for Robotics and AI, by Colledanchise and Ögren (available on ArXiv).
Research papers.
Online lectures on Youtube.
Examination and completion
If the course is discontinued, students may request to be examined during the following two academic years.
Grading scale
Examination
- EXA1 - Examination, 3.0 credits, grading scale: P, F
Based on recommendation from KTH’s coordinator for disabilities, the examiner will decide how to adapt an examination for students with documented disability.
The examiner may apply another examination format when re-examining individual students.
Examination will be in the form of homework sets and a small final project.
Opportunity to complete the requirements via supplementary examination
Opportunity to raise an approved grade via renewed examination
Examiner
Ethical approach
- All members of a group are responsible for the group's work.
- In any assessment, every student shall honestly disclose any help received and sources used.
- In an oral assessment, every student shall be able to present and answer questions about the entire assignment and solution.