ASE-7516 Dynamic Planning with Incomplete Information, 5 cr
Additional information
The course will be lectured in every second year.
Suitable for postgraduate studies.
Person responsible
Risto Ritala
Lessons
Implementation | Period | Person responsible | Requirements |
ASE-7516 2019-01 | 1 - 2 |
Risto Ritala |
Exam and exercise work. |
Learning Outcomes
This course expresses the planning problem as discrete-time dynamic programming with both action and information channel degrees of freedom. Problems studied include both discrete state and continuous state systems. Is capable (grade 3/5) 1. To formulate the control of the system and the selection of the information channels as a partially observable Markov decision process (POMDP) problem. 2. To produce approximate solutions to POMDP problems and to assess their suboptimality. 3. To explain the opportunities of approximate dynamic programming with POMDPs when system models are uncertain.
Content
Content | Core content | Complementary knowledge | Specialist knowledge |
1. | Concurrent planning action on system and selection of information channels. | Q-learning and model-free planning | |
2. | Partially observable Markov decision process (POMDP). | ||
3. | Solving dynamic programming problem for linear-quadratic-Gaussian POMDP. | ||
4. | Solving dynamic programming problem for discrete state POMDP | ||
5. | Dealing with model uncertainties by applying approximated dynamic programming methods. |
Instructions for students on how to achieve the learning outcomes
Exam and exercise work.
Assessment scale:
Numerical evaluation scale (0-5)
Study material
Type | Name | Author | ISBN | URL | Additional information | Examination material |
Lecture slides | No |
Prerequisites
Course | Mandatory/Advisable | Description |
ASE-4080 Systeemien hallinnan case study | Advisable | |
IHA-4306 Fundamentals of Mobile Robots | Advisable |
Correspondence of content
Course | Corresponds course | Description |
ASE-7516 Dynamic Planning with Incomplete Information, 5 cr | ASE-7516 Dynamic Planning with Incomplete Information, 5 cr |