Wir verwenden Google für unsere Suche. Mit Klick auf „Suche aktivieren“ aktivieren Sie das Suchfeld und akzeptieren die Nutzungsbedingungen.

Multi-Criteria Optimization and Decision Analysis for Embedded Systems Design

Lecturer (assistant) Nguyen Anh Vu Doan 2 SWS Wintersemester 2023/24 English

See TUMonline
Note: Registration via TUMonline starting on 25th September 2023.

Objectives

Upon successful completion of this module, students are able to: - understand the multi-criteria paradigm and its challenges for embedded systems design, - analyze and model encountered problems with this paradigm, - understand how different (multi-objective) optimization methods work, select and apply the most suitable one(s) depending on the situation, - understand how different (multi-criteria) decision making methods work, select and apply the most suitable one(s), as well as analyze the results obtained after the optimization process.

Description

Content of the lecture 1. Introduction to the multi-criteria paradigm for embedded systems design - Uni-criterion vs multi-criteria - Modeling and challenges 2. Optimization methods - Linear programming - Metaheuristics (e.g. genetic algorithms, simulated annealing) - Multi-objective optimization for design space exploration 3. Decision making processes - Voting theory - Multi-criteria decision analysis - Game theory - Decision under risk and uncertainty During the lecture, the theoretical content will be accompanied by examples illustrating the following concepts: problem abstraction and modeling, algorithm selection and implementation, multi-criteria decision making and analysis. Thereby both functional and non-functional aspects will be considered. More elaborate exercises on these topics will be done in self-study by the students.

Prerequisites

- Data structures - Basic programming skills in Python or Matlab; alternatively C/C++ or Java - Basic knowledge of probability and statistics (probability axioms and theorems, e.g. Bayes' Theorem and its applications; typical probability distributions, e.g. exponential, Gaussian, etc.)

Teaching and learning methods

The technical content will be introduced by means of lectures with PowerPoint presentations and will be illustrated with small examples that will be included in the slides. The students are encouraged to ask questions. In addition to the individual learning methods of the students, the transfer of the theoretical knowledge to its practical application will be achieved through illustrative examples during the lectures and additional exercises to be done in self study manner. All the course material will be made available to the students through Moodle.

Examination

Written examination (60 min.): The students will be examined through a written exam where they prove that they have understood the application of the multi-criteria paradigm and can apply it to perform analysis, modeling, optimization, and decision making for problems encountered in embedded systems design. The questions will cover the theoretical background presented during the lectures as well as exercises from the lecture. The exam lasts 60 minutes and will be carried out without helping material. Homework: As part of the self-study time, a homework in groups of 2-3 participants will be assessed, where the students have to demonstrate that they can solve real world optimization problems coming from a current research area. Given the size of such problems this cannot be covered in the written exam. The assessment process of this part will be carried out through deliverables and a report. The final grade is the weighted average of the written exam (60%) and the homework (40%).

Recommended literature

Optional literature recommendations: - XS. Yangi, &quot;Engineering Optimization: An Introduction with Metaheuristic Applications&quot;, Wiley 2010 - EG. Talbi, &quot;Metaheuristics: From Design to Implementation&quot;, Wiley 2009 - S. Greco, M. Ehrgott, J.R. Figueira (Eds.), &quot;Multiple Criteria Decision Analysis: State of the Art Surveys&quot;, Springer 2016