Change search
CiteExportLink to record
Permanent link

Direct link
Cite
Citation style
  • apa
  • ieee
  • modern-language-association-8th-edition
  • vancouver
  • Other style
More styles
Language
  • de-DE
  • en-GB
  • en-US
  • fi-FI
  • nn-NO
  • nn-NB
  • sv-SE
  • Other locale
More languages
Output format
  • html
  • text
  • asciidoc
  • rtf
Kontroll av inverterad pendel med inlärning
KTH, School of Engineering Sciences (SCI).
KTH, School of Engineering Sciences (SCI).
2019 (Swedish)Independent thesis Basic level (degree of Bachelor), 10 credits / 15 HE creditsStudent thesisAlternative title
Control of inverted pendelum by learning (English)
Abstract [sv]

Den inverterade pendeln fäst på en vagn som kan röra sig längs en riktning, även kallad CartPole, anses vara det mest grundläggande exemplet inom reglerteknik och optimering. I denna rapport var målet att försöka lösa detta instabila, olinjära stabiliseringsproblemet med hjälp av maskininlärning, mer specifikt med ett artificiellt neuronnät. I början av varje simulation ändrades vagnens massa samt pendelns massa och längd slumpmässigt. Dessutom undersöktes hur robust modellen var, genom att ta reda på hur långt man kunde ändra CartPole parametrarna utanför de gränser som modellen har tränats för. Detta uppnåddes genom att göra tusentals simuleringar, allt kodat i Python med paket som TFLearn, Tensorflow och gym från openAI. Resultaten visade att det faktiskt är möjligt att kontrollera CartPole-systemet med hjälp av ett neuronnät, samt visade att det är en robust modell som klarar av nya CartPole parametrar som går långt utanför det område som neuronnätet hade tränats för.

 

Abstract [en]

The inverted pendulum attached to a cart moving in one dimension, also called a CartPole, is considered to be the most fundamental example on control theory and optimization. In this report the goal was to try and solve the unstable, non-linear stabilization problem using machine learning, more specifically using artificial neural networks. Using this to control the CartPole to be able to swing up and stabilize the pendulum where the mass of the cart as well as the mass and length of the pendulum was set randomly at the start of each simulation. Furthermore, to find out how far it is possible to push the artificial neural network outside of the limits of training and still get a stable result. This was accomplished by doing thousands of simulations, coded in Python using packages like TFLearn, Tensorflow and the gym environment from openAI. The results showed that it is in fact possible to control the CartPole system using artificial neural networks, also showing that it is robust against new CartPole parameters going far outside of what the network had been trained for.

 

Place, publisher, year, edition, pages
2019.
Series
TRITA-SCI-GRU ; 2019:192
National Category
Engineering and Technology
Identifiers
URN: urn:nbn:se:kth:diva-254831OAI: oai:DiVA.org:kth-254831DiVA, id: diva2:1335510
Supervisors
Examiners
Available from: 2019-07-05 Created: 2019-07-05 Last updated: 2019-07-05Bibliographically approved

Open Access in DiVA

fulltext(3176 kB)8 downloads
File information
File name FULLTEXT01.pdfFile size 3176 kBChecksum SHA-512
3216d597aaa8228a539a2708c2092bb3f9ba9c420424a2f77d7253a74ad51bea14a60b3fea0f83b42bcd44c243b06b34b15e7846036c6deb5bdc54f5f371129c
Type fulltextMimetype application/pdf

By organisation
School of Engineering Sciences (SCI)
Engineering and Technology

Search outside of DiVA

GoogleGoogle Scholar
Total: 8 downloads
The number of downloads is the sum of all downloads of full texts. It may include eg previous versions that are now no longer available

urn-nbn

Altmetric score

urn-nbn
Total: 24 hits
CiteExportLink to record
Permanent link

Direct link
Cite
Citation style
  • apa
  • ieee
  • modern-language-association-8th-edition
  • vancouver
  • Other style
More styles
Language
  • de-DE
  • en-GB
  • en-US
  • fi-FI
  • nn-NO
  • nn-NB
  • sv-SE
  • Other locale
More languages
Output format
  • html
  • text
  • asciidoc
  • rtf