• Refine Query
  • Source
  • Publication year
  • to
  • Language
  • 1
  • Tagged with
  • 1
  • 1
  • 1
  • 1
  • 1
  • 1
  • 1
  • 1
  • 1
  • 1
  • 1
  • 1
  • 1
  • 1
  • 1
  • About
  • The Global ETD Search service is a free service for researchers to find electronic theses and dissertations. This service is provided by the Networked Digital Library of Theses and Dissertations.
    Our metadata is collected from universities around the world. If you manage a university/consortium/country archive and want to be added, details can be found on the NDLTD website.
1

Stuck state avoidance through PID estimation training of Q-learning agent / Förhindrande av odefinierade tillstånd vid Q-learning träning genom PID estimering

Moritz, Johan, Winkelmann, Albin January 2019 (has links)
Reinforcement learning is conceptually based on an agent learning through interaction with its environment. This trial-and-error learning method makes the process prone to situations in which the agent is stuck in a dead-end, from which it cannot keep learning. This thesis studies a method to diminish the risk that a wheeled inverted pendulum, or WIP, falls over during training by having a Qlearning based agent estimate a PID controller before training it on the balance problem. We show that our approach is equally stable compared to a Q-learning agent without estimation training, while having the WIP falling over less than half the number of times during training. Both agents succeeds in balancing the WIP for a full hour in repeated tests. / Reinforcement learning baseras på en agent som lär sig genom att interagera med sin omgivning. Denna inlärningsmetod kan göra att agenten hamnar i situationer där den fastnar och inte kan fortsätta träningen. I denna examensuppsats utforskas en metod för att minska risken att en självkörande robot faller under inlärning. Detta görs genom att en Q-learning agent tränas till att estimera en PID kontroller innan den tränar på balanseringsproblemet. Vi visar att vår metod är likvärdigt stabil jämfört med en Q-learning agent utan estimeringsträning. Under träning faller roboten färre än hälften så många gånger när den kontrolleras av vår metod. Båda agenterna lyckas balansera roboten under en hel timme.

Page generated in 0.0954 seconds