dc.contributor.advisor | Lekkas, Anastasios | |
dc.contributor.author | Vagle, Anders Haver | |
dc.date.accessioned | 2019-10-31T15:12:30Z | |
dc.date.issued | 2019 | |
dc.identifier | no.ntnu:inspera:35771502:14879818 | |
dc.identifier.uri | http://hdl.handle.net/11250/2625749 | |
dc.description | Full text not available | |
dc.description.abstract | Denne oppgaven tar for seg detaljer rundt implementasjon av PPO-algoritme for trening på egendefinerte miljøer designet for robotikk-basert manipulasjon. Resultatene er
lovende for de forenklede miljøene i simulering, men fungerer dårlig i den virkelige verden.
Simulering av ROS-implementerte roboter i Gazebo viser seg å være en treg prosess, og
sannsynligvis lite egnet for stor-skala operasjoner med mål om applikasjon i et virkelig
miljø. | |
dc.description.abstract | This thesis present the implementation details of how the PPO algorithm was used to
train on custom environments designed for robotic manipulation. The results are promising in the simulated environments, but transfer to the real-world yields generally weak
performance. Simulation of ROS implemented robots in Gazebo proves to be a very slow
process, and likely not suitable for large-scale tasks with goals of real-world application. | |
dc.language | eng | |
dc.publisher | NTNU | |
dc.title | Reinforcement Learning for Robotic Manipulation | |
dc.type | Master thesis | |