Posilované učení pro ovládání čtyřnohého robota s inovativní kinematikou
Reinforcement Learning for Quadrupedal Robot Control with Novel Kinematics
dc.contributor.advisor | Marko Rastislav | |
dc.contributor.author | Andrej Kružliak | |
dc.date.accessioned | 2024-01-23T23:52:19Z | |
dc.date.available | 2024-01-23T23:52:19Z | |
dc.date.issued | 2024-01-23 | |
dc.identifier | KOS-1240440828205 | |
dc.identifier.uri | http://hdl.handle.net/10467/113298 | |
dc.description.abstract | This thesis aims to implement and assess an innovative kinematics solution of the quadruped robotic platform Artaban by Panza Robotics, in a simulation environment within a deep reinforcement learning setting. The innovative kinematics include two distinct parallel mechanism types. One transfers the torque from the motor through a universal joint mechanism called Cardan mechanism, and the other is a four-link mechanism on the rear legs. The innovative kinematics are implemented and tested in three kinematic configurations: the original with distinct front and rear legs, a uniform front-legged configuration, and a front-legged Cardan configuration optimizing the transmission of torque. The Cost of Transport (CoT) metric, judging the amount of effort per robot velocity, is used for assessing the performance of gaits produced by those configurations. The original configuration encountered simulation issues due to improper loop-closure, leading to non-viable gaits. These challenges were absent in the front-legged configuration, which successfully generated a valid gait. The subsequent implementation of the Cardan mechanism yielded an even more efficient and visually pleasing gait. The Cardan mechanism's gait was identified as the most efficient from the point of view of the CoT metric, highlighting the mechanism's potential for enhancing robotic locomotion. The results obtained from the front-legged configuration with the Cardan mechanism are expected to translate effectively to the original kinematic configuration once a stable simulation of the rear legs is achieved. | cze |
dc.description.abstract | This thesis aims to implement and assess an innovative kinematics solution of the quadruped robotic platform Artaban by Panza Robotics, in a simulation environment within a deep reinforcement learning setting. The innovative kinematics include two distinct parallel mechanism types. One transfers the torque from the motor through a universal joint mechanism called Cardan mechanism, and the other is a four-link mechanism on the rear legs. The innovative kinematics are implemented and tested in three kinematic configurations: the original with distinct front and rear legs, a uniform front-legged configuration, and a front-legged Cardan configuration optimizing the transmission of torque. The Cost of Transport (CoT) metric, judging the amount of effort per robot velocity, is used for assessing the performance of gaits produced by those configurations. The original configuration encountered simulation issues due to improper loop-closure, leading to non-viable gaits. These challenges were absent in the front-legged configuration, which successfully generated a valid gait. The subsequent implementation of the Cardan mechanism yielded an even more efficient and visually pleasing gait. The Cardan mechanism's gait was identified as the most efficient from the point of view of the CoT metric, highlighting the mechanism's potential for enhancing robotic locomotion. The results obtained from the front-legged configuration with the Cardan mechanism are expected to translate effectively to the original kinematic configuration once a stable simulation of the rear legs is achieved. | eng |
dc.publisher | České vysoké učení technické v Praze. Vypočetní a informační centrum. | cze |
dc.publisher | Czech Technical University in Prague. Computing and Information Centre. | eng |
dc.rights | A university thesis is a work protected by the Copyright Act. Extracts, copies and transcripts of the thesis are allowed for personal use only and at one?s own expense. The use of thesis should be in compliance with the Copyright Act http://www.mkcr.cz/assets/autorske-pravo/01-3982006.pdf and the citation ethics http://knihovny.cvut.cz/vychova/vskp.html | eng |
dc.rights | Vysokoškolská závěrečná práce je dílo chráněné autorským zákonem. Je možné pořizovat z něj na své náklady a pro svoji osobní potřebu výpisy, opisy a rozmnoženiny. Jeho využití musí být v souladu s autorským zákonem http://www.mkcr.cz/assets/autorske-pravo/01-3982006.pdf a citační etikou http://knihovny.cvut.cz/vychova/vskp.html | cze |
dc.subject | deep reinforcement learning | cze |
dc.subject | Isaac Sim | cze |
dc.subject | Nvidia | cze |
dc.subject | quadruped legged robotics | cze |
dc.subject | Artaban robotic platform | cze |
dc.subject | Cardan mechanism | cze |
dc.subject | parallel mechanism | cze |
dc.subject | PPO | cze |
dc.subject | Proximal policy optimization | cze |
dc.subject | Actor-Critic methods | cze |
dc.subject | deep reinforcement learning | eng |
dc.subject | Isaac Sim | eng |
dc.subject | Nvidia | eng |
dc.subject | quadruped legged robotics | eng |
dc.subject | Artaban robotic platform | eng |
dc.subject | Cardan mechanism | eng |
dc.subject | parallel mechanism | eng |
dc.subject | PPO | eng |
dc.subject | Proximal policy optimization | eng |
dc.subject | Actor-Critic methods | eng |
dc.title | Posilované učení pro ovládání čtyřnohého robota s inovativní kinematikou | cze |
dc.title | Reinforcement Learning for Quadrupedal Robot Control with Novel Kinematics | eng |
dc.type | diplomová práce | cze |
dc.type | master thesis | eng |
dc.contributor.referee | Szadkowski Rudolf Jakub | |
theses.degree.grantor | katedra kybernetiky | cze |
theses.degree.programme | Kybernetika a robotika | cze |
Soubory tohoto záznamu
Tento záznam se objevuje v následujících kolekcích
-
Diplomové práce - 13133 [495]