BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//wp-events-plugin.com//7.3.5//EN
TZID:Europe/Paris
X-WR-TIMEZONE:Europe/Paris
BEGIN:VEVENT
UID:584@cran.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20241213T100000
DTEND;TZID=Europe/Paris:20241213T123000
DTSTAMP:20241107T134839Z
URL:https://www.cran.univ-lorraine.fr/events/soutenance-de-these-abir-boua
 ouda/
SUMMARY:Soutenance de these Abir BOUAOUDA
DESCRIPTION:Titre de la these : Reinforcement learning for controlling cabl
 e driven parallel robots.\n\nRapporteurs :\nJacques GANGLOFF : Université
  de Strasbourg – ICUBE\nAbdel-Illah MOUADDIB : Université de Caen – C
 REYC\n\nExaminateurs :\nOuiddad LABBANI-IGBIDA : Université de Limoges 
 – XLIM\nLaeticia MATIGNON : Université Claude Bernard Lyon 1 – Polyte
 ch / LIRIS\n\nDirecteurs – Co-encadrants de thèse :\nDominique MARTINEZ
  : Université Aix - Marseille – ISM\nMohamed BOUTAYEB : Université de 
 Lorraine – CRAN\nRémi PANNEQUIN : Université de Lorraine – CRAN\nFra
 nçois CHARPILLET : INRIA - Nancy\n\nAbstract: In this thesis\, we present
  a novel reinforcement learning-based control strategy designed specifical
 ly for cable-driven parallel robots. The development of this controller be
 gins with the derivation of the dynamic model of the robot available in ou
 r laboratory. This dynamic model serves as a testing platform for training
  the controller\, with validation being conducted against real data obtain
 ed using the existing PID-Type controller. Given our focus on addressing t
 he trajectory tracking problem\, we devised a methodology for generating t
 raining trajectories that ensure comprehensive coverage of all robot state
 s. Subsequently\, we designed a reward function that factors in considerat
 ions such as tracking error and other relevant metrics. One of the key cha
 llenges we encountered pertained to managing the action space – maintain
 ing cable tension while operating the end effector. Since our action space
  is defined by motor speed\, direct limitation of cable tension was not fe
 asible. Thus\, we formulated a strategy to calculate the current at each s
 tep and verify that it falls within predefined limits. Employing this setu
 p in conjunction with prominent reinforcement learning algorithms tailored
  for continuous spaces such as DDPG\, PPO\, and SAC\, we proceeded to trai
 n our controller. A comprehensive comparison was conducted among the three
  algorithms across various performance metrics. Given the extensive traini
 ng duration\, we introduced a novel approach leveraging our prior knowledg
 e of the robot to streamline the training process. The controller underwen
 t testing on the physical robot\, yielding results that demonstrate its ca
 pability to achieve precise trajectory tracking. Subsequently\, we contras
 ted the outcomes obtained from the three reinforcement learning algorithms
  with those derived from a PID-based controller\, analyzing factors includ
 ing tracking error\, energy efficiency\, and robustness. This comparative 
 evaluation provided insights into the strengths and weaknesses of each app
 roach\, shedding light on the efficacy of the proposed reinforcement learn
 ing-based control strategy for cable-driven parallel robots.
CATEGORIES:Département CID,Soutenances thèses et HDR
END:VEVENT
BEGIN:VTIMEZONE
TZID:Europe/Paris
X-LIC-LOCATION:Europe/Paris
BEGIN:STANDARD
DTSTART:20241027T020000
TZOFFSETFROM:+0200
TZOFFSETTO:+0100
TZNAME:CET
END:STANDARD
END:VTIMEZONE
END:VCALENDAR