본문

서브메뉴

The Robot Cerebellum: Toward Safe, Agile and Intelligent Legged Robotics
The Robot Cerebellum: Toward Safe, Agile and Intelligent Legged Robotics
The Robot Cerebellum: Toward Safe, Agile and Intelligent Legged Robotics

Detailed Information

자료유형  
 학위논문 서양
최종처리일시  
20260202103606
ISBN  
9798288862601
DDC  
629.8
저자명  
Li, Zhongyu.
서명/저자  
The Robot Cerebellum: Toward Safe, Agile and Intelligent Legged Robotics
발행사항  
[Sl] : University of California, Berkeley, 2025
발행사항  
Ann Arbor : ProQuest Dissertations & Theses, 2025
형태사항  
403 p
주기사항  
Source: Dissertations Abstracts International, Volume: 87-01, Section: B.
주기사항  
Advisor: Sreenath, Koushil.
학위논문주기  
Thesis (Ph.D.)--University of California, Berkeley, 2025.
초록/해제  
요약This dissertation develops a unified framework for building safe, agile, and intelligent legged robotic systems, inspired by the concept of the Robot Cerebellum which is the part of the brain responsible for body coordination and motion control. The core challenge addressed is how to enable versatile and dynamic maneuvers, i.e., to develop the cerebellum, in complex and underactuated robotic systems such as legged robots by integrating model-based optimal control and model-free reinforcement learning (RL). The work begins with structured motion control, introducing model-based locomotion frameworks to achieve expressive and robust locomotion. To unlock the full potential and agility of legged robots, this dissertation then presents a novel general locomotion control framework using model-free RL on a person-sized bipedal robot in the real world. This framework demonstrates a range of novel bipedal locomotion capabilities, including walking with compliance and persistent performance over a long timespan (1.5 years), running (including a 400-meter dash), and multi-axis targeted jumping such as standing long jumps and table jumps, with significant robustness and adaptivity. Building on this foundation, the dissertation introduces HiLMa, a hierarchical framework for whole-body loco-manipulation, enabling legged robots to function beyond locomotion and perform complex object interaction tasks using their legs as arms while in locomotion. It further extends to agent interaction, exploring both centralized and decentralized coordination strategies for human-robot and multi-robot collaboration in congested environments. Finally, the dissertation addresses safety by bridging model-based motion planning with learning-based controllers. Through system identification, it unveils hidden low-dimensional linear behavior of the closed-loop systems using RL policies, which is then embedded into an model-based framework to achieve certifiably safe navigation while preserving learned agility. Together, these contributions advance the state of the art across locomotion control, planning, learning, safety, and physical interaction using legged robots in real-world settings, paving the way to extend the cerebellar core into a broader Robot Brain for future robotics across a wide range of real-world tasks and environments.
일반주제명  
Robotics
일반주제명  
Computer science
일반주제명  
Electrical engineering
키워드  
Legged robotics
키워드  
Robot learning
키워드  
Reinforcement learning
키워드  
Robotic systems
키워드  
Robot Cerebellum
기타저자  
University of California, Berkeley Mechanical Engineering
기본자료저록  
Dissertations Abstracts International. 87-01B.
전자적 위치 및 접속  
로그인 후 원문을 볼 수 있습니다.

MARC

 008260126s2025        us                              c    eng  d
■001000017357834
■00520260202103606
■006m          o    d                
■007cr#unu||||||||
■020    ▼a9798288862601
■035    ▼a(MiAaPQ)AAI32042758
■040    ▼aMiAaPQ▼cMiAaPQ
■0820  ▼a629.8
■1001  ▼aLi,  Zhongyu.
■24510▼aThe  Robot  Cerebellum:  Toward  Safe,  Agile  and  Intelligent  Legged  Robotics
■260    ▼a[Sl]▼bUniversity  of  California,  Berkeley▼c2025
■260  1▼aAnn  Arbor▼bProQuest  Dissertations  &  Theses▼c2025
■300    ▼a403  p
■500    ▼aSource:  Dissertations  Abstracts  International,  Volume:  87-01,  Section:  B.
■500    ▼aAdvisor:  Sreenath,  Koushil.
■5021  ▼aThesis  (Ph.D.)--University  of  California,  Berkeley,  2025.
■520    ▼aThis  dissertation  develops  a  unified  framework  for  building  safe,  agile,  and  intelligent  legged  robotic  systems,  inspired  by  the  concept  of  the  Robot  Cerebellum  which  is  the  part  of  the  brain  responsible  for  body  coordination  and  motion  control.  The  core  challenge  addressed  is  how  to  enable  versatile  and  dynamic  maneuvers,  i.e.,  to  develop  the  cerebellum,  in  complex  and  underactuated  robotic  systems  such  as  legged  robots  by  integrating  model-based  optimal  control  and  model-free  reinforcement  learning  (RL).  The  work  begins  with  structured  motion  control,  introducing  model-based  locomotion  frameworks  to  achieve  expressive  and  robust  locomotion.  To  unlock  the  full  potential  and  agility  of  legged  robots,  this  dissertation  then  presents  a  novel  general  locomotion  control  framework  using  model-free  RL  on  a  person-sized  bipedal  robot  in  the  real  world.  This  framework  demonstrates  a  range  of  novel  bipedal  locomotion  capabilities,  including  walking  with  compliance  and  persistent  performance  over  a  long  timespan  (1.5  years),  running  (including  a  400-meter  dash),  and  multi-axis  targeted  jumping  such  as  standing  long  jumps  and  table  jumps,  with  significant  robustness  and  adaptivity.  Building  on  this  foundation,  the  dissertation  introduces  HiLMa,  a  hierarchical  framework  for  whole-body  loco-manipulation,  enabling  legged  robots  to  function  beyond  locomotion  and  perform  complex  object  interaction  tasks  using  their  legs  as  arms  while  in  locomotion.  It  further  extends  to  agent  interaction,  exploring  both  centralized  and  decentralized  coordination  strategies  for  human-robot  and  multi-robot  collaboration  in  congested  environments.  Finally,  the  dissertation  addresses  safety  by  bridging  model-based  motion  planning  with  learning-based  controllers.  Through  system  identification,  it  unveils  hidden  low-dimensional  linear  behavior  of  the  closed-loop  systems  using  RL  policies,  which  is  then  embedded  into  an  model-based  framework  to  achieve  certifiably  safe  navigation  while  preserving  learned  agility.  Together,  these  contributions  advance  the  state  of  the  art  across  locomotion  control,  planning,  learning,  safety,  and  physical  interaction  using  legged  robots  in  real-world  settings,  paving  the  way  to  extend  the  cerebellar  core  into  a  broader  Robot  Brain  for  future  robotics  across  a  wide  range  of  real-world  tasks  and  environments.
■590    ▼aSchool  code:  0028.
■650  4▼aRobotics
■650  4▼aComputer  science
■650  4▼aElectrical  engineering
■653    ▼aLegged  robotics
■653    ▼aRobot  learning
■653    ▼aReinforcement  learning
■653    ▼aRobotic  systems
■653    ▼aRobot  Cerebellum
■690    ▼a0771
■690    ▼a0984
■690    ▼a0544
■71020▼aUniversity  of  California,  Berkeley▼bMechanical  Engineering.
■7730  ▼tDissertations  Abstracts  International▼g87-01B.
■790    ▼a0028
■791    ▼aPh.D.
■792    ▼a2025
■793    ▼aEnglish
■85640▼uhttp://www.riss.kr/pdu/ddodLink.do?id=T17357834▼nKERIS▼z이  자료의  원문은  한국교육학술정보원에서  제공합니다.

Preview

Export

ChatGPT Discussion

AI Recommended Related Books


    New Books MORE
    Statistics for the past 3 years. Go to brief

    Подробнее информация.

    • Бронирование
    • не существует
    • моя папка
    • Первый запрос зрения
    • Non-Book Loan Application
    • Nighttime Book Loan Application
    материал
    Reg No. Количество платежных Местоположение статус Ленд информации
    TF14701 전자도서 대출가능 My Folder 부재도서신고 비도서대출신청 야간 도서대출신청

    * Бронирование доступны в заимствований книги. Чтобы сделать предварительный заказ, пожалуйста, нажмите кнопку бронирование

    Books borrowed together with this book

    Related Popular Books

    Available after logging in.