Skip to content
DraconiticResearch
Search Sources API
203,234 studies31 sourcesUpdated 3 min ago
Peer-reviewedCC BY 4.0PLOS One

Meta reinforcement learning for robust quadruped locomotion under joint failures

  • Likai Wu1
  • Chee-Onn Chow2
  • Wei Ru Wong3
  • Joon Huang Chuah
  • Jeevan Kanesan
  • Meina Zhang
  1. 1Department of Electrical Engineering, Faculty of Engineering, Universiti Malaya, Kuala Lumpur, Malaysia
  2. 2School of Electrical Engineering, Shandong Huayu University of Technology, Dezhou, China
  3. 3Department of Naval Architecture and Ocean Engineering, Pusan National University, Busan, South Korea
Open at sourcePDFDOICite

Abstract

Joint failures can disrupt gait coordination, velocity tracking, and body stability in quadruped robots by reducing the available control authority. This study investigates faults in which joints remain locked at fixed angles and presents Adjudicate Reconfiguration and Evolving Regularization Meta Reinforcement Learning (ARER-MetaRL). The framework combines task inference from context, modular routing tailored to each task through an Adjudicate Reconfiguration (AR) network, and Evolving Gradient Regularization (ER) to support rapid adaptation. Tasks differ in the identities and locking angles of the faulty joints, the number of simultaneous faults, the commanded motion direction, and the terrain. Experiments in NVIDIA Isaac Sim cover all 12 actuated joints, with 0–3 stuck joints during meta-training and severity tests involving up to 6 simultaneously stuck joints. ARER-MetaRL achieves higher Mean Episode Linear Velocity Tracking Reward (MELV) and Mean Episode Roll–Pitch Tracking Reward (MERP) than single-task reinforcement learning, ablated variants, and the evaluated learning and adaptation baselines. Cumulative rewards further demonstrate the benefit of combining the proposed components. In the study of fault severity, ARER-MetaRL achieves MELV/MERP values of 0.86/0.84 with one stuck joint and 0.55/0.54 with six stuck joints, compared with 0.21/0.18 and 0.05/0.03, respectively, for the single-task baseline. These results show improved adaptation across different configurations of locked joints and terrain conditions.

Related

  1. Robust disturbance-aware actor–critic reinforcement learning for multi-DOF robotic manipulators
    Viet Ngu Nguyen, Thi Minh Tam Le, Duc Hung Pham +1
    PLOS OnePeer-reviewedCC BY 4.0Full textNeurosciencePLOS
  2. Reln Haploinsufficiency Enhances Fentanyl-Induced Locomotion and Dorsal Striatal Activity Without Affecting Opioid Reinforcement and Relapse-Like Behavior.
    Carl Litif, Avraham M Libster, Shane Desfor +4
    Genes, brain, and behaviorPeer-reviewedCC BYFull textNeuroscienceEurope PMC
  3. Robust and Quality-of-Life-Aware Treatment Protocols in NSCLC using Deep Reinforcement Learning
    L. R. Jansen-Storbacka, A.-M. C. Dingemans, K. Stankova +2
    bioRxivPreprintCC BY-NC-NDFull textSystems & synthetic biologybioRxiv
  4. Robust pan-junctional reinforcement preserves the gut epithelial barrier under mechanical stress
    Vishnu Krishnakumar, Amalia Ferran, Sandra Bernat-Fabre +4
    Science AdvancesPeer-reviewedCC BYFull textCell biologySci. Adv.CrossrefEurope PMCEurope PMC+1
  5. Reliability-based arbitration reveals heterogeneous model-based and model-free reinforcement learning strategies linked to compulsivity and impulsivity
    Jeffrey Cockburn, Weilun Ding, Sarah Oh +4
    Nature CommunicationsPeer-reviewedCC BY-NC-ND 4.0Full textNeuroscienceCrossref

Cite

BibTeXRIS
BibTeX
@article{wu2026meta,
  title = {{Meta reinforcement learning for robust quadruped locomotion under joint failures}},
  author = {Likai Wu and Chee-Onn Chow and Wei Ru Wong and Joon Huang Chuah and Jeevan Kanesan and Meina Zhang},
  journal = {PLOS One},
  year = {2026},
  month = oct,
  doi = {10.1371/journal.pone.0361099},
  url = {https://doi.org/10.1371/journal.pone.0361099},
  note = {CC BY 4.0},
}
RIS
TY  - JOUR
TI  - Meta reinforcement learning for robust quadruped locomotion under joint failures
AU  - Likai Wu
AU  - Chee-Onn Chow
AU  - Wei Ru Wong
AU  - Joon Huang Chuah
AU  - Jeevan Kanesan
AU  - Meina Zhang
PY  - 2026
DA  - 2026/10/16
JO  - PLOS One
DO  - 10.1371/journal.pone.0361099
UR  - https://doi.org/10.1371/journal.pone.0361099
AB  - Joint failures can disrupt gait coordination, velocity tracking, and body stability in quadruped robots by reducing the available control authority. This study investigates faults in which joints remain locked at fixed angles and presents Adjudicate Reconfiguration and Evolving Regularization Meta Reinforcement Learning (ARER-MetaRL). The framework combines task inference from context, modular routing tailored to each task through an Adjudicate Reconfiguration (AR) network, and Evolving Gradient Regularization (ER) to support rapid adaptation. Tasks differ in the identities and locking angles of the faulty joints, the number of simultaneous faults, the commanded motion direction, and the terrain. Experiments in NVIDIA Isaac Sim cover all 12 actuated joints, with 0–3 stuck joints during meta-training and severity tests involving up to 6 simultaneously stuck joints. ARER-MetaRL achieves higher Mean Episode Linear Velocity Tracking Reward (MELV) and Mean Episode Roll–Pitch Tracking Reward (MERP) than single-task reinforcement learning, ablated variants, and the evaluated learning and adaptation baselines. Cumulative rewards further demonstrate the benefit of combining the proposed components. In the study of fault severity, ARER-MetaRL achieves MELV/MERP values of 0.86/0.84 with one stuck joint and 0.55/0.54 with six stuck joints, compared with 0.21/0.18 and 0.05/0.03, respectively, for the single-task baseline. These results show improved adaptation across different configurations of locked joints and terrain conditions.
SN  - 1932-6203
ER  -