Publication Details

SELECT * FROM publications WHERE Record_Number=11208
Reference TypeJournal Article
Author(s)Liu, Z.
Year2019
TitleLocal Online Motor Babbling: Learning Motor Abundance of A Musculoskeletal Robot Arm
Journal/Conference/Book TitleMaster Thesis
AbstractSensorimotor learning of continuum or soft robots has been a challenging problem due to the highly redundant, non- linear system with hysteresis. Recent works in goal babbling have demonstrated successful learning of inverse kinematics (IK) on such systems and suggest that babbling in the goal space better resolves motor redundancy by learning as few sensorimotor mapping as possible. However, for the musculoskeletal robot, which is a hard-bodied system with soft actuation, motor redundancy can be of useful information to explain muscle activation patterns, thus the term motor abundance. This thesis aims to learn the IK and motor abundance of a 10 degree-of-freedom (DoF) bio-inspired upper limb robot actuated by 24 pneumatic artificial muscles (PAMs), which is a highly redundant and over-actuated musculoskeletal system with an unknown task space. Firstly some simple heuristics are introduced to empirically estimate the unknown task space, so as to facilitate IK learning using directed online goal babbling. The results show that the learned IK is able to achieve 1.8 cm average accuracy given the best possible average is 1.2 cm. We then further propose local online motor babbling using the Covariance Matrix Adaptation Evolution Strategy (CMA-ES), which bootstraps on the collected samples in goal babbling for initialization, such that motor abundance can be queried for any static goal within the defined task space. The result shows that our motor babbling approach can efficiently explore motor abundance, and gives useful insights in terms of muscle stiffness and synergy.
Link to PDFhttps://www.ias.informatik.tu-darmstadt.de/uploads/Team/SvenjaStark/Liu_MSc_Thesis_2019.pdf

  

zum Seitenanfang