In the execution of repetitive tasks, humans can capitalize on experience to improve their motor performance. Prominent examples of this ability can be recognized in our capacity of grasping and manipulating in uncertain conditions. With the aim of providing a mathematical description for such behavior, experiments are considered where participants are required to lift an object with an unexpected mass distribution. By repeating multiple times the same lifting action, participants can learn the correct motor command for task accomplishment. Three models are proposed that combine reactive terms and a learned anticipatory action to explain experimental data. The models feature intratrial and intertrial memory, and the effect of slowly and fast adaptive sensory receptors. The architectures’ effectiveness in explaining experimental data is compared with a general‐purpose state of the art model. The proposed algorithms conspicuously outperform the state of the art in all the considered validation routines. Global and within‐trial human behavior is predicted with 88% of accuracy in nominal conditions. When the object's center of mass is moved, the accuracy is maintained up to 83%. Finally, convergence properties of proposed algorithms are analytically discussed, and their stability and robustness against measurement noise are evaluated in simulation.

Modeling Previous Trial Effect in Human Manipulation through Iterative Learning Control

Garabini, Manolo;Bianchi, Matteo;Bicchi, Antonio
2020-01-01

Abstract

In the execution of repetitive tasks, humans can capitalize on experience to improve their motor performance. Prominent examples of this ability can be recognized in our capacity of grasping and manipulating in uncertain conditions. With the aim of providing a mathematical description for such behavior, experiments are considered where participants are required to lift an object with an unexpected mass distribution. By repeating multiple times the same lifting action, participants can learn the correct motor command for task accomplishment. Three models are proposed that combine reactive terms and a learned anticipatory action to explain experimental data. The models feature intratrial and intertrial memory, and the effect of slowly and fast adaptive sensory receptors. The architectures’ effectiveness in explaining experimental data is compared with a general‐purpose state of the art model. The proposed algorithms conspicuously outperform the state of the art in all the considered validation routines. Global and within‐trial human behavior is predicted with 88% of accuracy in nominal conditions. When the object's center of mass is moved, the accuracy is maintained up to 83%. Finally, convergence properties of proposed algorithms are analytically discussed, and their stability and robustness against measurement noise are evaluated in simulation.
2020
Cenceschi, Lorenzo; Della Santina, Cosimo; Averta, Giuseppe; Garabini, Manolo; Fu, Qiushi; Santello, Marco; Bianchi, Matteo; Bicchi, Antonio
File in questo prodotto:
File Dimensione Formato  
BianchiMatteo_1050013.pdf

accesso aperto

Tipologia: Versione finale editoriale
Licenza: Creative commons
Dimensione 3.3 MB
Formato Adobe PDF
3.3 MB Adobe PDF Visualizza/Apri

I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.

Utilizza questo identificativo per citare o creare un link a questo documento: https://hdl.handle.net/11568/1050013
Citazioni
  • ???jsp.display-item.citation.pmc??? ND
  • Scopus ND
  • ???jsp.display-item.citation.isi??? 1
social impact