In this study, a hierarchical structure is proposed to model human movement control during sit-to-stand transfer. At the highest level the desired movement is planned. Then, the task to be performed is decomposed to its constitutive sub-tasks. To decompose the sit-to-stand movement, the spatial trajectory of the body center of mass is automatically approximated by partially linearized trajectories. Each linearized part defines a sub-task. At the second level, corresponding to each sub-task a module is developed that learns to control the movement during the performance of that sub-task. Since the procedure of decomposition is performed automatically, the number of modules and assessment of suitable data to train the modules are also determined automatically. This feature is one of the main differences between the proposed structure and the MOdular Selection And Identification for Control (MOSAIC) structure [M. Haruno, D.M. Wolpert, M. Kawato, MOSAIC model for sensorimotor learning and control, Neural Computation 13 (2001) 2201–2220.]. Our proposed model is in conformity with the recent physiological and neurobehavioral findings and provides a framework for examining a given movement under different conditions.
AMA-MOSAICI: An automatic module assigning hierarchical structure to control human motion based on movement decomposition
EMADI ANDANI, Mehran;
2009-01-01
Abstract
In this study, a hierarchical structure is proposed to model human movement control during sit-to-stand transfer. At the highest level the desired movement is planned. Then, the task to be performed is decomposed to its constitutive sub-tasks. To decompose the sit-to-stand movement, the spatial trajectory of the body center of mass is automatically approximated by partially linearized trajectories. Each linearized part defines a sub-task. At the second level, corresponding to each sub-task a module is developed that learns to control the movement during the performance of that sub-task. Since the procedure of decomposition is performed automatically, the number of modules and assessment of suitable data to train the modules are also determined automatically. This feature is one of the main differences between the proposed structure and the MOdular Selection And Identification for Control (MOSAIC) structure [M. Haruno, D.M. Wolpert, M. Kawato, MOSAIC model for sensorimotor learning and control, Neural Computation 13 (2001) 2201–2220.]. Our proposed model is in conformity with the recent physiological and neurobehavioral findings and provides a framework for examining a given movement under different conditions.
I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
