TY - GEN
T1 - Cognitive-based terminal state prediction for human motion planning
AU - Howard, Brad
AU - Yang, James
AU - Yang, Guolai
PY - 2013
Y1 - 2013
N2 - Every day, people are presented with tasks that are completed with very little mechanical effort such as turning a door knob, turning a screw driver, and grabbing a cup to move it to a new location and/or orientation. These tasks are often overlooked in the mechanical study of human movement due to the fact they carry with them very little biomechanical costs or effort. However, from a cognitive standpoint, these tasks carry high complexity. For example, the simple task of grabbing a cup and flipping it over is very easy mechanically and can be effortlessly achieved by a two year old. However, it is impossible to predict or simulate this motion without major intervention in the form of explicit constraints defining the task. The model itself cannot decide in which orientation the hand should assume in order to grasp the object. Also, it cannot decide where on the object the hand should be placed. These aspects must be assumed by the researcher and constrained in the formulation. In other words, digital human models, such as optimization-based motion prediction models, are unable to plan actions. This implies that for each task, a unique optimization formulation is needed in order to predict the motion/posture needed to complete each task. This paper presents a new method for task planning prediction within optimization based posture and motion prediction. It provides a new single optimization formulation that allows for the prediction of multiple unique manual manipulation tasks. The method is based on observations made from experimental studies on cognitive motor planning.
AB - Every day, people are presented with tasks that are completed with very little mechanical effort such as turning a door knob, turning a screw driver, and grabbing a cup to move it to a new location and/or orientation. These tasks are often overlooked in the mechanical study of human movement due to the fact they carry with them very little biomechanical costs or effort. However, from a cognitive standpoint, these tasks carry high complexity. For example, the simple task of grabbing a cup and flipping it over is very easy mechanically and can be effortlessly achieved by a two year old. However, it is impossible to predict or simulate this motion without major intervention in the form of explicit constraints defining the task. The model itself cannot decide in which orientation the hand should assume in order to grasp the object. Also, it cannot decide where on the object the hand should be placed. These aspects must be assumed by the researcher and constrained in the formulation. In other words, digital human models, such as optimization-based motion prediction models, are unable to plan actions. This implies that for each task, a unique optimization formulation is needed in order to predict the motion/posture needed to complete each task. This paper presents a new method for task planning prediction within optimization based posture and motion prediction. It provides a new single optimization formulation that allows for the prediction of multiple unique manual manipulation tasks. The method is based on observations made from experimental studies on cognitive motor planning.
KW - Digital human
KW - Ground reaction forces
KW - Multiobjective optimization
KW - Zero moment point
UR - http://www.scopus.com/inward/record.url?scp=84896978923&partnerID=8YFLogxK
U2 - 10.1115/DETC2013-13440
DO - 10.1115/DETC2013-13440
M3 - Conference contribution
AN - SCOPUS:84896978923
SN - 9780791855898
T3 - Proceedings of the ASME Design Engineering Technical Conference
BT - 39th Design Automation Conference
PB - American Society of Mechanical Engineers
T2 - ASME 2013 International Design Engineering Technical Conferences and Computers and Information in Engineering Conference, IDETC/CIE 2013
Y2 - 4 August 2013 through 7 August 2013
ER -