Kinematic and kinetic evaluation of a throwing model using motion primitives

Joo H. Kim, Dustyn P. Roberts, John Meusch, Salam Rahmatalla

Research output: Contribution to journalArticlepeer-review


A human model of a dynamic overarm throwing motion is evaluated in this work with novel motion primitives including determinants, key frames, and ground reaction force (GRF) profiles. The kinematic and kinetic data of overarm throwing from five subjects are used for motion primitives. The evaluation process uses combined qualitative key frames (picture-based) and quantitative analyses to capture the differences in the experimental and generated throwing motion. The model shows potential application in simulating human-like throwing strategies and humanoid motion planning and control. Significant correlation between the generated and experimental GRFs with R2 = 0.90 and 0.82 were found for the right and left feet, respectively, and the total vertical GRF calculated from the model was approximately equal to the weight of the avatars. The generated right shoulder flexion, shoulder rotation, and right elbow flexion showed near perfect correlation with R2 values larger than 0.90, while the right shoulder abduction, left hip abduction, and left hip rotation showed moderate correlations with R2 values between 0.25 and 0.55. Due to strategy differences, the generated left hip flexion and left knee flexion showed weak correlations with R2 values equal to 0.17 and 0.03, respectively.

Original languageEnglish (US)
Pages (from-to)98-109
Number of pages12
JournalInternational Journal of Robotics and Automation
Issue number1
StatePublished - Jan 1 2015


  • Inverse kinematics
  • Key frames
  • Motion determinants
  • Motion primitives
  • Optimization
  • Overarm throw

ASJC Scopus subject areas

  • Software
  • Control and Systems Engineering
  • Modeling and Simulation
  • Mechanical Engineering
  • Electrical and Electronic Engineering
  • Artificial Intelligence


Dive into the research topics of 'Kinematic and kinetic evaluation of a throwing model using motion primitives'. Together they form a unique fingerprint.

Cite this