Human motion has been analyzed for decades based on experimentally collected subject data, serving various purposes, from enhancing athletic performance to assisting patients' recovery in rehabilitation and many individuals can benefit significantly from study advancements. Human motion prediction, is a more challenging task because no experimental data are available in advance, particularly concerning repetitive tasks, such as box lifting and tossing, to prevent injury risks. Tossing, a common task in various industries, involves the simultaneous vertical and horizontal movement of objects but often results in bodily strain. This paper presents an optimization-based method for predicting two-dimensional (2D) symmetric tossing motion without relying on experimental data. The method employs sequential quadratic programming, which optimizes dynamic effort by incorporating both static and dynamic joint torque limits. To validate the proposed model, experimental data were collected from 10 subjects performing tossing tasks using a motion capture system and force plates. The predicted joint angles and ground reaction forces considering dynamic joint strength constraints were compared with their corresponding experimental data to validate the model. In addition, the predicted joint torques differences are compared between joint dynamics strengths and static strengths. The results showed that the predicted optimal tossing motions range between the maximum and minimum of the experimental standard deviation for kinematic data across all subjects and the ground reaction forces are also within the experimental data range. This supports the validity of the prediction model. The findings of this study could have practical applications, especially in preventing the potential risk of injuries among workers who have daily tossing jobs.
Read full abstract